diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -3,16023 +3,32023 @@
   "best_model_checkpoint": null,
   "epoch": 1.0,
   "eval_steps": 500,
-  "global_step": 1000,
+  "global_step": 2000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "clip_ratio": 0.0,
-      "completion_length": 503.75,
-      "epoch": 0.001,
-      "grad_norm": 2.9832106604609847,
+      "completion_length": 328.70833587646484,
+      "epoch": 0.0005,
+      "grad_norm": 6.251723469874294,
       "kl": 0.0,
-      "learning_rate": 1e-08,
-      "loss": -0.0412,
-      "reward": 1.165066421031952,
-      "reward_std": 0.4164372831583023,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.6388888955116272,
-      "rewards/repetition_penalty_reward": -0.03632250800728798,
-      "rewards/tag_count_reward": 0.5208333730697632,
+      "learning_rate": 5e-09,
+      "loss": -0.0315,
+      "reward": 0.6384468078613281,
+      "reward_std": 0.2296978384256363,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.1597222313284874,
+      "rewards/repetition_penalty_reward": -0.05252542719244957,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 1
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 601.9375305175781,
-      "epoch": 0.002,
-      "grad_norm": 2.6050906286301534,
+      "completion_length": 333.6666717529297,
+      "epoch": 0.001,
+      "grad_norm": 4.439731592620521,
       "kl": 0.0,
-      "learning_rate": 2e-08,
-      "loss": -0.0073,
-      "reward": 1.1193318367004395,
-      "reward_std": 0.2575419098138809,
+      "learning_rate": 1e-08,
+      "loss": -0.0813,
+      "reward": 0.6252729892730713,
+      "reward_std": 0.29075783491134644,
       "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.659722238779068,
-      "rewards/repetition_penalty_reward": -0.0508070383220911,
-      "rewards/tag_count_reward": 0.5104166865348816,
+      "rewards/reasoning_steps_reward": 0.194444440305233,
+      "rewards/repetition_penalty_reward": -0.06917147152125835,
+      "rewards/tag_count_reward": 0.5,
       "step": 2
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 624.4583740234375,
-      "epoch": 0.003,
-      "grad_norm": 2.4392990354195887,
-      "kl": 0.00021076202392578125,
-      "learning_rate": 3e-08,
-      "loss": 0.0166,
-      "reward": 1.1063403487205505,
-      "reward_std": 0.29120244085788727,
+      "completion_length": 315.2291717529297,
+      "epoch": 0.0015,
+      "grad_norm": 4.648660461286372,
+      "kl": 0.00014925003051757812,
+      "learning_rate": 1.5e-08,
+      "loss": -0.0918,
+      "reward": 0.578189492225647,
+      "reward_std": 0.19131075590848923,
       "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.6597222089767456,
-      "rewards/repetition_penalty_reward": -0.0533819030970335,
+      "rewards/reasoning_steps_reward": 0.1319444552063942,
+      "rewards/repetition_penalty_reward": -0.05375497601926327,
       "rewards/tag_count_reward": 0.5,
       "step": 3
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 505.97918701171875,
-      "epoch": 0.004,
-      "grad_norm": 2.6745634307308377,
-      "kl": 0.000171661376953125,
-      "learning_rate": 4e-08,
-      "loss": -0.0265,
-      "reward": 1.1932607293128967,
-      "reward_std": 0.3999044597148895,
-      "rewards/accuracy_reward": 0.06250000186264515,
-      "rewards/reasoning_steps_reward": 0.6527777910232544,
-      "rewards/repetition_penalty_reward": -0.04805881343781948,
-      "rewards/tag_count_reward": 0.5260416865348816,
+      "completion_length": 391.8333435058594,
+      "epoch": 0.002,
+      "grad_norm": 3.069933909424734,
+      "kl": 0.00018310546875,
+      "learning_rate": 2e-08,
+      "loss": -0.0881,
+      "reward": 0.6503687500953674,
+      "reward_std": 0.2582377791404724,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.180555559694767,
+      "rewards/repetition_penalty_reward": -0.061436835676431656,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 4
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 604.6458435058594,
-      "epoch": 0.005,
-      "grad_norm": 2.7835442273591835,
-      "kl": 0.00020742416381835938,
-      "learning_rate": 5e-08,
-      "loss": -0.0375,
-      "reward": 1.0637533068656921,
-      "reward_std": 0.3286152780056,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.569444477558136,
-      "rewards/repetition_penalty_reward": -0.04735783860087395,
-      "rewards/tag_count_reward": 0.5208333432674408,
+      "completion_length": 415.8333435058594,
+      "epoch": 0.0025,
+      "grad_norm": 2.784498697916323,
+      "kl": 0.00012993812561035156,
+      "learning_rate": 2.5e-08,
+      "loss": -0.066,
+      "reward": 0.5441770553588867,
+      "reward_std": 0.21253511309623718,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.1527777835726738,
+      "rewards/repetition_penalty_reward": -0.10860074311494827,
+      "rewards/tag_count_reward": 0.5,
       "step": 5
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 487.22918701171875,
-      "epoch": 0.006,
-      "grad_norm": 3.0412926302974874,
-      "kl": 0.00020647048950195312,
-      "learning_rate": 6e-08,
-      "loss": -0.0389,
-      "reward": 1.0956225991249084,
-      "reward_std": 0.38825175166130066,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.604166716337204,
-      "rewards/repetition_penalty_reward": -0.039794087409973145,
-      "rewards/tag_count_reward": 0.5104166865348816,
+      "completion_length": 432.1041717529297,
+      "epoch": 0.003,
+      "grad_norm": 3.090569400029858,
+      "kl": 0.00017309188842773438,
+      "learning_rate": 3e-08,
+      "loss": -0.0504,
+      "reward": 0.7367339134216309,
+      "reward_std": 0.3796353191137314,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.2500000149011612,
+      "rewards/repetition_penalty_reward": -0.07576615735888481,
+      "rewards/tag_count_reward": 0.5208333432674408,
       "step": 6
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 613.5833435058594,
-      "epoch": 0.007,
-      "grad_norm": 2.444811156842584,
-      "kl": 0.00019121170043945312,
-      "learning_rate": 7e-08,
-      "loss": -0.013,
-      "reward": 1.1998452544212341,
-      "reward_std": 0.3531786799430847,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.6736111640930176,
-      "rewards/repetition_penalty_reward": -0.05189092084765434,
-      "rewards/tag_count_reward": 0.5364583730697632,
+      "completion_length": 390.5833435058594,
+      "epoch": 0.0035,
+      "grad_norm": 2.8425497031556843,
+      "kl": 0.00014519691467285156,
+      "learning_rate": 3.5e-08,
+      "loss": -0.0414,
+      "reward": 0.6315869688987732,
+      "reward_std": 0.1821143701672554,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.1875,
+      "rewards/repetition_penalty_reward": -0.06632974371314049,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 7
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 526.9166870117188,
-      "epoch": 0.008,
-      "grad_norm": 2.577759663416605,
-      "kl": 0.00021648406982421875,
-      "learning_rate": 8e-08,
-      "loss": -0.1157,
-      "reward": 1.1657525897026062,
-      "reward_std": 0.40865209698677063,
-      "rewards/accuracy_reward": 0.06250000186264515,
-      "rewards/reasoning_steps_reward": 0.597222238779068,
-      "rewards/repetition_penalty_reward": -0.040844724513590336,
-      "rewards/tag_count_reward": 0.5468750298023224,
+      "completion_length": 406.2083435058594,
+      "epoch": 0.004,
+      "grad_norm": 2.877982305967783,
+      "kl": 0.0001347064971923828,
+      "learning_rate": 4e-08,
+      "loss": -0.0406,
+      "reward": 0.6656659245491028,
+      "reward_std": 0.26930323243141174,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2638889104127884,
+      "rewards/repetition_penalty_reward": -0.0982230119407177,
+      "rewards/tag_count_reward": 0.5,
       "step": 8
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.5208435058594,
-      "epoch": 0.009,
-      "grad_norm": 2.8811327535844984,
-      "kl": 0.00019359588623046875,
-      "learning_rate": 9e-08,
-      "loss": -0.0489,
-      "reward": 1.1031688451766968,
-      "reward_std": 0.41312260925769806,
-      "rewards/accuracy_reward": 0.0833333358168602,
-      "rewards/reasoning_steps_reward": 0.506944477558136,
-      "rewards/repetition_penalty_reward": -0.0444006510078907,
-      "rewards/tag_count_reward": 0.5572916865348816,
+      "completion_length": 402.97918701171875,
+      "epoch": 0.0045,
+      "grad_norm": 2.796921444089777,
+      "kl": 0.00016069412231445312,
+      "learning_rate": 4.5e-08,
+      "loss": 0.0211,
+      "reward": 0.619064211845398,
+      "reward_std": 0.21635551750659943,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.1597222313284874,
+      "rewards/repetition_penalty_reward": -0.061491381376981735,
+      "rewards/tag_count_reward": 0.5,
       "step": 9
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 556.5208435058594,
-      "epoch": 0.01,
-      "grad_norm": 2.4760963797873985,
-      "kl": 0.00020551681518554688,
-      "learning_rate": 1e-07,
-      "loss": 0.1082,
-      "reward": 1.057469666004181,
-      "reward_std": 0.2826480269432068,
+      "completion_length": 428.52085876464844,
+      "epoch": 0.005,
+      "grad_norm": 2.732127974984369,
+      "kl": 0.00013709068298339844,
+      "learning_rate": 5e-08,
+      "loss": -0.0253,
+      "reward": 0.5867983400821686,
+      "reward_std": 0.2477174624800682,
       "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.604166716337204,
-      "rewards/repetition_penalty_reward": -0.057113731279969215,
-      "rewards/tag_count_reward": 0.5104166865348816,
+      "rewards/reasoning_steps_reward": 0.1875000149011612,
+      "rewards/repetition_penalty_reward": -0.10070168599486351,
+      "rewards/tag_count_reward": 0.5,
       "step": 10
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 505.10418701171875,
-      "epoch": 0.011,
-      "grad_norm": 2.7426745766245304,
-      "kl": 0.00023126602172851562,
-      "learning_rate": 1.0999999999999999e-07,
-      "loss": -0.0164,
-      "reward": 1.1937065124511719,
-      "reward_std": 0.39100518822669983,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.6736111342906952,
-      "rewards/repetition_penalty_reward": -0.042404673993587494,
-      "rewards/tag_count_reward": 0.5208333730697632,
+      "completion_length": 483.6041717529297,
+      "epoch": 0.0055,
+      "grad_norm": 2.556791644892623,
+      "kl": 0.00016498565673828125,
+      "learning_rate": 5.4999999999999996e-08,
+      "loss": -0.046,
+      "reward": 0.5901573300361633,
+      "reward_std": 0.25573817640542984,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.180555559694767,
+      "rewards/repetition_penalty_reward": -0.09039826691150665,
+      "rewards/tag_count_reward": 0.5,
       "step": 11
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 606.3333435058594,
-      "epoch": 0.012,
-      "grad_norm": 2.547656089396865,
-      "kl": 0.000209808349609375,
-      "learning_rate": 1.2e-07,
-      "loss": -0.1084,
-      "reward": 1.0744403004646301,
-      "reward_std": 0.3415149748325348,
+      "completion_length": 389.81251525878906,
+      "epoch": 0.006,
+      "grad_norm": 2.970269557370756,
+      "kl": 0.00014209747314453125,
+      "learning_rate": 6e-08,
+      "loss": -0.0726,
+      "reward": 0.5915650725364685,
+      "reward_std": 0.227043054997921,
       "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.6250000298023224,
-      "rewards/repetition_penalty_reward": -0.06097651459276676,
-      "rewards/tag_count_reward": 0.5104166865348816,
+      "rewards/reasoning_steps_reward": 0.1458333432674408,
+      "rewards/repetition_penalty_reward": -0.0594765804708004,
+      "rewards/tag_count_reward": 0.5052083432674408,
       "step": 12
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 624.8958740234375,
-      "epoch": 0.013,
-      "grad_norm": 2.5452173789968726,
-      "kl": 0.00021314620971679688,
-      "learning_rate": 1.3e-07,
-      "loss": -0.0677,
-      "reward": 1.0435009598731995,
-      "reward_std": 0.2849584221839905,
+      "completion_length": 425.5833435058594,
+      "epoch": 0.0065,
+      "grad_norm": 3.55225958110278,
+      "kl": 0.00015544891357421875,
+      "learning_rate": 6.5e-08,
+      "loss": -0.1838,
+      "reward": 0.5981267392635345,
+      "reward_std": 0.2597131133079529,
       "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.5902778208255768,
-      "rewards/repetition_penalty_reward": -0.057193491607904434,
-      "rewards/tag_count_reward": 0.5104166865348816,
+      "rewards/reasoning_steps_reward": 0.1527777835726738,
+      "rewards/repetition_penalty_reward": -0.05985940620303154,
+      "rewards/tag_count_reward": 0.5052083432674408,
       "step": 13
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 534.3125152587891,
-      "epoch": 0.014,
-      "grad_norm": 2.5568403376430475,
-      "kl": 0.000244140625,
-      "learning_rate": 1.4e-07,
-      "loss": -0.0267,
-      "reward": 1.033350259065628,
-      "reward_std": 0.36252032220363617,
+      "completion_length": 383.5,
+      "epoch": 0.007,
+      "grad_norm": 2.8654839696962204,
+      "kl": 0.00014352798461914062,
+      "learning_rate": 7e-08,
+      "loss": -0.0181,
+      "reward": 0.6581193804740906,
+      "reward_std": 0.33123770356178284,
       "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.5069444626569748,
-      "rewards/repetition_penalty_reward": -0.03609427623450756,
-      "rewards/tag_count_reward": 0.5208333730697632,
+      "rewards/reasoning_steps_reward": 0.1875,
+      "rewards/repetition_penalty_reward": -0.08146397396922112,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 14
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 633.8541870117188,
-      "epoch": 0.015,
-      "grad_norm": 2.4845137409121465,
-      "kl": 0.00019216537475585938,
-      "learning_rate": 1.5e-07,
-      "loss": -0.0353,
-      "reward": 1.1506143808364868,
-      "reward_std": 0.3063247799873352,
-      "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.6805556118488312,
-      "rewards/repetition_penalty_reward": -0.045566244050860405,
-      "rewards/tag_count_reward": 0.5156250298023224,
+      "completion_length": 452.35418701171875,
+      "epoch": 0.0075,
+      "grad_norm": 2.735812818011658,
+      "kl": 0.0001621246337890625,
+      "learning_rate": 7.5e-08,
+      "loss": 0.0199,
+      "reward": 0.7007659077644348,
+      "reward_std": 0.2738788276910782,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.2500000149011612,
+      "rewards/repetition_penalty_reward": -0.08048411272466183,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 15
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 570.9375152587891,
-      "epoch": 0.016,
-      "grad_norm": 2.7140477200825273,
-      "kl": 0.0002346038818359375,
-      "learning_rate": 1.6e-07,
-      "loss": -0.0089,
-      "reward": 1.0886216163635254,
-      "reward_std": 0.3162507861852646,
+      "completion_length": 538.1666717529297,
+      "epoch": 0.008,
+      "grad_norm": 2.4984072149490264,
+      "kl": 0.00014781951904296875,
+      "learning_rate": 8e-08,
+      "loss": 0.0239,
+      "reward": 0.5570363700389862,
+      "reward_std": 0.23343774676322937,
       "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.6388888955116272,
-      "rewards/repetition_penalty_reward": -0.06068398430943489,
-      "rewards/tag_count_reward": 0.5104166865348816,
+      "rewards/reasoning_steps_reward": 0.1527777835726738,
+      "rewards/repetition_penalty_reward": -0.09574145823717117,
+      "rewards/tag_count_reward": 0.5,
       "step": 16
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 585.2708740234375,
-      "epoch": 0.017,
-      "grad_norm": 2.4713266581431625,
-      "kl": 0.0001678466796875,
-      "learning_rate": 1.7000000000000001e-07,
-      "loss": -0.0658,
-      "reward": 1.1053802967071533,
-      "reward_std": 0.3452056348323822,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.631944477558136,
-      "rewards/repetition_penalty_reward": -0.06823080591857433,
-      "rewards/tag_count_reward": 0.5208333432674408,
+      "completion_length": 386.2916717529297,
+      "epoch": 0.0085,
+      "grad_norm": 3.2191965365951885,
+      "kl": 0.0001621246337890625,
+      "learning_rate": 8.500000000000001e-08,
+      "loss": -0.0334,
+      "reward": 0.5502374768257141,
+      "reward_std": 0.200188048183918,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.1250000074505806,
+      "rewards/repetition_penalty_reward": -0.07997088506817818,
+      "rewards/tag_count_reward": 0.5052083432674408,
       "step": 17
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 513.6458587646484,
-      "epoch": 0.018,
-      "grad_norm": 2.751825490594234,
-      "kl": 0.00019788742065429688,
-      "learning_rate": 1.8e-07,
-      "loss": -0.0599,
-      "reward": 1.0880020260810852,
-      "reward_std": 0.318176731467247,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.61111119389534,
-      "rewards/repetition_penalty_reward": -0.04394245892763138,
+      "completion_length": 387.35418701171875,
+      "epoch": 0.009,
+      "grad_norm": 3.501374136164966,
+      "kl": 0.00019168853759765625,
+      "learning_rate": 9e-08,
+      "loss": -0.0635,
+      "reward": 0.5605765283107758,
+      "reward_std": 0.15551594644784927,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.1319444514811039,
+      "rewards/repetition_penalty_reward": -0.07136795669794083,
       "rewards/tag_count_reward": 0.5,
       "step": 18
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 586.6250305175781,
-      "epoch": 0.019,
-      "grad_norm": 2.39448557038407,
-      "kl": 0.0001811981201171875,
-      "learning_rate": 1.8999999999999998e-07,
-      "loss": 0.0141,
-      "reward": 1.0959751605987549,
-      "reward_std": 0.43034467101097107,
-      "rewards/accuracy_reward": 0.06250000186264515,
-      "rewards/reasoning_steps_reward": 0.5486111640930176,
-      "rewards/repetition_penalty_reward": -0.05159439332783222,
-      "rewards/tag_count_reward": 0.5364583432674408,
+      "completion_length": 461.7083435058594,
+      "epoch": 0.0095,
+      "grad_norm": 2.8703850021404986,
+      "kl": 0.0001609325408935547,
+      "learning_rate": 9.499999999999999e-08,
+      "loss": 0.0089,
+      "reward": 0.6698747277259827,
+      "reward_std": 0.21975713968276978,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2638889104127884,
+      "rewards/repetition_penalty_reward": -0.09401418641209602,
+      "rewards/tag_count_reward": 0.5,
       "step": 19
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 652.1250305175781,
-      "epoch": 0.02,
-      "grad_norm": 2.0210344240592066,
-      "kl": 0.00016498565673828125,
-      "learning_rate": 2e-07,
-      "loss": -0.0102,
-      "reward": 1.2613900899887085,
-      "reward_std": 0.5556788444519043,
-      "rewards/accuracy_reward": 0.1250000037252903,
-      "rewards/reasoning_steps_reward": 0.6527777910232544,
-      "rewards/repetition_penalty_reward": -0.06847099214792252,
-      "rewards/tag_count_reward": 0.5520833432674408,
+      "completion_length": 434.0625,
+      "epoch": 0.01,
+      "grad_norm": 2.857781731829004,
+      "kl": 0.00016641616821289062,
+      "learning_rate": 1e-07,
+      "loss": -0.0404,
+      "reward": 0.6144264340400696,
+      "reward_std": 0.23488393425941467,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2013889029622078,
+      "rewards/repetition_penalty_reward": -0.08696247264742851,
+      "rewards/tag_count_reward": 0.5,
       "step": 20
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 534.3333587646484,
-      "epoch": 0.021,
-      "grad_norm": 2.723236250331092,
-      "kl": 0.000263214111328125,
-      "learning_rate": 2.0999999999999997e-07,
-      "loss": -0.0982,
-      "reward": 1.113393783569336,
-      "reward_std": 0.3082204759120941,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.6319444477558136,
-      "rewards/repetition_penalty_reward": -0.03938402608036995,
+      "completion_length": 453.7291717529297,
+      "epoch": 0.0105,
+      "grad_norm": 2.945964454449734,
+      "kl": 0.0001544952392578125,
+      "learning_rate": 1.0499999999999999e-07,
+      "loss": -0.028,
+      "reward": 0.6437437832355499,
+      "reward_std": 0.2618062347173691,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2222222313284874,
+      "rewards/repetition_penalty_reward": -0.07847847789525986,
       "rewards/tag_count_reward": 0.5,
       "step": 21
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 593.5208435058594,
-      "epoch": 0.022,
-      "grad_norm": 2.802437133759923,
-      "kl": 0.00021219253540039062,
-      "learning_rate": 2.1999999999999998e-07,
-      "loss": 0.025,
-      "reward": 1.141821563243866,
-      "reward_std": 0.30135248601436615,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.6527777910232544,
-      "rewards/repetition_penalty_reward": -0.042206283658742905,
-      "rewards/tag_count_reward": 0.5104166865348816,
+      "completion_length": 412.50001525878906,
+      "epoch": 0.011,
+      "grad_norm": 2.5290188531303004,
+      "kl": 0.00010466575622558594,
+      "learning_rate": 1.0999999999999999e-07,
+      "loss": -0.0099,
+      "reward": 0.6274481117725372,
+      "reward_std": 0.2595982700586319,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2361111268401146,
+      "rewards/repetition_penalty_reward": -0.10866303369402885,
+      "rewards/tag_count_reward": 0.5,
       "step": 22
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 569.9166870117188,
-      "epoch": 0.023,
-      "grad_norm": 2.513753582519348,
-      "kl": 0.00023412704467773438,
-      "learning_rate": 2.3e-07,
-      "loss": -0.0394,
-      "reward": 1.033397912979126,
-      "reward_std": 0.3224620223045349,
+      "completion_length": 435.1666717529297,
+      "epoch": 0.0115,
+      "grad_norm": 2.757687382488654,
+      "kl": 0.00016546249389648438,
+      "learning_rate": 1.15e-07,
+      "loss": 0.0177,
+      "reward": 0.6414158344268799,
+      "reward_std": 0.22706189006567,
       "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.5833333730697632,
-      "rewards/repetition_penalty_reward": -0.04993540979921818,
+      "rewards/reasoning_steps_reward": 0.2291666939854622,
+      "rewards/repetition_penalty_reward": -0.08775084465742111,
       "rewards/tag_count_reward": 0.5,
       "step": 23
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 538.8125305175781,
-      "epoch": 0.024,
-      "grad_norm": 2.5472422948610527,
-      "kl": 0.0001964569091796875,
-      "learning_rate": 2.4e-07,
-      "loss": -0.0033,
-      "reward": 1.2361397743225098,
-      "reward_std": 0.4402560144662857,
-      "rewards/accuracy_reward": 0.0833333358168602,
-      "rewards/reasoning_steps_reward": 0.6736111640930176,
-      "rewards/repetition_penalty_reward": -0.052054738625884056,
-      "rewards/tag_count_reward": 0.5312500298023224,
+      "completion_length": 385.7708435058594,
+      "epoch": 0.012,
+      "grad_norm": 3.020715752986674,
+      "kl": 0.0001327991485595703,
+      "learning_rate": 1.2e-07,
+      "loss": 0.0064,
+      "reward": 0.5981995165348053,
+      "reward_std": 0.22115938365459442,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.173611119389534,
+      "rewards/repetition_penalty_reward": -0.075411606580019,
+      "rewards/tag_count_reward": 0.5,
       "step": 24
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 537.1041870117188,
-      "epoch": 0.025,
-      "grad_norm": 2.734296760351383,
-      "kl": 0.00023365020751953125,
-      "learning_rate": 2.5e-07,
-      "loss": -0.0501,
-      "reward": 1.1702337265014648,
-      "reward_std": 0.4403166174888611,
-      "rewards/accuracy_reward": 0.08333333395421505,
-      "rewards/reasoning_steps_reward": 0.6041666865348816,
-      "rewards/repetition_penalty_reward": -0.053724685683846474,
-      "rewards/tag_count_reward": 0.5364583432674408,
+      "completion_length": 410.3333435058594,
+      "epoch": 0.0125,
+      "grad_norm": 2.6737799977246337,
+      "kl": 0.00013828277587890625,
+      "learning_rate": 1.25e-07,
+      "loss": 0.0381,
+      "reward": 0.7896432876586914,
+      "reward_std": 0.43896663188934326,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.2916666865348816,
+      "rewards/repetition_penalty_reward": -0.06452339142560959,
+      "rewards/tag_count_reward": 0.5208333730697632,
       "step": 25
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 561.1041870117188,
-      "epoch": 0.026,
-      "grad_norm": 2.8294150280268386,
-      "kl": 0.0002675056457519531,
-      "learning_rate": 2.6e-07,
-      "loss": -0.007,
-      "reward": 1.1117247343063354,
-      "reward_std": 0.32203447818756104,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.6319444477558136,
-      "rewards/repetition_penalty_reward": -0.05146980658173561,
+      "completion_length": 455.2708435058594,
+      "epoch": 0.013,
+      "grad_norm": 2.7398842512902424,
+      "kl": 0.0001811981201171875,
+      "learning_rate": 1.3e-07,
+      "loss": -0.0147,
+      "reward": 0.6110469698905945,
+      "reward_std": 0.21338298171758652,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2083333358168602,
+      "rewards/repetition_penalty_reward": -0.1077030710875988,
       "rewards/tag_count_reward": 0.5104166865348816,
       "step": 26
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.0,
-      "epoch": 0.027,
-      "grad_norm": 2.762760874256859,
-      "kl": 0.00025463104248046875,
-      "learning_rate": 2.7e-07,
-      "loss": -0.0483,
-      "reward": 1.1202597618103027,
-      "reward_std": 0.36672815680503845,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.5972222685813904,
-      "rewards/repetition_penalty_reward": -0.03946254029870033,
-      "rewards/tag_count_reward": 0.5208333432674408,
+      "completion_length": 414.1041717529297,
+      "epoch": 0.0135,
+      "grad_norm": 2.8122894279350015,
+      "kl": 0.00015592575073242188,
+      "learning_rate": 1.35e-07,
+      "loss": -0.0376,
+      "reward": 0.6806438565254211,
+      "reward_std": 0.22845971584320068,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2430555671453476,
+      "rewards/repetition_penalty_reward": -0.06762006506323814,
+      "rewards/tag_count_reward": 0.5052083432674408,
       "step": 27
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 532.3125,
-      "epoch": 0.028,
-      "grad_norm": 2.5927248535988077,
-      "kl": 0.00019502639770507812,
-      "learning_rate": 2.8e-07,
-      "loss": -0.0126,
-      "reward": 1.1283650398254395,
-      "reward_std": 0.4593771994113922,
-      "rewards/accuracy_reward": 0.06250000186264515,
-      "rewards/reasoning_steps_reward": 0.5833333730697632,
-      "rewards/repetition_penalty_reward": -0.04871835932135582,
-      "rewards/tag_count_reward": 0.5312500298023224,
+      "completion_length": 442.8958435058594,
+      "epoch": 0.014,
+      "grad_norm": 2.792569424252216,
+      "kl": 0.00018310546875,
+      "learning_rate": 1.4e-07,
+      "loss": -0.0449,
+      "reward": 0.6144725680351257,
+      "reward_std": 0.2537280172109604,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.1944444552063942,
+      "rewards/repetition_penalty_reward": -0.09038857370615005,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 28
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 587.5416870117188,
-      "epoch": 0.029,
-      "grad_norm": 2.379176256693628,
-      "kl": 0.00020503997802734375,
-      "learning_rate": 2.9e-07,
-      "loss": -0.0035,
-      "reward": 0.9848364293575287,
-      "reward_std": 0.31292901933193207,
-      "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.541666716337204,
-      "rewards/repetition_penalty_reward": -0.06203855946660042,
-      "rewards/tag_count_reward": 0.5052083432674408,
+      "completion_length": 469.0833435058594,
+      "epoch": 0.0145,
+      "grad_norm": 2.573870828705473,
+      "kl": 0.00014352798461914062,
+      "learning_rate": 1.45e-07,
+      "loss": -0.0278,
+      "reward": 0.6352183520793915,
+      "reward_std": 0.31960034370422363,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.1736111268401146,
+      "rewards/repetition_penalty_reward": -0.11130945011973381,
+      "rewards/tag_count_reward": 0.5312500298023224,
       "step": 29
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 521.4791870117188,
-      "epoch": 0.03,
-      "grad_norm": 2.711197411130166,
-      "kl": 0.0002446174621582031,
-      "learning_rate": 3e-07,
-      "loss": 0.0671,
-      "reward": 1.1310882568359375,
-      "reward_std": 0.3909059464931488,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.6180555820465088,
-      "rewards/repetition_penalty_reward": -0.049467260017991066,
-      "rewards/tag_count_reward": 0.5208333432674408,
+      "completion_length": 391.125,
+      "epoch": 0.015,
+      "grad_norm": 2.895308039349145,
+      "kl": 0.00012111663818359375,
+      "learning_rate": 1.5e-07,
+      "loss": -0.0451,
+      "reward": 0.6583640277385712,
+      "reward_std": 0.37743693590164185,
+      "rewards/accuracy_reward": 0.06250000186264515,
+      "rewards/reasoning_steps_reward": 0.173611119389534,
+      "rewards/repetition_penalty_reward": -0.09337210655212402,
+      "rewards/tag_count_reward": 0.515625,
       "step": 30
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 650.3125,
-      "epoch": 0.031,
-      "grad_norm": 2.094653949008266,
-      "kl": 0.00021648406982421875,
-      "learning_rate": 3.1e-07,
-      "loss": -0.0303,
-      "reward": 1.3080359101295471,
-      "reward_std": 0.36785976588726044,
-      "rewards/accuracy_reward": 0.06250000186264515,
-      "rewards/reasoning_steps_reward": 0.784722238779068,
-      "rewards/repetition_penalty_reward": -0.05481144040822983,
-      "rewards/tag_count_reward": 0.515625,
+      "completion_length": 417.41668701171875,
+      "epoch": 0.0155,
+      "grad_norm": 3.3217228750439722,
+      "kl": 0.00017309188842773438,
+      "learning_rate": 1.55e-07,
+      "loss": -0.0388,
+      "reward": 0.6766975820064545,
+      "reward_std": 0.3049939051270485,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.173611119389534,
+      "rewards/repetition_penalty_reward": -0.06983024999499321,
+      "rewards/tag_count_reward": 0.5312500298023224,
       "step": 31
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 492.2708435058594,
-      "epoch": 0.032,
-      "grad_norm": 2.6034268800973854,
-      "kl": 0.0003032684326171875,
-      "learning_rate": 3.2e-07,
-      "loss": -0.0383,
-      "reward": 1.2631294131278992,
-      "reward_std": 0.42681366205215454,
-      "rewards/accuracy_reward": 0.06250000186264515,
-      "rewards/reasoning_steps_reward": 0.701388955116272,
-      "rewards/repetition_penalty_reward": -0.03200950939208269,
-      "rewards/tag_count_reward": 0.5312500298023224,
+      "completion_length": 467.0625,
+      "epoch": 0.016,
+      "grad_norm": 2.592972499186617,
+      "kl": 0.00017213821411132812,
+      "learning_rate": 1.6e-07,
+      "loss": 0.002,
+      "reward": 0.6280147135257721,
+      "reward_std": 0.2920844256877899,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.1666666716337204,
+      "rewards/repetition_penalty_reward": -0.09073532372713089,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 32
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 517.9375305175781,
-      "epoch": 0.033,
-      "grad_norm": 2.6134583516084144,
-      "kl": 0.0002884864807128906,
-      "learning_rate": 3.3e-07,
-      "loss": 0.085,
-      "reward": 1.3313005566596985,
-      "reward_std": 0.5551705211400986,
-      "rewards/accuracy_reward": 0.10416666977107525,
-      "rewards/reasoning_steps_reward": 0.7152777910232544,
-      "rewards/repetition_penalty_reward": -0.04022728279232979,
-      "rewards/tag_count_reward": 0.5520833730697632,
+      "completion_length": 368.2708435058594,
+      "epoch": 0.0165,
+      "grad_norm": 2.885134740538339,
+      "kl": 0.00016164779663085938,
+      "learning_rate": 1.65e-07,
+      "loss": -0.0322,
+      "reward": 0.5996805429458618,
+      "reward_std": 0.2433435320854187,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.1458333358168602,
+      "rewards/repetition_penalty_reward": -0.07219448685646057,
+      "rewards/tag_count_reward": 0.5052083432674408,
       "step": 33
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 575.625,
-      "epoch": 0.034,
-      "grad_norm": 2.785389250577513,
-      "kl": 0.00032901763916015625,
-      "learning_rate": 3.4000000000000003e-07,
-      "loss": -0.0602,
-      "reward": 1.2880616784095764,
-      "reward_std": 0.4542257487773895,
-      "rewards/accuracy_reward": 0.0833333358168602,
-      "rewards/reasoning_steps_reward": 0.694444477558136,
-      "rewards/repetition_penalty_reward": -0.04179951548576355,
-      "rewards/tag_count_reward": 0.5520833730697632,
+      "completion_length": 412.5208435058594,
+      "epoch": 0.017,
+      "grad_norm": 3.0967481185727936,
+      "kl": 0.00016927719116210938,
+      "learning_rate": 1.7000000000000001e-07,
+      "loss": -0.0958,
+      "reward": 0.6694463491439819,
+      "reward_std": 0.224090114235878,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2222222462296486,
+      "rewards/repetition_penalty_reward": -0.06319255381822586,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 34
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 521.4791870117188,
-      "epoch": 0.035,
-      "grad_norm": 2.5702386275125564,
-      "kl": 0.00037384033203125,
-      "learning_rate": 3.5e-07,
-      "loss": -0.0291,
-      "reward": 1.2401779294013977,
-      "reward_std": 0.3075665980577469,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.7500000596046448,
-      "rewards/repetition_penalty_reward": -0.04628048092126846,
-      "rewards/tag_count_reward": 0.5156250298023224,
+      "completion_length": 456.2291717529297,
+      "epoch": 0.0175,
+      "grad_norm": 2.5514196985566495,
+      "kl": 0.00011515617370605469,
+      "learning_rate": 1.75e-07,
+      "loss": -0.0634,
+      "reward": 0.7589674890041351,
+      "reward_std": 0.3829844295978546,
+      "rewards/accuracy_reward": 0.06250000186264515,
+      "rewards/reasoning_steps_reward": 0.2638889104127884,
+      "rewards/repetition_penalty_reward": -0.08825474977493286,
+      "rewards/tag_count_reward": 0.5208333730697632,
       "step": 35
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 583.4166870117188,
-      "epoch": 0.036,
-      "grad_norm": 2.4872780913924584,
-      "kl": 0.000377655029296875,
-      "learning_rate": 3.6e-07,
-      "loss": -0.0549,
-      "reward": 1.0867574214935303,
-      "reward_std": 0.3429463729262352,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.5972222685813904,
-      "rewards/repetition_penalty_reward": -0.052131447941064835,
-      "rewards/tag_count_reward": 0.5208333730697632,
+      "completion_length": 428.6458435058594,
+      "epoch": 0.018,
+      "grad_norm": 2.596517737983663,
+      "kl": 0.00011324882507324219,
+      "learning_rate": 1.8e-07,
+      "loss": -0.0258,
+      "reward": 0.6948606371879578,
+      "reward_std": 0.2351381480693817,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.284722238779068,
+      "rewards/repetition_penalty_reward": -0.08986162021756172,
+      "rewards/tag_count_reward": 0.5,
       "step": 36
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 576.4583435058594,
-      "epoch": 0.037,
-      "grad_norm": 2.5982342131854668,
-      "kl": 0.00039768218994140625,
-      "learning_rate": 3.7e-07,
-      "loss": -0.0891,
-      "reward": 1.198279321193695,
-      "reward_std": 0.26803672313690186,
+      "completion_length": 434.5416717529297,
+      "epoch": 0.0185,
+      "grad_norm": 2.6997865552883082,
+      "kl": 0.00014734268188476562,
+      "learning_rate": 1.85e-07,
+      "loss": 0.0079,
+      "reward": 0.6786134541034698,
+      "reward_std": 0.34496983885765076,
       "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.7152778208255768,
-      "rewards/repetition_penalty_reward": -0.0638735331594944,
-      "rewards/tag_count_reward": 0.5052083432674408,
+      "rewards/reasoning_steps_reward": 0.1944444626569748,
+      "rewards/repetition_penalty_reward": -0.07833104580640793,
+      "rewards/tag_count_reward": 0.5208333730697632,
       "step": 37
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 573.3541870117188,
-      "epoch": 0.038,
-      "grad_norm": 2.416596419208738,
-      "kl": 0.0005702972412109375,
-      "learning_rate": 3.7999999999999996e-07,
-      "loss": 0.021,
-      "reward": 1.1571301221847534,
-      "reward_std": 0.37196002900600433,
+      "completion_length": 420.7916717529297,
+      "epoch": 0.019,
+      "grad_norm": 2.8889113372764026,
+      "kl": 0.00016307830810546875,
+      "learning_rate": 1.8999999999999998e-07,
+      "loss": -0.0861,
+      "reward": 0.6670835018157959,
+      "reward_std": 0.297781839966774,
       "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.6527777910232544,
-      "rewards/repetition_penalty_reward": -0.04773101769387722,
-      "rewards/tag_count_reward": 0.5312500298023224,
+      "rewards/reasoning_steps_reward": 0.1944444477558136,
+      "rewards/repetition_penalty_reward": -0.0690276212990284,
+      "rewards/tag_count_reward": 0.5208333432674408,
       "step": 38
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 489.2500305175781,
-      "epoch": 0.039,
-      "grad_norm": 2.7311163851954614,
-      "kl": 0.0006694793701171875,
-      "learning_rate": 3.8999999999999997e-07,
-      "loss": 0.0014,
-      "reward": 1.1946178078651428,
-      "reward_std": 0.42158831655979156,
-      "rewards/accuracy_reward": 0.06250000186264515,
-      "rewards/reasoning_steps_reward": 0.6319445073604584,
-      "rewards/repetition_penalty_reward": -0.03628497198224068,
-      "rewards/tag_count_reward": 0.5364583730697632,
+      "completion_length": 391.50001525878906,
+      "epoch": 0.0195,
+      "grad_norm": 3.0222077809885683,
+      "kl": 0.00017881393432617188,
+      "learning_rate": 1.9499999999999999e-07,
+      "loss": -0.0421,
+      "reward": 0.6102947890758514,
+      "reward_std": 0.2821989879012108,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.1527777835726738,
+      "rewards/repetition_penalty_reward": -0.07373302057385445,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 39
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 557.4791870117188,
-      "epoch": 0.04,
-      "grad_norm": 2.5731368960029664,
-      "kl": 0.000751495361328125,
-      "learning_rate": 4e-07,
-      "loss": 0.0449,
-      "reward": 1.1931411623954773,
-      "reward_std": 0.4069296419620514,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.694444477558136,
-      "rewards/repetition_penalty_reward": -0.05338665284216404,
-      "rewards/tag_count_reward": 0.5104166865348816,
+      "completion_length": 438.60418701171875,
+      "epoch": 0.02,
+      "grad_norm": 2.8665067076530604,
+      "kl": 0.00017309188842773438,
+      "learning_rate": 2e-07,
+      "loss": -0.066,
+      "reward": 0.6099069118499756,
+      "reward_std": 0.23168403655290604,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2083333432674408,
+      "rewards/repetition_penalty_reward": -0.0984264425933361,
+      "rewards/tag_count_reward": 0.5,
       "step": 40
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 630.5625,
-      "epoch": 0.041,
-      "grad_norm": 2.394979741576189,
-      "kl": 0.0008411407470703125,
-      "learning_rate": 4.0999999999999994e-07,
-      "loss": -0.0326,
-      "reward": 1.2281219959259033,
-      "reward_std": 0.4181075543165207,
-      "rewards/accuracy_reward": 0.08333333395421505,
-      "rewards/reasoning_steps_reward": 0.6458333432674408,
-      "rewards/repetition_penalty_reward": -0.05312805995345116,
-      "rewards/tag_count_reward": 0.5520833730697632,
+      "completion_length": 552.5416717529297,
+      "epoch": 0.0205,
+      "grad_norm": 2.7988133120372214,
+      "kl": 0.00020265579223632812,
+      "learning_rate": 2.0499999999999997e-07,
+      "loss": -0.0244,
+      "reward": 0.712624192237854,
+      "reward_std": 0.2695208936929703,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2986111417412758,
+      "rewards/repetition_penalty_reward": -0.08598695322871208,
+      "rewards/tag_count_reward": 0.5,
       "step": 41
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 608.2083740234375,
-      "epoch": 0.042,
-      "grad_norm": 2.3967041495309647,
-      "kl": 0.0009288787841796875,
-      "learning_rate": 4.1999999999999995e-07,
-      "loss": 0.0268,
-      "reward": 1.1586747765541077,
-      "reward_std": 0.34190115332603455,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.6666666865348816,
-      "rewards/repetition_penalty_reward": -0.04445029981434345,
-      "rewards/tag_count_reward": 0.5156250298023224,
+      "completion_length": 429.68751525878906,
+      "epoch": 0.021,
+      "grad_norm": 3.1003673488903027,
+      "kl": 0.0001983642578125,
+      "learning_rate": 2.0999999999999997e-07,
+      "loss": -0.0657,
+      "reward": 0.6647423803806305,
+      "reward_std": 0.3343476206064224,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.1944444589316845,
+      "rewards/repetition_penalty_reward": -0.08178546652197838,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 42
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 545.5208435058594,
-      "epoch": 0.043,
-      "grad_norm": 2.7138542170454003,
-      "kl": 0.00109100341796875,
-      "learning_rate": 4.2999999999999996e-07,
-      "loss": -0.0442,
-      "reward": 1.2215397357940674,
-      "reward_std": 0.330724373459816,
-      "rewards/accuracy_reward": 0.02083333395421505,
-      "rewards/reasoning_steps_reward": 0.729166716337204,
-      "rewards/repetition_penalty_reward": -0.04929363913834095,
-      "rewards/tag_count_reward": 0.5208333730697632,
+      "completion_length": 388.29168701171875,
+      "epoch": 0.0215,
+      "grad_norm": 3.1597548965345204,
+      "kl": 0.000186920166015625,
+      "learning_rate": 2.1499999999999998e-07,
+      "loss": 0.0158,
+      "reward": 0.6541115939617157,
+      "reward_std": 0.2999124675989151,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.1527777872979641,
+      "rewards/repetition_penalty_reward": -0.05595788359642029,
+      "rewards/tag_count_reward": 0.515625,
       "step": 43
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.8125,
-      "epoch": 0.044,
-      "grad_norm": 2.9707533290622923,
-      "kl": 0.001163482666015625,
-      "learning_rate": 4.3999999999999997e-07,
-      "loss": -0.0372,
-      "reward": 1.193300485610962,
-      "reward_std": 0.3503916561603546,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.6527778208255768,
-      "rewards/repetition_penalty_reward": -0.0428106477484107,
-      "rewards/tag_count_reward": 0.5416666865348816,
+      "completion_length": 447.18751525878906,
+      "epoch": 0.022,
+      "grad_norm": 2.558177785664247,
+      "kl": 0.00014591217041015625,
+      "learning_rate": 2.1999999999999998e-07,
+      "loss": -0.0045,
+      "reward": 0.6448527276515961,
+      "reward_std": 0.25322096794843674,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2430555745959282,
+      "rewards/repetition_penalty_reward": -0.0982028879225254,
+      "rewards/tag_count_reward": 0.5,
       "step": 44
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 592.2291870117188,
-      "epoch": 0.045,
-      "grad_norm": 2.8301357058974603,
-      "kl": 0.001316070556640625,
-      "learning_rate": 4.5e-07,
-      "loss": -0.0149,
-      "reward": 1.5306417346000671,
-      "reward_std": 0.6089198887348175,
-      "rewards/accuracy_reward": 0.2291666716337204,
-      "rewards/reasoning_steps_reward": 0.7430555522441864,
-      "rewards/repetition_penalty_reward": -0.04574713110923767,
-      "rewards/tag_count_reward": 0.6041666865348816,
+      "completion_length": 430.3125,
+      "epoch": 0.0225,
+      "grad_norm": 2.8810601257270263,
+      "kl": 0.00017499923706054688,
+      "learning_rate": 2.25e-07,
+      "loss": -0.0455,
+      "reward": 0.6615873873233795,
+      "reward_std": 0.2432490661740303,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.222222238779068,
+      "rewards/repetition_penalty_reward": -0.06063482351601124,
+      "rewards/tag_count_reward": 0.5,
       "step": 45
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 560.0833435058594,
-      "epoch": 0.046,
-      "grad_norm": 2.9830534003583997,
-      "kl": 0.001560211181640625,
-      "learning_rate": 4.6e-07,
-      "loss": -0.0654,
-      "reward": 1.2372803688049316,
-      "reward_std": 0.3439289480447769,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.7013889253139496,
-      "rewards/repetition_penalty_reward": -0.04744199104607105,
-      "rewards/tag_count_reward": 0.5416666865348816,
+      "completion_length": 420.18751525878906,
+      "epoch": 0.023,
+      "grad_norm": 2.727537687409915,
+      "kl": 0.00017786026000976562,
+      "learning_rate": 2.3e-07,
+      "loss": -0.0767,
+      "reward": 0.7279665172100067,
+      "reward_std": 0.3034803867340088,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.2708333507180214,
+      "rewards/repetition_penalty_reward": -0.07411682605743408,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 46
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 595.5416870117188,
-      "epoch": 0.047,
-      "grad_norm": 2.62124053942929,
-      "kl": 0.001735687255859375,
-      "learning_rate": 4.6999999999999995e-07,
-      "loss": 0.0004,
-      "reward": 1.289455771446228,
-      "reward_std": 0.41812919080257416,
-      "rewards/accuracy_reward": 0.06250000186264515,
-      "rewards/reasoning_steps_reward": 0.7222222685813904,
-      "rewards/repetition_penalty_reward": -0.04734986647963524,
-      "rewards/tag_count_reward": 0.5520833730697632,
+      "completion_length": 348.72918701171875,
+      "epoch": 0.0235,
+      "grad_norm": 3.1997159072003227,
+      "kl": 0.00018548965454101562,
+      "learning_rate": 2.3499999999999997e-07,
+      "loss": 0.0492,
+      "reward": 0.598628580570221,
+      "reward_std": 0.2303108423948288,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.11111111752688885,
+      "rewards/repetition_penalty_reward": -0.043732548132538795,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 47
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 603.2500305175781,
-      "epoch": 0.048,
-      "grad_norm": 2.313447720284229,
-      "kl": 0.001773834228515625,
-      "learning_rate": 4.8e-07,
-      "loss": 0.0089,
-      "reward": 1.3387857675552368,
-      "reward_std": 0.3809618651866913,
-      "rewards/accuracy_reward": 0.12500000558793545,
-      "rewards/reasoning_steps_reward": 0.7083333432674408,
-      "rewards/repetition_penalty_reward": -0.05183934420347214,
-      "rewards/tag_count_reward": 0.5572916865348816,
+      "completion_length": 383.6458435058594,
+      "epoch": 0.024,
+      "grad_norm": 2.6573394355964783,
+      "kl": 0.00020503997802734375,
+      "learning_rate": 2.4e-07,
+      "loss": 0.033,
+      "reward": 0.6143307387828827,
+      "reward_std": 0.28714819252491,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.166666679084301,
+      "rewards/repetition_penalty_reward": -0.08358598873019218,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 48
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 523.3541717529297,
-      "epoch": 0.049,
-      "grad_norm": 2.8891626678471063,
-      "kl": 0.00232696533203125,
-      "learning_rate": 4.9e-07,
-      "loss": -0.089,
-      "reward": 1.2625147104263306,
-      "reward_std": 0.4176745116710663,
-      "rewards/accuracy_reward": 0.0416666679084301,
-      "rewards/reasoning_steps_reward": 0.7083334028720856,
-      "rewards/repetition_penalty_reward": -0.039568664506077766,
-      "rewards/tag_count_reward": 0.5520833730697632,
+      "completion_length": 436.62501525878906,
+      "epoch": 0.0245,
+      "grad_norm": 2.7867155541781488,
+      "kl": 0.00023317337036132812,
+      "learning_rate": 2.45e-07,
+      "loss": -0.0491,
+      "reward": 0.6386054158210754,
+      "reward_std": 0.24428075551986694,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.1875000149011612,
+      "rewards/repetition_penalty_reward": -0.06972793489694595,
+      "rewards/tag_count_reward": 0.5,
       "step": 49
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.0625,
-      "epoch": 0.05,
-      "grad_norm": 2.6847464499943308,
-      "kl": 0.0029754638671875,
-      "learning_rate": 5e-07,
-      "loss": 0.0743,
-      "reward": 1.615307629108429,
-      "reward_std": 0.6159101724624634,
-      "rewards/accuracy_reward": 0.3125000149011612,
-      "rewards/reasoning_steps_reward": 0.7083333730697632,
-      "rewards/repetition_penalty_reward": -0.046150704845786095,
-      "rewards/tag_count_reward": 0.6406250298023224,
+      "completion_length": 449.9166717529297,
+      "epoch": 0.025,
+      "grad_norm": 3.006240341565843,
+      "kl": 0.00022745132446289062,
+      "learning_rate": 2.5e-07,
+      "loss": -0.1239,
+      "reward": 0.7317107021808624,
+      "reward_std": 0.27121981978416443,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.3125000149011612,
+      "rewards/repetition_penalty_reward": -0.08078934252262115,
+      "rewards/tag_count_reward": 0.5,
       "step": 50
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.1875305175781,
-      "epoch": 0.051,
-      "grad_norm": 2.4123895595368547,
-      "kl": 0.0031280517578125,
-      "learning_rate": 5.1e-07,
-      "loss": 0.0147,
-      "reward": 1.4028043746948242,
-      "reward_std": 0.43062111735343933,
-      "rewards/accuracy_reward": 0.1041666679084301,
-      "rewards/reasoning_steps_reward": 0.7430555522441864,
-      "rewards/repetition_penalty_reward": -0.04858442768454552,
-      "rewards/tag_count_reward": 0.6041666865348816,
+      "completion_length": 472.2291717529297,
+      "epoch": 0.0255,
+      "grad_norm": 2.6548371678838856,
+      "kl": 0.00025463104248046875,
+      "learning_rate": 2.55e-07,
+      "loss": -0.0904,
+      "reward": 0.6300550401210785,
+      "reward_std": 0.31470321863889694,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.1666666716337204,
+      "rewards/repetition_penalty_reward": -0.1043199859559536,
+      "rewards/tag_count_reward": 0.5260416865348816,
       "step": 51
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 575.7291870117188,
-      "epoch": 0.052,
-      "grad_norm": 2.8249120385470197,
-      "kl": 0.00350189208984375,
-      "learning_rate": 5.2e-07,
-      "loss": -0.0312,
-      "reward": 1.5281450748443604,
-      "reward_std": 0.6317126750946045,
-      "rewards/accuracy_reward": 0.2708333358168602,
-      "rewards/reasoning_steps_reward": 0.6805556118488312,
-      "rewards/repetition_penalty_reward": -0.058660563081502914,
-      "rewards/tag_count_reward": 0.6354166865348816,
+      "completion_length": 410.37501525878906,
+      "epoch": 0.026,
+      "grad_norm": 2.891679932666362,
+      "kl": 0.00026416778564453125,
+      "learning_rate": 2.6e-07,
+      "loss": -0.0272,
+      "reward": 0.5869153141975403,
+      "reward_std": 0.2003287822008133,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.173611119389534,
+      "rewards/repetition_penalty_reward": -0.0866958275437355,
+      "rewards/tag_count_reward": 0.5,
       "step": 52
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 519.4166870117188,
-      "epoch": 0.053,
-      "grad_norm": 2.8646116438456453,
-      "kl": 0.00429534912109375,
-      "learning_rate": 5.3e-07,
-      "loss": 0.0331,
-      "reward": 1.6907492280006409,
-      "reward_std": 0.6883328557014465,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.6875000298023224,
-      "rewards/repetition_penalty_reward": -0.03841756656765938,
-      "rewards/tag_count_reward": 0.6875,
+      "completion_length": 434.125,
+      "epoch": 0.0265,
+      "grad_norm": 2.87497343902247,
+      "kl": 0.0002574920654296875,
+      "learning_rate": 2.65e-07,
+      "loss": 0.0957,
+      "reward": 0.6155928373336792,
+      "reward_std": 0.31732793152332306,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.1875000149011612,
+      "rewards/repetition_penalty_reward": -0.0979488454759121,
+      "rewards/tag_count_reward": 0.5052083432674408,
       "step": 53
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 578.1250305175781,
-      "epoch": 0.054,
-      "grad_norm": 2.532736796747683,
-      "kl": 0.0043792724609375,
-      "learning_rate": 5.4e-07,
-      "loss": 0.0327,
-      "reward": 1.4663925766944885,
-      "reward_std": 0.4452537000179291,
-      "rewards/accuracy_reward": 0.12500000558793545,
-      "rewards/reasoning_steps_reward": 0.7986111342906952,
-      "rewards/repetition_penalty_reward": -0.06138528883457184,
-      "rewards/tag_count_reward": 0.6041666865348816,
+      "completion_length": 388.5833435058594,
+      "epoch": 0.027,
+      "grad_norm": 3.1249467609329815,
+      "kl": 0.00028514862060546875,
+      "learning_rate": 2.7e-07,
+      "loss": -0.0707,
+      "reward": 0.6530064940452576,
+      "reward_std": 0.26522205770015717,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2361111268401146,
+      "rewards/repetition_penalty_reward": -0.08831300958991051,
+      "rewards/tag_count_reward": 0.5052083432674408,
       "step": 54
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 556.3750305175781,
-      "epoch": 0.055,
-      "grad_norm": 2.626064949357054,
-      "kl": 0.0063018798828125,
-      "learning_rate": 5.5e-07,
-      "loss": 0.0412,
-      "reward": 1.5763724446296692,
-      "reward_std": 0.5313678234815598,
-      "rewards/accuracy_reward": 0.2500000149011612,
-      "rewards/reasoning_steps_reward": 0.7083333730697632,
-      "rewards/repetition_penalty_reward": -0.05904424749314785,
-      "rewards/tag_count_reward": 0.6770833432674408,
+      "completion_length": 391.7916717529297,
+      "epoch": 0.0275,
+      "grad_norm": 3.7182582598169582,
+      "kl": 0.00043487548828125,
+      "learning_rate": 2.75e-07,
+      "loss": -0.1307,
+      "reward": 0.6811047792434692,
+      "reward_std": 0.3341375142335892,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.2083333432674408,
+      "rewards/repetition_penalty_reward": -0.08972860500216484,
+      "rewards/tag_count_reward": 0.5208333730697632,
       "step": 55
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.9166870117188,
-      "epoch": 0.056,
-      "grad_norm": 2.6301403493903748,
-      "kl": 0.0066680908203125,
-      "learning_rate": 5.6e-07,
-      "loss": 0.0203,
-      "reward": 1.794304072856903,
-      "reward_std": 0.6143919825553894,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.7361111640930176,
-      "rewards/repetition_penalty_reward": -0.04597374238073826,
-      "rewards/tag_count_reward": 0.7291666865348816,
+      "completion_length": 431.72918701171875,
+      "epoch": 0.028,
+      "grad_norm": 2.9876795440276056,
+      "kl": 0.00030040740966796875,
+      "learning_rate": 2.8e-07,
+      "loss": -0.0276,
+      "reward": 0.6388545334339142,
+      "reward_std": 0.1873469203710556,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2152777835726738,
+      "rewards/repetition_penalty_reward": -0.0764232836663723,
+      "rewards/tag_count_reward": 0.5,
       "step": 56
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 581.75,
-      "epoch": 0.057,
-      "grad_norm": 2.377807933864504,
-      "kl": 0.0074462890625,
-      "learning_rate": 5.699999999999999e-07,
-      "loss": 0.0249,
-      "reward": 1.681954801082611,
-      "reward_std": 0.5288814753293991,
-      "rewards/accuracy_reward": 0.291666679084301,
-      "rewards/reasoning_steps_reward": 0.7430556118488312,
-      "rewards/repetition_penalty_reward": -0.05589243024587631,
-      "rewards/tag_count_reward": 0.703125,
+      "completion_length": 446.7916717529297,
+      "epoch": 0.0285,
+      "grad_norm": 2.5753158131037117,
+      "kl": 0.00035953521728515625,
+      "learning_rate": 2.8499999999999997e-07,
+      "loss": -0.0241,
+      "reward": 0.740548849105835,
+      "reward_std": 0.40591859817504883,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.2500000074505806,
+      "rewards/repetition_penalty_reward": -0.09278450906276703,
+      "rewards/tag_count_reward": 0.5416666865348816,
       "step": 57
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 564.7916870117188,
-      "epoch": 0.058,
-      "grad_norm": 2.9351818937282297,
-      "kl": 0.0069580078125,
-      "learning_rate": 5.8e-07,
-      "loss": 0.0467,
-      "reward": 2.0512834787368774,
-      "reward_std": 0.659260630607605,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.7708333730697632,
-      "rewards/repetition_penalty_reward": -0.04767502471804619,
-      "rewards/tag_count_reward": 0.7864583432674408,
+      "completion_length": 292.2291793823242,
+      "epoch": 0.029,
+      "grad_norm": 5.591928815640096,
+      "kl": 0.0007524490356445312,
+      "learning_rate": 2.9e-07,
+      "loss": 0.0557,
+      "reward": 0.6187903136014938,
+      "reward_std": 0.20580651611089706,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.19444445706903934,
+      "rewards/repetition_penalty_reward": -0.07565413787961006,
+      "rewards/tag_count_reward": 0.5,
       "step": 58
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 611.2916870117188,
-      "epoch": 0.059,
-      "grad_norm": 2.5837982836551467,
-      "kl": 0.00921630859375,
-      "learning_rate": 5.9e-07,
-      "loss": -0.0409,
-      "reward": 1.758675456047058,
-      "reward_std": 0.5324381589889526,
-      "rewards/accuracy_reward": 0.2500000111758709,
-      "rewards/reasoning_steps_reward": 0.7222222089767456,
-      "rewards/repetition_penalty_reward": -0.05729677900671959,
-      "rewards/tag_count_reward": 0.84375,
+      "completion_length": 421.5416717529297,
+      "epoch": 0.0295,
+      "grad_norm": 3.1936915993711903,
+      "kl": 0.00032329559326171875,
+      "learning_rate": 2.95e-07,
+      "loss": -0.1173,
+      "reward": 0.7119120061397552,
+      "reward_std": 0.47539106011390686,
+      "rewards/accuracy_reward": 0.08333333395421505,
+      "rewards/reasoning_steps_reward": 0.1875000111758709,
+      "rewards/repetition_penalty_reward": -0.07975470274686813,
+      "rewards/tag_count_reward": 0.5208333730697632,
       "step": 59
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.6666870117188,
-      "epoch": 0.06,
-      "grad_norm": 3.05141171791299,
-      "kl": 0.011322021484375,
-      "learning_rate": 6e-07,
-      "loss": 0.0112,
-      "reward": 2.136493146419525,
-      "reward_std": 0.587219700217247,
-      "rewards/accuracy_reward": 0.5625000149011612,
-      "rewards/reasoning_steps_reward": 0.756944477558136,
-      "rewards/repetition_penalty_reward": -0.0475346464663744,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 506.2708435058594,
+      "epoch": 0.03,
+      "grad_norm": 3.162339001956615,
+      "kl": 0.0004940032958984375,
+      "learning_rate": 3e-07,
+      "loss": 0.0028,
+      "reward": 0.7199562191963196,
+      "reward_std": 0.25821222364902496,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.3125000149011612,
+      "rewards/repetition_penalty_reward": -0.09254380315542221,
+      "rewards/tag_count_reward": 0.5,
       "step": 60
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 547.7916870117188,
-      "epoch": 0.061,
-      "grad_norm": 2.7073330792983072,
-      "kl": 0.011810302734375,
-      "learning_rate": 6.1e-07,
-      "loss": 0.0477,
-      "reward": 2.2359567284584045,
-      "reward_std": 0.5013462156057358,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.6458333134651184,
-      "rewards/repetition_penalty_reward": -0.04008489940315485,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 439.12501525878906,
+      "epoch": 0.0305,
+      "grad_norm": 2.6085011916809364,
+      "kl": 0.0004892349243164062,
+      "learning_rate": 3.05e-07,
+      "loss": -0.0387,
+      "reward": 0.6562853455543518,
+      "reward_std": 0.2614954710006714,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2291666865348816,
+      "rewards/repetition_penalty_reward": -0.09371470659971237,
+      "rewards/tag_count_reward": 0.5208333432674408,
       "step": 61
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 449.5208435058594,
-      "epoch": 0.062,
-      "grad_norm": 3.257750731526075,
-      "kl": 0.019744873046875,
-      "learning_rate": 6.2e-07,
-      "loss": 0.0639,
-      "reward": 2.321463108062744,
-      "reward_std": 0.4783380925655365,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.7013889253139496,
-      "rewards/repetition_penalty_reward": -0.04138421919196844,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 438.9791717529297,
+      "epoch": 0.031,
+      "grad_norm": 2.933119213416396,
+      "kl": 0.00045871734619140625,
+      "learning_rate": 3.1e-07,
+      "loss": 0.0025,
+      "reward": 0.7185687720775604,
+      "reward_std": 0.2700326144695282,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.3125000298023224,
+      "rewards/repetition_penalty_reward": -0.09393121674656868,
+      "rewards/tag_count_reward": 0.5,
       "step": 62
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 547.2500305175781,
-      "epoch": 0.063,
-      "grad_norm": 2.7432912908420732,
-      "kl": 0.011871337890625,
-      "learning_rate": 6.3e-07,
-      "loss": 0.0445,
-      "reward": 2.019491195678711,
-      "reward_std": 0.5333467572927475,
-      "rewards/accuracy_reward": 0.458333358168602,
-      "rewards/reasoning_steps_reward": 0.6736111640930176,
-      "rewards/repetition_penalty_reward": -0.04995330423116684,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 403.1458435058594,
+      "epoch": 0.0315,
+      "grad_norm": 2.897102185777002,
+      "kl": 0.000621795654296875,
+      "learning_rate": 3.15e-07,
+      "loss": -0.0791,
+      "reward": 0.6670123040676117,
+      "reward_std": 0.237733893096447,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.2569444477558136,
+      "rewards/repetition_penalty_reward": -0.08993213623762131,
+      "rewards/tag_count_reward": 0.5,
       "step": 63
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 533.8125305175781,
-      "epoch": 0.064,
-      "grad_norm": 2.775011190038876,
-      "kl": 0.01220703125,
-      "learning_rate": 6.4e-07,
-      "loss": -0.0156,
-      "reward": 2.413369059562683,
-      "reward_std": 0.517508327960968,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.79861119389534,
-      "rewards/repetition_penalty_reward": -0.041492147371172905,
-      "rewards/tag_count_reward": 0.96875,
+      "completion_length": 433.6041717529297,
+      "epoch": 0.032,
+      "grad_norm": 2.446714560125192,
+      "kl": 0.0006542205810546875,
+      "learning_rate": 3.2e-07,
+      "loss": 0.0159,
+      "reward": 0.7158828973770142,
+      "reward_std": 0.33434779942035675,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.2638888955116272,
+      "rewards/repetition_penalty_reward": -0.10529769212007523,
+      "rewards/tag_count_reward": 0.5156250298023224,
       "step": 64
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 511.16668701171875,
-      "epoch": 0.065,
-      "grad_norm": 2.745662806864412,
-      "kl": 0.014892578125,
-      "learning_rate": 6.5e-07,
-      "loss": -0.0064,
-      "reward": 2.327776312828064,
-      "reward_std": 0.5232683420181274,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.694444477558136,
-      "rewards/repetition_penalty_reward": -0.038543105125427246,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 401.54168701171875,
+      "epoch": 0.0325,
+      "grad_norm": 2.7732151443967163,
+      "kl": 0.000652313232421875,
+      "learning_rate": 3.25e-07,
+      "loss": -0.0358,
+      "reward": 0.8164660930633545,
+      "reward_std": 0.3192940354347229,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.326388917863369,
+      "rewards/repetition_penalty_reward": -0.07242286205291748,
+      "rewards/tag_count_reward": 0.5208333730697632,
       "step": 65
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 538.875,
-      "epoch": 0.066,
-      "grad_norm": 2.5297084422395555,
-      "kl": 0.0113525390625,
-      "learning_rate": 6.6e-07,
-      "loss": -0.0286,
-      "reward": 2.1816118955612183,
-      "reward_std": 0.38229691982269287,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.6805555820465088,
-      "rewards/repetition_penalty_reward": -0.03540213964879513,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 320.4583435058594,
+      "epoch": 0.033,
+      "grad_norm": 6.165543609500131,
+      "kl": 0.0016994476318359375,
+      "learning_rate": 3.3e-07,
+      "loss": -0.0995,
+      "reward": 0.7236150503158569,
+      "reward_std": 0.33078788220882416,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.2361111268401146,
+      "rewards/repetition_penalty_reward": -0.0645794328302145,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 66
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 559.8958435058594,
-      "epoch": 0.067,
-      "grad_norm": 2.672746029201746,
-      "kl": 0.012908935546875,
-      "learning_rate": 6.7e-07,
-      "loss": -0.035,
-      "reward": 2.2214502096176147,
-      "reward_std": 0.5396545231342316,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.7152777910232544,
-      "rewards/repetition_penalty_reward": -0.04591094329953194,
-      "rewards/tag_count_reward": 0.9895833730697632,
+      "completion_length": 391.9791717529297,
+      "epoch": 0.0335,
+      "grad_norm": 2.658500757905026,
+      "kl": 0.001007080078125,
+      "learning_rate": 3.35e-07,
+      "loss": -0.0184,
+      "reward": 0.6389293074607849,
+      "reward_std": 0.27491873502731323,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.229166679084301,
+      "rewards/repetition_penalty_reward": -0.12148737907409668,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 67
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 537.3125,
-      "epoch": 0.068,
-      "grad_norm": 2.920361106100061,
-      "kl": 0.019287109375,
-      "learning_rate": 6.800000000000001e-07,
-      "loss": 0.0507,
-      "reward": 2.291020631790161,
-      "reward_std": 0.49400143325328827,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.7361111342906952,
-      "rewards/repetition_penalty_reward": -0.049257127568125725,
-      "rewards/tag_count_reward": 0.9583333730697632,
+      "completion_length": 444.3333435058594,
+      "epoch": 0.034,
+      "grad_norm": 2.6477800642599996,
+      "kl": 0.0009365081787109375,
+      "learning_rate": 3.4000000000000003e-07,
+      "loss": -0.0267,
+      "reward": 0.7566591203212738,
+      "reward_std": 0.3201362192630768,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.270833358168602,
+      "rewards/repetition_penalty_reward": -0.08709090948104858,
+      "rewards/tag_count_reward": 0.5520833432674408,
       "step": 68
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 469.2708435058594,
-      "epoch": 0.069,
-      "grad_norm": 2.901120243600892,
-      "kl": 0.017242431640625,
-      "learning_rate": 6.9e-07,
-      "loss": 0.0654,
-      "reward": 2.4167280197143555,
-      "reward_std": 0.5367273986339569,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.7638889849185944,
-      "rewards/repetition_penalty_reward": -0.03466090187430382,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 457.5833435058594,
+      "epoch": 0.0345,
+      "grad_norm": 2.633615712345152,
+      "kl": 0.001323699951171875,
+      "learning_rate": 3.45e-07,
+      "loss": -0.0703,
+      "reward": 0.7424019575119019,
+      "reward_std": 0.3558398336172104,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.3194444477558136,
+      "rewards/repetition_penalty_reward": -0.11350089311599731,
+      "rewards/tag_count_reward": 0.5156250298023224,
       "step": 69
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 504.6666717529297,
-      "epoch": 0.07,
-      "grad_norm": 3.137395976127796,
-      "kl": 0.018524169921875,
-      "learning_rate": 7e-07,
-      "loss": 0.0147,
-      "reward": 2.4454265832901,
-      "reward_std": 0.5369938015937805,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.7083333730697632,
-      "rewards/repetition_penalty_reward": -0.04415685310959816,
-      "rewards/tag_count_reward": 0.9895833432674408,
-      "step": 70
-    },
+      "completion_length": 436.72918701171875,
+      "epoch": 0.035,
+      "grad_norm": 2.949344418221538,
+      "kl": 0.0012359619140625,
+      "learning_rate": 3.5e-07,
+      "loss": -0.0637,
+      "reward": 0.7435359060764313,
+      "reward_std": 0.20930662006139755,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.3194444626569748,
+      "rewards/repetition_penalty_reward": -0.0759085863828659,
+      "rewards/tag_count_reward": 0.5,
+      "step": 70
+    },
     {
       "clip_ratio": 0.0,
-      "completion_length": 535.5000305175781,
-      "epoch": 0.071,
-      "grad_norm": 2.6854335457859206,
-      "kl": 0.01251220703125,
-      "learning_rate": 7.1e-07,
-      "loss": 0.0458,
-      "reward": 2.375571846961975,
-      "reward_std": 0.4342511296272278,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.7222222685813904,
-      "rewards/repetition_penalty_reward": -0.028942234814167023,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 423.5833435058594,
+      "epoch": 0.0355,
+      "grad_norm": 2.869137528454137,
+      "kl": 0.001552581787109375,
+      "learning_rate": 3.55e-07,
+      "loss": -0.0516,
+      "reward": 0.7146015167236328,
+      "reward_std": 0.29403047263622284,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.298611119389534,
+      "rewards/repetition_penalty_reward": -0.08400958776473999,
+      "rewards/tag_count_reward": 0.5,
       "step": 71
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.6250305175781,
-      "epoch": 0.072,
-      "grad_norm": 2.736209803149782,
-      "kl": 0.01373291015625,
-      "learning_rate": 7.2e-07,
-      "loss": 0.0066,
-      "reward": 2.4244872331619263,
-      "reward_std": 0.4428471177816391,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.7361111640930176,
-      "rewards/repetition_penalty_reward": -0.04079079441726208,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 490.0208435058594,
+      "epoch": 0.036,
+      "grad_norm": 2.479062073282576,
+      "kl": 0.001407623291015625,
+      "learning_rate": 3.6e-07,
+      "loss": 0.0264,
+      "reward": 0.766609251499176,
+      "reward_std": 0.3444522023200989,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.3263889104127884,
+      "rewards/repetition_penalty_reward": -0.09102966263890266,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 72
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 540.7708740234375,
-      "epoch": 0.073,
-      "grad_norm": 3.26658788883033,
-      "kl": 0.0230712890625,
-      "learning_rate": 7.3e-07,
-      "loss": 0.05,
-      "reward": 2.218494176864624,
-      "reward_std": 0.44005706906318665,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.7291666865348816,
-      "rewards/repetition_penalty_reward": -0.036714269779622555,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 403.35418701171875,
+      "epoch": 0.0365,
+      "grad_norm": 2.55373473630631,
+      "kl": 0.00188446044921875,
+      "learning_rate": 3.65e-07,
+      "loss": 0.0128,
+      "reward": 0.8594351708889008,
+      "reward_std": 0.38698044419288635,
+      "rewards/accuracy_reward": 0.06250000186264515,
+      "rewards/reasoning_steps_reward": 0.361111119389534,
+      "rewards/repetition_penalty_reward": -0.10063427314162254,
+      "rewards/tag_count_reward": 0.5364583730697632,
       "step": 73
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 477.5833435058594,
-      "epoch": 0.074,
-      "grad_norm": 2.85387562733274,
-      "kl": 0.015350341796875,
-      "learning_rate": 7.4e-07,
-      "loss": -0.0614,
-      "reward": 2.5582029819488525,
-      "reward_std": 0.46813952922821045,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.784722238779068,
-      "rewards/repetition_penalty_reward": -0.03901927825063467,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 441.31251525878906,
+      "epoch": 0.037,
+      "grad_norm": 2.763489892021762,
+      "kl": 0.00213623046875,
+      "learning_rate": 3.7e-07,
+      "loss": -0.0037,
+      "reward": 0.8403445482254028,
+      "reward_std": 0.22306300699710846,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.409722238779068,
+      "rewards/repetition_penalty_reward": -0.10062775760889053,
+      "rewards/tag_count_reward": 0.53125,
       "step": 74
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 530.1041870117188,
-      "epoch": 0.075,
-      "grad_norm": 2.963663960294806,
-      "kl": 0.017822265625,
-      "learning_rate": 7.5e-07,
-      "loss": 0.0242,
-      "reward": 2.356380820274353,
-      "reward_std": 0.5158264935016632,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.7083334028720856,
-      "rewards/repetition_penalty_reward": -0.03945251181721687,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 430.0416717529297,
+      "epoch": 0.0375,
+      "grad_norm": 2.9294440220568005,
+      "kl": 0.00238800048828125,
+      "learning_rate": 3.75e-07,
+      "loss": -0.0807,
+      "reward": 0.8580312430858612,
+      "reward_std": 0.39887402951717377,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.3680555820465088,
+      "rewards/repetition_penalty_reward": -0.08294104412198067,
+      "rewards/tag_count_reward": 0.53125,
       "step": 75
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 542.7500305175781,
-      "epoch": 0.076,
-      "grad_norm": 8.251127258415144,
-      "kl": 0.02813720703125,
-      "learning_rate": 7.599999999999999e-07,
-      "loss": 0.0219,
-      "reward": 2.4178924560546875,
-      "reward_std": 0.44688859581947327,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.7222222089767456,
-      "rewards/repetition_penalty_reward": -0.043913234025239944,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 356.8541717529297,
+      "epoch": 0.038,
+      "grad_norm": 2.1979386745725438,
+      "kl": 0.00208282470703125,
+      "learning_rate": 3.7999999999999996e-07,
+      "loss": -0.0118,
+      "reward": 0.8219205439090729,
+      "reward_std": 0.33333858847618103,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.3055555745959282,
+      "rewards/repetition_penalty_reward": -0.0669684112071991,
+      "rewards/tag_count_reward": 0.5416666865348816,
       "step": 76
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 547.0625,
-      "epoch": 0.077,
-      "grad_norm": 3.4452973677390144,
-      "kl": 0.021881103515625,
-      "learning_rate": 7.699999999999999e-07,
-      "loss": 0.0425,
-      "reward": 2.3173056840896606,
-      "reward_std": 0.49892735481262207,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.75,
-      "rewards/repetition_penalty_reward": -0.04727764055132866,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 499.41668701171875,
+      "epoch": 0.0385,
+      "grad_norm": 2.4412843501817862,
+      "kl": 0.00276947021484375,
+      "learning_rate": 3.8499999999999997e-07,
+      "loss": 0.0243,
+      "reward": 0.9450303316116333,
+      "reward_std": 0.35558322072029114,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.472222238779068,
+      "rewards/repetition_penalty_reward": -0.10531692206859589,
+      "rewards/tag_count_reward": 0.5364583432674408,
       "step": 77
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 554.3958435058594,
-      "epoch": 0.078,
-      "grad_norm": 2.5762620979216746,
-      "kl": 0.017181396484375,
-      "learning_rate": 7.799999999999999e-07,
-      "loss": -0.0855,
-      "reward": 2.4648728370666504,
-      "reward_std": 0.5124194473028183,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.8472222685813904,
-      "rewards/repetition_penalty_reward": -0.04380778409540653,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 395.62501525878906,
+      "epoch": 0.039,
+      "grad_norm": 2.8263728812238895,
+      "kl": 0.00323486328125,
+      "learning_rate": 3.8999999999999997e-07,
+      "loss": -0.0533,
+      "reward": 0.7922153770923615,
+      "reward_std": 0.2505815550684929,
+      "rewards/accuracy_reward": 0.0,
+      "rewards/reasoning_steps_reward": 0.3750000298023224,
+      "rewards/repetition_penalty_reward": -0.08278463035821915,
+      "rewards/tag_count_reward": 0.5,
       "step": 78
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 490.83335876464844,
-      "epoch": 0.079,
-      "grad_norm": 2.8936310930522597,
-      "kl": 0.0186767578125,
-      "learning_rate": 7.9e-07,
-      "loss": 0.0224,
-      "reward": 2.5715928077697754,
-      "reward_std": 0.4545498937368393,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.7708334028720856,
-      "rewards/repetition_penalty_reward": -0.03257405199110508,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 408.9791717529297,
+      "epoch": 0.0395,
+      "grad_norm": 2.623388298754155,
+      "kl": 0.00296783447265625,
+      "learning_rate": 3.95e-07,
+      "loss": -0.0619,
+      "reward": 0.8954714834690094,
+      "reward_std": 0.3817988187074661,
+      "rewards/accuracy_reward": 0.06250000186264515,
+      "rewards/reasoning_steps_reward": 0.3819444477558136,
+      "rewards/repetition_penalty_reward": -0.07501471415162086,
+      "rewards/tag_count_reward": 0.5260416865348816,
       "step": 79
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 505.16668701171875,
-      "epoch": 0.08,
-      "grad_norm": 3.386283681350372,
-      "kl": 0.018951416015625,
-      "learning_rate": 8e-07,
-      "loss": 0.0251,
-      "reward": 2.380759596824646,
-      "reward_std": 0.40306712687015533,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.7708333432674408,
-      "rewards/repetition_penalty_reward": -0.03590711485594511,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 440.35418701171875,
+      "epoch": 0.04,
+      "grad_norm": 3.2358128766367216,
+      "kl": 0.0038299560546875,
+      "learning_rate": 4e-07,
+      "loss": -0.0908,
+      "reward": 1.0643112063407898,
+      "reward_std": 0.43053002655506134,
+      "rewards/accuracy_reward": 0.14583333395421505,
+      "rewards/reasoning_steps_reward": 0.4166666567325592,
+      "rewards/repetition_penalty_reward": -0.07110553234815598,
+      "rewards/tag_count_reward": 0.5729166865348816,
       "step": 80
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 518.7708435058594,
-      "epoch": 0.081,
-      "grad_norm": 2.774591505446817,
-      "kl": 0.016845703125,
-      "learning_rate": 8.1e-07,
-      "loss": 0.1045,
-      "reward": 2.4150757789611816,
-      "reward_std": 0.34570978581905365,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.7638889253139496,
-      "rewards/repetition_penalty_reward": -0.03631327673792839,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 464.41668701171875,
+      "epoch": 0.0405,
+      "grad_norm": 2.469210493769818,
+      "kl": 0.0035552978515625,
+      "learning_rate": 4.05e-07,
+      "loss": 0.0004,
+      "reward": 0.8626311421394348,
+      "reward_std": 0.3268212229013443,
+      "rewards/accuracy_reward": 0.0625,
+      "rewards/reasoning_steps_reward": 0.3680555820465088,
+      "rewards/repetition_penalty_reward": -0.09917446598410606,
+      "rewards/tag_count_reward": 0.53125,
       "step": 81
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 542.4791717529297,
-      "epoch": 0.082,
-      "grad_norm": 2.9728145387844758,
-      "kl": 0.01812744140625,
-      "learning_rate": 8.199999999999999e-07,
-      "loss": -0.0635,
-      "reward": 2.5557461977005005,
-      "reward_std": 0.4178060442209244,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.8680555820465088,
-      "rewards/repetition_penalty_reward": -0.04147614166140556,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 439.0833435058594,
+      "epoch": 0.041,
+      "grad_norm": 3.0805939483605678,
+      "kl": 0.0042877197265625,
+      "learning_rate": 4.0999999999999994e-07,
+      "loss": -0.019,
+      "reward": 0.9204289317131042,
+      "reward_std": 0.3671407848596573,
+      "rewards/accuracy_reward": 0.08333333395421505,
+      "rewards/reasoning_steps_reward": 0.4027777910232544,
+      "rewards/repetition_penalty_reward": -0.10214058682322502,
+      "rewards/tag_count_reward": 0.5364583432674408,
       "step": 82
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 540.0208435058594,
-      "epoch": 0.083,
-      "grad_norm": 2.624113800462916,
-      "kl": 0.02069091796875,
-      "learning_rate": 8.299999999999999e-07,
-      "loss": -0.0095,
-      "reward": 2.4058659076690674,
-      "reward_std": 0.49703338742256165,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.7708334028720856,
-      "rewards/repetition_penalty_reward": -0.04205096513032913,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 521.9166870117188,
+      "epoch": 0.0415,
+      "grad_norm": 2.597320274415619,
+      "kl": 0.0041046142578125,
+      "learning_rate": 4.1499999999999994e-07,
+      "loss": 0.0645,
+      "reward": 1.1805400252342224,
+      "reward_std": 0.43789660930633545,
+      "rewards/accuracy_reward": 0.16666667722165585,
+      "rewards/reasoning_steps_reward": 0.5277778059244156,
+      "rewards/repetition_penalty_reward": -0.08682116121053696,
+      "rewards/tag_count_reward": 0.5729166865348816,
       "step": 83
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 538.7291870117188,
-      "epoch": 0.084,
-      "grad_norm": 2.654590798347164,
-      "kl": 0.0184326171875,
-      "learning_rate": 8.399999999999999e-07,
-      "loss": -0.0225,
-      "reward": 2.4977999925613403,
-      "reward_std": 0.5065296292304993,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.847222238779068,
-      "rewards/repetition_penalty_reward": -0.03692223224788904,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 475.2708435058594,
+      "epoch": 0.042,
+      "grad_norm": 2.652905559142146,
+      "kl": 0.004302978515625,
+      "learning_rate": 4.1999999999999995e-07,
+      "loss": -0.0013,
+      "reward": 1.1051982939243317,
+      "reward_std": 0.4141012579202652,
+      "rewards/accuracy_reward": 0.12500000558793545,
+      "rewards/reasoning_steps_reward": 0.5138888657093048,
+      "rewards/repetition_penalty_reward": -0.09619061276316643,
+      "rewards/tag_count_reward": 0.5625000298023224,
       "step": 84
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.7916870117188,
-      "epoch": 0.085,
-      "grad_norm": 2.5236300531262033,
-      "kl": 0.02142333984375,
-      "learning_rate": 8.499999999999999e-07,
-      "loss": 0.0286,
-      "reward": 2.3374500274658203,
-      "reward_std": 0.4774511754512787,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.854166716337204,
-      "rewards/repetition_penalty_reward": -0.03755011223256588,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 401.0833435058594,
+      "epoch": 0.0425,
+      "grad_norm": 2.7752719784771798,
+      "kl": 0.00518798828125,
+      "learning_rate": 4.2499999999999995e-07,
+      "loss": -0.0391,
+      "reward": 1.0055087208747864,
+      "reward_std": 0.2745797038078308,
+      "rewards/accuracy_reward": 0.02083333395421505,
+      "rewards/reasoning_steps_reward": 0.5555556118488312,
+      "rewards/repetition_penalty_reward": -0.09171349555253983,
+      "rewards/tag_count_reward": 0.5208333730697632,
       "step": 85
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 557.6458435058594,
-      "epoch": 0.086,
-      "grad_norm": 2.4520679561302434,
-      "kl": 0.01904296875,
-      "learning_rate": 8.599999999999999e-07,
-      "loss": -0.0126,
-      "reward": 2.469316601753235,
-      "reward_std": 0.3855314701795578,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.8958333432674408,
-      "rewards/repetition_penalty_reward": -0.046308472752571106,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 419.4791717529297,
+      "epoch": 0.043,
+      "grad_norm": 2.640366127433102,
+      "kl": 0.0047607421875,
+      "learning_rate": 4.2999999999999996e-07,
+      "loss": 0.0125,
+      "reward": 1.0041911602020264,
+      "reward_std": 0.25805340707302094,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.5486111342906952,
+      "rewards/repetition_penalty_reward": -0.12254500389099121,
+      "rewards/tag_count_reward": 0.5364583432674408,
       "step": 86
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 537.1458435058594,
-      "epoch": 0.087,
-      "grad_norm": 2.8696069886274858,
-      "kl": 0.02032470703125,
-      "learning_rate": 8.699999999999999e-07,
-      "loss": 0.0288,
-      "reward": 2.775725483894348,
-      "reward_std": 0.3478439748287201,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9097222685813904,
-      "rewards/repetition_penalty_reward": -0.04545513913035393,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 428.2083435058594,
+      "epoch": 0.0435,
+      "grad_norm": 2.780462703093036,
+      "kl": 0.0068511962890625,
+      "learning_rate": 4.3499999999999996e-07,
+      "loss": -0.0254,
+      "reward": 0.9412573575973511,
+      "reward_std": 0.3015138581395149,
+      "rewards/accuracy_reward": 0.0625,
+      "rewards/reasoning_steps_reward": 0.4722222536802292,
+      "rewards/repetition_penalty_reward": -0.1247149184346199,
+      "rewards/tag_count_reward": 0.53125,
       "step": 87
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 552.0625152587891,
-      "epoch": 0.088,
-      "grad_norm": 2.7328359228843984,
-      "kl": 0.0185546875,
-      "learning_rate": 8.799999999999999e-07,
-      "loss": 0.1113,
-      "reward": 2.5936840772628784,
-      "reward_std": 0.2873719036579132,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.881944477558136,
-      "rewards/repetition_penalty_reward": -0.03826039098203182,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 452.4583435058594,
+      "epoch": 0.044,
+      "grad_norm": 2.9854860146208924,
+      "kl": 0.0058135986328125,
+      "learning_rate": 4.3999999999999997e-07,
+      "loss": -0.0644,
+      "reward": 1.1855429410934448,
+      "reward_std": 0.4236704409122467,
+      "rewards/accuracy_reward": 0.12500000558793545,
+      "rewards/reasoning_steps_reward": 0.604166716337204,
+      "rewards/repetition_penalty_reward": -0.10091547667980194,
+      "rewards/tag_count_reward": 0.5572916865348816,
       "step": 88
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 540.8125305175781,
-      "epoch": 0.089,
-      "grad_norm": 2.6685590803479307,
-      "kl": 0.02239990234375,
-      "learning_rate": 8.9e-07,
-      "loss": 0.0224,
-      "reward": 2.3886401653289795,
-      "reward_std": 0.48963838815689087,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.9305555522441864,
-      "rewards/repetition_penalty_reward": -0.03670708276331425,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 460.37501525878906,
+      "epoch": 0.0445,
+      "grad_norm": 2.6660126680917506,
+      "kl": 0.0048370361328125,
+      "learning_rate": 4.45e-07,
+      "loss": 0.0538,
+      "reward": 1.1993393301963806,
+      "reward_std": 0.45567750930786133,
+      "rewards/accuracy_reward": 0.2083333432674408,
+      "rewards/reasoning_steps_reward": 0.4583333432674408,
+      "rewards/repetition_penalty_reward": -0.11316069215536118,
+      "rewards/tag_count_reward": 0.6458333432674408,
       "step": 89
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 490.06251525878906,
-      "epoch": 0.09,
-      "grad_norm": 3.839427427471072,
-      "kl": 0.024169921875,
-      "learning_rate": 9e-07,
-      "loss": 0.0817,
-      "reward": 2.394170045852661,
-      "reward_std": 0.5097104758024216,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.8680555820465088,
-      "rewards/repetition_penalty_reward": -0.03638560324907303,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 503.18751525878906,
+      "epoch": 0.045,
+      "grad_norm": 2.6170875678464642,
+      "kl": 0.0070648193359375,
+      "learning_rate": 4.5e-07,
+      "loss": -0.0717,
+      "reward": 1.1190320253372192,
+      "reward_std": 0.36526739597320557,
+      "rewards/accuracy_reward": 0.0625,
+      "rewards/reasoning_steps_reward": 0.6111111640930176,
+      "rewards/repetition_penalty_reward": -0.08582913875579834,
+      "rewards/tag_count_reward": 0.53125,
       "step": 90
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 524.9791870117188,
-      "epoch": 0.091,
-      "grad_norm": 2.553614340168228,
-      "kl": 0.02398681640625,
-      "learning_rate": 9.1e-07,
-      "loss": 0.0725,
-      "reward": 2.6662930250167847,
-      "reward_std": 0.20492929220199585,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9097222685813904,
-      "rewards/repetition_penalty_reward": -0.03509608097374439,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 437.7083435058594,
+      "epoch": 0.0455,
+      "grad_norm": 2.6475065375904956,
+      "kl": 0.00689697265625,
+      "learning_rate": 4.55e-07,
+      "loss": 0.0389,
+      "reward": 1.283356487751007,
+      "reward_std": 0.5063433349132538,
+      "rewards/accuracy_reward": 0.1666666716337204,
+      "rewards/reasoning_steps_reward": 0.5833333730697632,
+      "rewards/repetition_penalty_reward": -0.08122700080275536,
+      "rewards/tag_count_reward": 0.6145833432674408,
       "step": 91
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 515.0625305175781,
-      "epoch": 0.092,
-      "grad_norm": 2.79663742057304,
-      "kl": 0.02203369140625,
-      "learning_rate": 9.2e-07,
-      "loss": 0.0525,
-      "reward": 2.7029329538345337,
-      "reward_std": 0.3332049995660782,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.04185891151428223,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 419.4583435058594,
+      "epoch": 0.046,
+      "grad_norm": 2.893081594976591,
+      "kl": 0.0062255859375,
+      "learning_rate": 4.6e-07,
+      "loss": 0.0274,
+      "reward": 1.465721845626831,
+      "reward_std": 0.415210023522377,
+      "rewards/accuracy_reward": 0.3333333432674408,
+      "rewards/reasoning_steps_reward": 0.5277778208255768,
+      "rewards/repetition_penalty_reward": -0.08288927376270294,
+      "rewards/tag_count_reward": 0.6875000298023224,
       "step": 92
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 586.6875,
-      "epoch": 0.093,
-      "grad_norm": 2.51813341735105,
-      "kl": 0.02301025390625,
-      "learning_rate": 9.3e-07,
-      "loss": 0.0461,
-      "reward": 2.7553551197052,
-      "reward_std": 0.41791823506355286,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.04499218240380287,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 496.0833435058594,
+      "epoch": 0.0465,
+      "grad_norm": 3.0157369307824924,
+      "kl": 0.0055694580078125,
+      "learning_rate": 4.65e-07,
+      "loss": -0.0453,
+      "reward": 1.174148678779602,
+      "reward_std": 0.43805110454559326,
+      "rewards/accuracy_reward": 0.14583333395421505,
+      "rewards/reasoning_steps_reward": 0.5069444626569748,
+      "rewards/repetition_penalty_reward": -0.07758747413754463,
+      "rewards/tag_count_reward": 0.5989583432674408,
       "step": 93
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.2708435058594,
-      "epoch": 0.094,
-      "grad_norm": 2.5842043324661863,
-      "kl": 0.0238037109375,
-      "learning_rate": 9.399999999999999e-07,
-      "loss": 0.0528,
-      "reward": 2.252669334411621,
-      "reward_std": 0.3811512589454651,
-      "rewards/accuracy_reward": 0.3333333432674408,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.04073341749608517,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 400.6458435058594,
+      "epoch": 0.047,
+      "grad_norm": 3.021710643706005,
+      "kl": 0.00909423828125,
+      "learning_rate": 4.6999999999999995e-07,
+      "loss": 0.0763,
+      "reward": 1.587534487247467,
+      "reward_std": 0.6594474613666534,
+      "rewards/accuracy_reward": 0.4375000149011612,
+      "rewards/reasoning_steps_reward": 0.479166716337204,
+      "rewards/repetition_penalty_reward": -0.07392388582229614,
+      "rewards/tag_count_reward": 0.7447916865348816,
       "step": 94
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 564.7708435058594,
-      "epoch": 0.095,
-      "grad_norm": 2.5488718443741667,
-      "kl": 0.02752685546875,
-      "learning_rate": 9.499999999999999e-07,
-      "loss": 0.1107,
-      "reward": 2.6501858234405518,
-      "reward_std": 0.27840781956911087,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.902777761220932,
-      "rewards/repetition_penalty_reward": -0.04425876401364803,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 419.3958435058594,
+      "epoch": 0.0475,
+      "grad_norm": 2.8068774316465617,
+      "kl": 0.0103759765625,
+      "learning_rate": 4.7499999999999995e-07,
+      "loss": -0.0172,
+      "reward": 1.508695363998413,
+      "reward_std": 0.5535295158624649,
+      "rewards/accuracy_reward": 0.3333333432674408,
+      "rewards/reasoning_steps_reward": 0.5694444477558136,
+      "rewards/repetition_penalty_reward": -0.0763741172850132,
+      "rewards/tag_count_reward": 0.6822916865348816,
       "step": 95
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 559.2708740234375,
-      "epoch": 0.096,
-      "grad_norm": 2.631139494749351,
-      "kl": 0.02752685546875,
-      "learning_rate": 9.6e-07,
-      "loss": 0.017,
-      "reward": 2.412258505821228,
-      "reward_std": 0.4537012577056885,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.04607492312788963,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 470.35418701171875,
+      "epoch": 0.048,
+      "grad_norm": 2.8744733124062676,
+      "kl": 0.0072174072265625,
+      "learning_rate": 4.8e-07,
+      "loss": -0.0036,
+      "reward": 1.1895955204963684,
+      "reward_std": 0.4534749835729599,
+      "rewards/accuracy_reward": 0.1875000074505806,
+      "rewards/reasoning_steps_reward": 0.3888889104127884,
+      "rewards/repetition_penalty_reward": -0.11075177043676376,
+      "rewards/tag_count_reward": 0.7239583432674408,
       "step": 96
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 606.6666870117188,
-      "epoch": 0.097,
-      "grad_norm": 2.8652564620163288,
-      "kl": 0.02764892578125,
-      "learning_rate": 9.7e-07,
-      "loss": 0.0402,
-      "reward": 2.6432669162750244,
-      "reward_std": 0.47565995156764984,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9027778208255768,
-      "rewards/repetition_penalty_reward": -0.04596925526857376,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 427.00001525878906,
+      "epoch": 0.0485,
+      "grad_norm": 2.820798846547903,
+      "kl": 0.010498046875,
+      "learning_rate": 4.85e-07,
+      "loss": 0.0096,
+      "reward": 1.331734836101532,
+      "reward_std": 0.6030288934707642,
+      "rewards/accuracy_reward": 0.3125,
+      "rewards/reasoning_steps_reward": 0.4027778059244156,
+      "rewards/repetition_penalty_reward": -0.10750139504671097,
+      "rewards/tag_count_reward": 0.7239583432674408,
       "step": 97
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 506.4583435058594,
-      "epoch": 0.098,
-      "grad_norm": 2.701632712968642,
-      "kl": 0.032470703125,
-      "learning_rate": 9.8e-07,
-      "loss": -0.0494,
-      "reward": 2.5521618127822876,
-      "reward_std": 0.4033654034137726,
-      "rewards/accuracy_reward": 0.708333358168602,
-      "rewards/reasoning_steps_reward": 0.8888889253139496,
-      "rewards/repetition_penalty_reward": -0.0450606532394886,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 486.1666717529297,
+      "epoch": 0.049,
+      "grad_norm": 2.607464524310001,
+      "kl": 0.010772705078125,
+      "learning_rate": 4.9e-07,
+      "loss": 0.0265,
+      "reward": 1.5102072954177856,
+      "reward_std": 0.5378637164831161,
+      "rewards/accuracy_reward": 0.458333358168602,
+      "rewards/reasoning_steps_reward": 0.298611119389534,
+      "rewards/repetition_penalty_reward": -0.12173716351389885,
+      "rewards/tag_count_reward": 0.8750000298023224,
       "step": 98
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 570.2916870117188,
-      "epoch": 0.099,
-      "grad_norm": 2.578693808208252,
-      "kl": 0.03271484375,
-      "learning_rate": 9.9e-07,
-      "loss": -0.0427,
-      "reward": 2.4133933782577515,
-      "reward_std": 0.30207425355911255,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03973175957798958,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 504.04168701171875,
+      "epoch": 0.0495,
+      "grad_norm": 2.6750653025151925,
+      "kl": 0.013946533203125,
+      "learning_rate": 4.95e-07,
+      "loss": 0.0658,
+      "reward": 1.610666811466217,
+      "reward_std": 0.6144271492958069,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.3541667014360428,
+      "rewards/repetition_penalty_reward": -0.10287501662969589,
+      "rewards/tag_count_reward": 0.8593750298023224,
       "step": 99
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 614.3333740234375,
-      "epoch": 0.1,
-      "grad_norm": 2.67358470309149,
-      "kl": 0.028564453125,
-      "learning_rate": 1e-06,
-      "loss": -0.0598,
-      "reward": 2.3288190364837646,
-      "reward_std": 0.3720841705799103,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.8819445073604584,
-      "rewards/repetition_penalty_reward": -0.05312554910778999,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 510.72918701171875,
+      "epoch": 0.05,
+      "grad_norm": 2.7156863679632885,
+      "kl": 0.009613037109375,
+      "learning_rate": 5e-07,
+      "loss": 0.2283,
+      "reward": 1.464966893196106,
+      "reward_std": 0.5712087154388428,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.3402777910232544,
+      "rewards/repetition_penalty_reward": -0.1096859984099865,
+      "rewards/tag_count_reward": 0.8177083432674408,
       "step": 100
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.4375,
-      "epoch": 0.101,
-      "grad_norm": 2.557702270091156,
-      "kl": 0.0347900390625,
-      "learning_rate": 9.999972584460056e-07,
-      "loss": 0.0339,
-      "reward": 2.7004672288894653,
-      "reward_std": 0.3648904711008072,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.92361119389534,
-      "rewards/repetition_penalty_reward": -0.05126911960542202,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 423.0416717529297,
+      "epoch": 0.0505,
+      "grad_norm": 2.919566854448967,
+      "kl": 0.0150146484375,
+      "learning_rate": 5.049999999999999e-07,
+      "loss": -0.0017,
+      "reward": 1.971518337726593,
+      "reward_std": 0.5628243684768677,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.375,
+      "rewards/repetition_penalty_reward": -0.08056501299142838,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 101
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 596.1458435058594,
-      "epoch": 0.102,
-      "grad_norm": 2.7419447112432644,
-      "kl": 0.033935546875,
-      "learning_rate": 9.999890338174275e-07,
-      "loss": 0.0276,
-      "reward": 2.4844894409179688,
-      "reward_std": 0.4377788305282593,
-      "rewards/accuracy_reward": 0.583333358168602,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.043288541957736015,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 370.5416717529297,
+      "epoch": 0.051,
+      "grad_norm": 2.864551722431305,
+      "kl": 0.017913818359375,
+      "learning_rate": 5.1e-07,
+      "loss": 0.037,
+      "reward": 1.7677536010742188,
+      "reward_std": 0.4400963932275772,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.3194444626569748,
+      "rewards/repetition_penalty_reward": -0.07252424210309982,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 102
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 549.3125305175781,
-      "epoch": 0.103,
-      "grad_norm": 2.5449102277843556,
-      "kl": 0.0384521484375,
-      "learning_rate": 9.999753262144804e-07,
-      "loss": 0.091,
-      "reward": 2.214192271232605,
-      "reward_std": 0.41757629811763763,
-      "rewards/accuracy_reward": 0.3333333358168602,
-      "rewards/reasoning_steps_reward": 0.930555522441864,
-      "rewards/repetition_penalty_reward": -0.04969660937786102,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 413.66668701171875,
+      "epoch": 0.0515,
+      "grad_norm": 2.962558584851364,
+      "kl": 0.012969970703125,
+      "learning_rate": 5.149999999999999e-07,
+      "loss": -0.0576,
+      "reward": 1.84916353225708,
+      "reward_std": 0.4907376766204834,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.3263888955116272,
+      "rewards/repetition_penalty_reward": -0.07097543030977249,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 103
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 503.6458435058594,
-      "epoch": 0.104,
-      "grad_norm": 2.5988919390766134,
-      "kl": 0.038330078125,
-      "learning_rate": 9.999561358041868e-07,
-      "loss": -0.0306,
-      "reward": 2.473710536956787,
-      "reward_std": 0.36740100383758545,
-      "rewards/accuracy_reward": 0.5833333730697632,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.04712275043129921,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 383.3958435058594,
+      "epoch": 0.052,
+      "grad_norm": 3.174614828168193,
+      "kl": 0.015167236328125,
+      "learning_rate": 5.2e-07,
+      "loss": -0.0749,
+      "reward": 1.8075725436210632,
+      "reward_std": 0.28401315957307816,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.2916666865348816,
+      "rewards/repetition_penalty_reward": -0.057010795921087265,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 104
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 584.5833435058594,
-      "epoch": 0.105,
-      "grad_norm": 2.2253173220096847,
-      "kl": 0.03759765625,
-      "learning_rate": 9.99931462820376e-07,
-      "loss": -0.0418,
-      "reward": 2.6660202741622925,
-      "reward_std": 0.39214441180229187,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.04231319762766361,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 413.75,
+      "epoch": 0.0525,
+      "grad_norm": 2.9070150258769845,
+      "kl": 0.0205078125,
+      "learning_rate": 5.25e-07,
+      "loss": -0.0433,
+      "reward": 1.7146551609039307,
+      "reward_std": 0.5036576092243195,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.36805559694767,
+      "rewards/repetition_penalty_reward": -0.07527554780244827,
+      "rewards/tag_count_reward": 0.9427083730697632,
       "step": 105
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 580.5833435058594,
-      "epoch": 0.106,
-      "grad_norm": 2.740740338946049,
-      "kl": 0.03857421875,
-      "learning_rate": 9.999013075636804e-07,
-      "loss": -0.0301,
-      "reward": 2.671258807182312,
-      "reward_std": 0.35463356226682663,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.044019101187586784,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 506.22918701171875,
+      "epoch": 0.053,
+      "grad_norm": 2.7140962159705353,
+      "kl": 0.01116943359375,
+      "learning_rate": 5.3e-07,
+      "loss": -0.0411,
+      "reward": 1.7457041144371033,
+      "reward_std": 0.43025586009025574,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.236111119389534,
+      "rewards/repetition_penalty_reward": -0.0997820794582367,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 106
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 552.8958740234375,
-      "epoch": 0.107,
-      "grad_norm": 2.6710939233020636,
-      "kl": 0.0406494140625,
-      "learning_rate": 9.998656704015323e-07,
-      "loss": 0.0481,
-      "reward": 2.4425666332244873,
-      "reward_std": 0.36045634746551514,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.04528068192303181,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 390.7708435058594,
+      "epoch": 0.0535,
+      "grad_norm": 2.5890826142014376,
+      "kl": 0.016693115234375,
+      "learning_rate": 5.35e-07,
+      "loss": 0.048,
+      "reward": 1.9064030051231384,
+      "reward_std": 0.3108719140291214,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.2916666865348816,
+      "rewards/repetition_penalty_reward": -0.08838870003819466,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 107
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 521.7500305175781,
-      "epoch": 0.108,
-      "grad_norm": 2.9062446352100086,
-      "kl": 0.0419921875,
-      "learning_rate": 9.998245517681593e-07,
-      "loss": 0.0707,
-      "reward": 2.725650668144226,
-      "reward_std": 0.3536549210548401,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.05212709680199623,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 457.6458435058594,
+      "epoch": 0.054,
+      "grad_norm": 2.4716329573423796,
+      "kl": 0.01641845703125,
+      "learning_rate": 5.4e-07,
+      "loss": -0.0065,
+      "reward": 1.9147586822509766,
+      "reward_std": 0.30019962787628174,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.2986111342906952,
+      "rewards/repetition_penalty_reward": -0.08697742223739624,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 108
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 473.10418701171875,
-      "epoch": 0.109,
-      "grad_norm": 2.840139368335502,
-      "kl": 0.048583984375,
-      "learning_rate": 9.997779521645791e-07,
-      "loss": 0.0222,
-      "reward": 2.534466028213501,
-      "reward_std": 0.2846931293606758,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.028034010902047157,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 413.6458435058594,
+      "epoch": 0.0545,
+      "grad_norm": 2.954116545096682,
+      "kl": 0.0186767578125,
+      "learning_rate": 5.45e-07,
+      "loss": 0.0025,
+      "reward": 2.002891957759857,
+      "reward_std": 0.39341901242733,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.2430555671453476,
+      "rewards/repetition_penalty_reward": -0.06828875839710236,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 109
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 565.9375305175781,
-      "epoch": 0.11,
-      "grad_norm": 2.2512653144943884,
-      "kl": 0.0401611328125,
-      "learning_rate": 9.997258721585931e-07,
-      "loss": 0.0075,
-      "reward": 2.7475154399871826,
-      "reward_std": 0.3540599048137665,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.04935969039797783,
-      "rewards/tag_count_reward": 0.9843750298023224,
+      "completion_length": 431.93751525878906,
+      "epoch": 0.055,
+      "grad_norm": 2.796630595163655,
+      "kl": 0.013763427734375,
+      "learning_rate": 5.5e-07,
+      "loss": -0.0342,
+      "reward": 2.094782531261444,
+      "reward_std": 0.32828887552022934,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.395833358168602,
+      "rewards/repetition_penalty_reward": -0.10834262520074844,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 110
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 553.3125,
-      "epoch": 0.111,
-      "grad_norm": 2.527203584806038,
-      "kl": 0.04296875,
-      "learning_rate": 9.996683123847795e-07,
-      "loss": 0.0512,
-      "reward": 2.7313212156295776,
-      "reward_std": 0.2547169327735901,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.04124833457171917,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 455.0,
+      "epoch": 0.0555,
+      "grad_norm": 3.3329482247244755,
+      "kl": 0.013427734375,
+      "learning_rate": 5.55e-07,
+      "loss": 0.1221,
+      "reward": 2.00520521402359,
+      "reward_std": 0.48712170124053955,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.3541666865348816,
+      "rewards/repetition_penalty_reward": -0.07812817022204399,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 111
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 536.8958740234375,
-      "epoch": 0.112,
-      "grad_norm": 2.7442985147195307,
-      "kl": 0.044189453125,
-      "learning_rate": 9.996052735444862e-07,
-      "loss": 0.0349,
-      "reward": 2.7351317405700684,
-      "reward_std": 0.31766992807388306,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.03917396813631058,
-      "rewards/tag_count_reward": 0.9895833730697632,
+      "completion_length": 410.37501525878906,
+      "epoch": 0.056,
+      "grad_norm": 2.9159306101645845,
+      "kl": 0.01458740234375,
+      "learning_rate": 5.6e-07,
+      "loss": 0.0302,
+      "reward": 1.945899486541748,
+      "reward_std": 0.5274121165275574,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.3680555522441864,
+      "rewards/repetition_penalty_reward": -0.09923946484923363,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 112
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 633.9375,
-      "epoch": 0.113,
-      "grad_norm": 2.206627822892764,
-      "kl": 0.0399169921875,
-      "learning_rate": 9.995367564058216e-07,
-      "loss": 0.0002,
-      "reward": 2.4217275381088257,
-      "reward_std": 0.37440885603427887,
-      "rewards/accuracy_reward": 0.5000000223517418,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.050494687631726265,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 430.8958435058594,
+      "epoch": 0.0565,
+      "grad_norm": 3.0211550455327836,
+      "kl": 0.02276611328125,
+      "learning_rate": 5.649999999999999e-07,
+      "loss": 0.0349,
+      "reward": 2.0766223669052124,
+      "reward_std": 0.3339201509952545,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.3263888955116272,
+      "rewards/repetition_penalty_reward": -0.07789156958460808,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 113
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 559.5000305175781,
-      "epoch": 0.114,
-      "grad_norm": 2.9323115322073567,
-      "kl": 0.048583984375,
-      "learning_rate": 9.994627618036452e-07,
-      "loss": 0.0838,
-      "reward": 2.7629553079605103,
-      "reward_std": 0.3546500727534294,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.04260038584470749,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 421.3541717529297,
+      "epoch": 0.057,
+      "grad_norm": 2.946457972762416,
+      "kl": 0.0147705078125,
+      "learning_rate": 5.699999999999999e-07,
+      "loss": 0.012,
+      "reward": 1.9260443449020386,
+      "reward_std": 0.4190318286418915,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.402777761220932,
+      "rewards/repetition_penalty_reward": -0.07569188624620438,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 114
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 600.7291870117188,
-      "epoch": 0.115,
-      "grad_norm": 2.281029530481033,
-      "kl": 0.0433349609375,
-      "learning_rate": 9.993832906395582e-07,
-      "loss": 0.0435,
-      "reward": 2.6696189641952515,
-      "reward_std": 0.3239249736070633,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.05260329693555832,
+      "completion_length": 401.47918701171875,
+      "epoch": 0.0575,
+      "grad_norm": 2.776628513370855,
+      "kl": 0.01708984375,
+      "learning_rate": 5.749999999999999e-07,
+      "loss": -0.0217,
+      "reward": 2.1737382411956787,
+      "reward_std": 0.416288822889328,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.3680555820465088,
+      "rewards/repetition_penalty_reward": -0.09015080332756042,
       "rewards/tag_count_reward": 1.0,
       "step": 115
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 571.4791870117188,
-      "epoch": 0.116,
-      "grad_norm": 2.339412143248174,
-      "kl": 0.0447998046875,
-      "learning_rate": 9.992983438818915e-07,
-      "loss": -0.0459,
-      "reward": 2.695494771003723,
-      "reward_std": 0.3602631092071533,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.05276927165687084,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 462.1458435058594,
+      "epoch": 0.058,
+      "grad_norm": 2.5888150833378383,
+      "kl": 0.01837158203125,
+      "learning_rate": 5.8e-07,
+      "loss": -0.0648,
+      "reward": 2.0337727069854736,
+      "reward_std": 0.3629211187362671,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.291666679084301,
+      "rewards/repetition_penalty_reward": -0.09122735634446144,
+      "rewards/tag_count_reward": 1.0,
       "step": 116
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 569.2291870117188,
-      "epoch": 0.117,
-      "grad_norm": 2.6819516393809444,
-      "kl": 0.0435791015625,
-      "learning_rate": 9.992079225656944e-07,
-      "loss": 0.0445,
-      "reward": 2.677255153656006,
-      "reward_std": 0.2550307661294937,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.038022806867957115,
+      "completion_length": 395.1041717529297,
+      "epoch": 0.0585,
+      "grad_norm": 3.1651475575486065,
+      "kl": 0.02288818359375,
+      "learning_rate": 5.849999999999999e-07,
+      "loss": 0.0276,
+      "reward": 2.0250861048698425,
+      "reward_std": 0.385862335562706,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.284722238779068,
+      "rewards/repetition_penalty_reward": -0.07213617861270905,
       "rewards/tag_count_reward": 1.0,
       "step": 117
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 606.5,
-      "epoch": 0.118,
-      "grad_norm": 2.7553746000465016,
-      "kl": 0.0418701171875,
-      "learning_rate": 9.991120277927223e-07,
-      "loss": 0.0683,
-      "reward": 2.683447241783142,
-      "reward_std": 0.4267844557762146,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.045719537883996964,
+      "completion_length": 405.50001525878906,
+      "epoch": 0.059,
+      "grad_norm": 2.995948787512775,
+      "kl": 0.01568603515625,
+      "learning_rate": 5.9e-07,
+      "loss": 0.0545,
+      "reward": 2.079026162624359,
+      "reward_std": 0.4058743417263031,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.3680555745959282,
+      "rewards/repetition_penalty_reward": -0.08069606870412827,
       "rewards/tag_count_reward": 1.0,
       "step": 118
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.4583740234375,
-      "epoch": 0.119,
-      "grad_norm": 2.5283730612634168,
-      "kl": 0.0426025390625,
-      "learning_rate": 9.990106607314225e-07,
-      "loss": -0.0626,
-      "reward": 2.5398285388946533,
-      "reward_std": 0.37042136490345,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.05044914782047272,
+      "completion_length": 427.1458435058594,
+      "epoch": 0.0595,
+      "grad_norm": 3.2921667824300074,
+      "kl": 0.02056884765625,
+      "learning_rate": 5.949999999999999e-07,
+      "loss": 0.0068,
+      "reward": 2.110425293445587,
+      "reward_std": 0.4844990372657776,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.4305555671453476,
+      "rewards/repetition_penalty_reward": -0.07013046741485596,
       "rewards/tag_count_reward": 1.0,
       "step": 119
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 581.7500305175781,
-      "epoch": 0.12,
-      "grad_norm": 2.649956713864194,
-      "kl": 0.04931640625,
-      "learning_rate": 9.989038226169207e-07,
-      "loss": 0.0635,
-      "reward": 2.569565773010254,
-      "reward_std": 0.4911084771156311,
-      "rewards/accuracy_reward": 0.6458333730697632,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.04154558852314949,
+      "completion_length": 394.79168701171875,
+      "epoch": 0.06,
+      "grad_norm": 3.0710450999617622,
+      "kl": 0.01605224609375,
+      "learning_rate": 6e-07,
+      "loss": 0.0034,
+      "reward": 1.7129506468772888,
+      "reward_std": 0.3703618347644806,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.3472222238779068,
+      "rewards/repetition_penalty_reward": -0.07177170366048813,
       "rewards/tag_count_reward": 1.0,
       "step": 120
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 636.6666870117188,
-      "epoch": 0.121,
-      "grad_norm": 2.7429948870181233,
-      "kl": 0.0418701171875,
-      "learning_rate": 9.98791514751006e-07,
-      "loss": 0.13,
-      "reward": 2.663628339767456,
-      "reward_std": 0.396682009100914,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.060329992324113846,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 441.18751525878906,
+      "epoch": 0.0605,
+      "grad_norm": 2.7064381875532835,
+      "kl": 0.02081298828125,
+      "learning_rate": 6.049999999999999e-07,
+      "loss": 0.0009,
+      "reward": 1.8438260555267334,
+      "reward_std": 0.46181726455688477,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.3055555522441864,
+      "rewards/repetition_penalty_reward": -0.08672957122325897,
+      "rewards/tag_count_reward": 1.0,
       "step": 121
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 596.2083435058594,
-      "epoch": 0.122,
-      "grad_norm": 2.4221170538253816,
-      "kl": 0.045166015625,
-      "learning_rate": 9.98673738502114e-07,
-      "loss": 0.0342,
-      "reward": 2.6282721757888794,
-      "reward_std": 0.46036189794540405,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.04707522317767143,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 453.5833435058594,
+      "epoch": 0.061,
+      "grad_norm": 2.9942145930271735,
+      "kl": 0.01763916015625,
+      "learning_rate": 6.1e-07,
+      "loss": -0.0401,
+      "reward": 1.858453392982483,
+      "reward_std": 0.4963291585445404,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.361111119389534,
+      "rewards/repetition_penalty_reward": -0.07036612182855606,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 122
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 610.0416870117188,
-      "epoch": 0.123,
-      "grad_norm": 2.268992560964926,
-      "kl": 0.045166015625,
-      "learning_rate": 9.985504953053113e-07,
-      "loss": -0.0075,
-      "reward": 2.5075987577438354,
-      "reward_std": 0.4839261472225189,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.047956960275769234,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 440.72918701171875,
+      "epoch": 0.0615,
+      "grad_norm": 2.591891405826925,
+      "kl": 0.0142822265625,
+      "learning_rate": 6.149999999999999e-07,
+      "loss": 0.0201,
+      "reward": 2.2280489206314087,
+      "reward_std": 0.35533128678798676,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.4583333730697632,
+      "rewards/repetition_penalty_reward": -0.08445119112730026,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 123
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 637.8333435058594,
-      "epoch": 0.124,
-      "grad_norm": 2.3192412067695276,
-      "kl": 0.040283203125,
-      "learning_rate": 9.98421786662277e-07,
-      "loss": 0.0276,
-      "reward": 2.580567240715027,
-      "reward_std": 0.40886104106903076,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.060057852417230606,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 507.8958435058594,
+      "epoch": 0.062,
+      "grad_norm": 2.5201772155206448,
+      "kl": 0.01611328125,
+      "learning_rate": 6.2e-07,
+      "loss": 0.0014,
+      "reward": 2.040831744670868,
+      "reward_std": 0.43686023354530334,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.4375,
+      "rewards/repetition_penalty_reward": -0.10500159859657288,
+      "rewards/tag_count_reward": 1.0,
       "step": 124
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 550.6875305175781,
-      "epoch": 0.125,
-      "grad_norm": 2.6428669262747935,
-      "kl": 0.0528564453125,
-      "learning_rate": 9.982876141412855e-07,
-      "loss": -0.021,
-      "reward": 2.6972975730895996,
-      "reward_std": 0.30313703417778015,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.0353414136916399,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 403.1041717529297,
+      "epoch": 0.0625,
+      "grad_norm": 2.930044284566675,
+      "kl": 0.0194091796875,
+      "learning_rate": 6.249999999999999e-07,
+      "loss": 0.0418,
+      "reward": 1.9074211716651917,
+      "reward_std": 0.5366443991661072,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.3611111342906952,
+      "rewards/repetition_penalty_reward": -0.08389833942055702,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 125
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 680.6041870117188,
-      "epoch": 0.126,
-      "grad_norm": 2.3227570325460047,
-      "kl": 0.0430908203125,
-      "learning_rate": 9.981479793771866e-07,
-      "loss": 0.0752,
-      "reward": 2.591596007347107,
-      "reward_std": 0.49062085151672363,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.04902906343340874,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 401.6458435058594,
+      "epoch": 0.063,
+      "grad_norm": 3.0687595839670165,
+      "kl": 0.02783203125,
+      "learning_rate": 6.3e-07,
+      "loss": -0.1223,
+      "reward": 2.101858079433441,
+      "reward_std": 0.38748544454574585,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.3958333879709244,
+      "rewards/repetition_penalty_reward": -0.09085042029619217,
+      "rewards/tag_count_reward": 0.984375,
       "step": 126
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 567.7291870117188,
-      "epoch": 0.127,
-      "grad_norm": 2.5341618125591894,
-      "kl": 0.0494384765625,
-      "learning_rate": 9.98002884071386e-07,
-      "loss": 0.0338,
-      "reward": 2.735344886779785,
-      "reward_std": 0.3433926999568939,
+      "completion_length": 404.7291717529297,
+      "epoch": 0.0635,
+      "grad_norm": 2.876756319174346,
+      "kl": 0.019561767578125,
+      "learning_rate": 6.35e-07,
+      "loss": 0.0507,
+      "reward": 2.075514793395996,
+      "reward_std": 0.3558611124753952,
       "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.04243295639753342,
+      "rewards/reasoning_steps_reward": 0.3541666716337204,
+      "rewards/repetition_penalty_reward": -0.0703187882900238,
       "rewards/tag_count_reward": 1.0,
       "step": 127
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 611.7291870117188,
-      "epoch": 0.128,
-      "grad_norm": 84.62793212916526,
-      "kl": 0.50634765625,
-      "learning_rate": 9.97852329991824e-07,
-      "loss": 0.0161,
-      "reward": 2.649895191192627,
-      "reward_std": 0.4249277561903,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.04628550261259079,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 483.1875,
+      "epoch": 0.064,
+      "grad_norm": 2.722649342936597,
+      "kl": 0.0159912109375,
+      "learning_rate": 6.4e-07,
+      "loss": 0.0157,
+      "reward": 1.942257821559906,
+      "reward_std": 0.45219163596630096,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.4375000149011612,
+      "rewards/repetition_penalty_reward": -0.09940891712903976,
+      "rewards/tag_count_reward": 1.0,
       "step": 128
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 587.2083740234375,
-      "epoch": 0.129,
-      "grad_norm": 8.417728255751756,
-      "kl": 0.0889892578125,
-      "learning_rate": 9.976963189729547e-07,
-      "loss": 0.0735,
-      "reward": 2.7577908039093018,
-      "reward_std": 0.3766750693321228,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.04082038067281246,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 428.5416717529297,
+      "epoch": 0.0645,
+      "grad_norm": 2.721100537467751,
+      "kl": 0.02197265625,
+      "learning_rate": 6.45e-07,
+      "loss": -0.0335,
+      "reward": 1.7678640484809875,
+      "reward_std": 0.4747384488582611,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.3125000149011612,
+      "rewards/repetition_penalty_reward": -0.0706777349114418,
+      "rewards/tag_count_reward": 0.984375,
       "step": 129
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 592.8958435058594,
-      "epoch": 0.13,
-      "grad_norm": 2.422421343225394,
-      "kl": 0.047607421875,
-      "learning_rate": 9.975348529157229e-07,
-      "loss": 0.0476,
-      "reward": 2.71326220035553,
-      "reward_std": 0.3403441533446312,
+      "completion_length": 435.2083435058594,
+      "epoch": 0.065,
+      "grad_norm": 2.617452029309224,
+      "kl": 0.0174560546875,
+      "learning_rate": 6.5e-07,
+      "loss": -0.0038,
+      "reward": 2.130853533744812,
+      "reward_std": 0.4503345489501953,
       "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.05236278846859932,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "rewards/reasoning_steps_reward": 0.4166666865348816,
+      "rewards/repetition_penalty_reward": -0.07747986912727356,
+      "rewards/tag_count_reward": 1.0,
       "step": 130
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 592.1250305175781,
-      "epoch": 0.131,
-      "grad_norm": 3.1437399948807276,
-      "kl": 0.0548095703125,
-      "learning_rate": 9.973679337875418e-07,
-      "loss": 0.0466,
-      "reward": 2.6631596088409424,
-      "reward_std": 0.37047071754932404,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.04170159809291363,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 397.7083435058594,
+      "epoch": 0.0655,
+      "grad_norm": 2.808732238238703,
+      "kl": 0.02032470703125,
+      "learning_rate": 6.55e-07,
+      "loss": -0.011,
+      "reward": 2.1196643710136414,
+      "reward_std": 0.40009158849716187,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.4027778059244156,
+      "rewards/repetition_penalty_reward": -0.07478011026978493,
+      "rewards/tag_count_reward": 1.0,
       "step": 131
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 531.9791870117188,
-      "epoch": 0.132,
-      "grad_norm": 2.5320955473297437,
-      "kl": 0.052490234375,
-      "learning_rate": 9.971955636222684e-07,
-      "loss": 0.0346,
-      "reward": 2.73369300365448,
-      "reward_std": 0.3214843422174454,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.04408488981425762,
+      "completion_length": 428.5625,
+      "epoch": 0.066,
+      "grad_norm": 2.6151338968076194,
+      "kl": 0.0184326171875,
+      "learning_rate": 6.6e-07,
+      "loss": 0.0189,
+      "reward": 1.877784252166748,
+      "reward_std": 0.3097234219312668,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.2916666865348816,
+      "rewards/repetition_penalty_reward": -0.08054918050765991,
       "rewards/tag_count_reward": 1.0,
       "step": 132
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 654.1875305175781,
-      "epoch": 0.133,
-      "grad_norm": 2.3742871546515545,
-      "kl": 0.047607421875,
-      "learning_rate": 9.970177445201783e-07,
-      "loss": 0.0042,
-      "reward": 2.523491144180298,
-      "reward_std": 0.49983178079128265,
-      "rewards/accuracy_reward": 0.5833333730697632,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.045953478664159775,
+      "completion_length": 434.8333435058594,
+      "epoch": 0.0665,
+      "grad_norm": 3.154952772902957,
+      "kl": 0.02117919921875,
+      "learning_rate": 6.65e-07,
+      "loss": 0.0818,
+      "reward": 2.2022292613983154,
+      "reward_std": 0.3770638406276703,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.4375000149011612,
+      "rewards/repetition_penalty_reward": -0.06860406324267387,
       "rewards/tag_count_reward": 1.0,
       "step": 133
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 567.6458740234375,
-      "epoch": 0.134,
-      "grad_norm": 2.5745090813340257,
-      "kl": 0.052734375,
-      "learning_rate": 9.968344786479415e-07,
-      "loss": 0.0497,
-      "reward": 2.654099702835083,
-      "reward_std": 0.32977308332920074,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.05770585313439369,
-      "rewards/tag_count_reward": 0.9895833730697632,
+      "completion_length": 401.87501525878906,
+      "epoch": 0.067,
+      "grad_norm": 2.764724446368728,
+      "kl": 0.01763916015625,
+      "learning_rate": 6.7e-07,
+      "loss": -0.0843,
+      "reward": 1.9817876815795898,
+      "reward_std": 0.39721402525901794,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.409722238779068,
+      "rewards/repetition_penalty_reward": -0.09460122510790825,
+      "rewards/tag_count_reward": 1.0,
       "step": 134
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 591.4375305175781,
-      "epoch": 0.135,
-      "grad_norm": 2.613972316646787,
-      "kl": 0.0504150390625,
-      "learning_rate": 9.96645768238595e-07,
-      "loss": 0.0264,
-      "reward": 2.715443730354309,
-      "reward_std": 0.4562966376543045,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.04323691129684448,
+      "completion_length": 465.7708435058594,
+      "epoch": 0.0675,
+      "grad_norm": 2.3989523347162534,
+      "kl": 0.01953125,
+      "learning_rate": 6.75e-07,
+      "loss": -0.0177,
+      "reward": 2.228469967842102,
+      "reward_std": 0.3507627248764038,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.534722238779068,
+      "rewards/repetition_penalty_reward": -0.0927107036113739,
       "rewards/tag_count_reward": 0.9947916865348816,
       "step": 135
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 608.5208435058594,
-      "epoch": 0.136,
-      "grad_norm": 2.405934822243009,
-      "kl": 0.0482177734375,
-      "learning_rate": 9.964516155915151e-07,
-      "loss": 0.0707,
-      "reward": 2.7309588193893433,
-      "reward_std": 0.42438623309135437,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.06244398467242718,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 391.06251525878906,
+      "epoch": 0.068,
+      "grad_norm": 3.0855932954768965,
+      "kl": 0.01800537109375,
+      "learning_rate": 6.800000000000001e-07,
+      "loss": 0.0166,
+      "reward": 2.255845546722412,
+      "reward_std": 0.4071827381849289,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.4166666716337204,
+      "rewards/repetition_penalty_reward": -0.07748781517148018,
+      "rewards/tag_count_reward": 1.0,
       "step": 136
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 698.9375305175781,
-      "epoch": 0.137,
-      "grad_norm": 2.4898466567094255,
-      "kl": 0.041259765625,
-      "learning_rate": 9.962520230723906e-07,
-      "loss": -0.0047,
-      "reward": 2.7930511236190796,
-      "reward_std": 0.3140558898448944,
-      "rewards/accuracy_reward": 0.8958333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.06458798050880432,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 446.2708435058594,
+      "epoch": 0.0685,
+      "grad_norm": 2.9929865598477066,
+      "kl": 0.019287109375,
+      "learning_rate": 6.85e-07,
+      "loss": -0.0368,
+      "reward": 2.1877795457839966,
+      "reward_std": 0.5341714322566986,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.4513889104127884,
+      "rewards/repetition_penalty_reward": -0.07610936462879181,
+      "rewards/tag_count_reward": 1.0,
       "step": 137
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 529.0208435058594,
-      "epoch": 0.138,
-      "grad_norm": 2.8868557758723306,
-      "kl": 0.0540771484375,
-      "learning_rate": 9.960469931131936e-07,
-      "loss": 0.0636,
-      "reward": 2.6102973222732544,
-      "reward_std": 0.42000116407871246,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.0372721990570426,
+      "completion_length": 397.7083435058594,
+      "epoch": 0.069,
+      "grad_norm": 2.954697616837276,
+      "kl": 0.02471923828125,
+      "learning_rate": 6.9e-07,
+      "loss": 0.009,
+      "reward": 1.824935793876648,
+      "reward_std": 0.49531859159469604,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.5000000596046448,
+      "rewards/repetition_penalty_reward": -0.08652260527014732,
       "rewards/tag_count_reward": 0.9947916865348816,
       "step": 138
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 525.6875305175781,
-      "epoch": 0.139,
-      "grad_norm": 2.8464042704145114,
-      "kl": 0.058349609375,
-      "learning_rate": 9.958365282121496e-07,
-      "loss": 0.0587,
-      "reward": 2.6924026012420654,
-      "reward_std": 0.3652355223894119,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.04370862618088722,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 548.0625152587891,
+      "epoch": 0.0695,
+      "grad_norm": 2.506876897622573,
+      "kl": 0.014892578125,
+      "learning_rate": 6.949999999999999e-07,
+      "loss": -0.054,
+      "reward": 1.6296937465667725,
+      "reward_std": 0.4967469274997711,
+      "rewards/accuracy_reward": 0.2708333432674408,
+      "rewards/reasoning_steps_reward": 0.451388880610466,
+      "rewards/repetition_penalty_reward": -0.08732018992304802,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 139
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 623.0416870117188,
-      "epoch": 0.14,
-      "grad_norm": 2.5535766431944555,
-      "kl": 0.05078125,
-      "learning_rate": 9.956206309337066e-07,
-      "loss": -0.0369,
-      "reward": 2.5890008211135864,
-      "reward_std": 0.43424585461616516,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.049888189882040024,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 435.0833435058594,
+      "epoch": 0.07,
+      "grad_norm": 2.7570820251386037,
+      "kl": 0.02325439453125,
+      "learning_rate": 7e-07,
+      "loss": 0.0036,
+      "reward": 2.1854411363601685,
+      "reward_std": 0.4105776250362396,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.5138888657093048,
+      "rewards/repetition_penalty_reward": -0.0732395276427269,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 140
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.7291870117188,
-      "epoch": 0.141,
-      "grad_norm": 2.5331968568084804,
-      "kl": 0.055419921875,
-      "learning_rate": 9.953993039085048e-07,
-      "loss": 0.0256,
-      "reward": 2.5839684009552,
-      "reward_std": 0.30562296509742737,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.0427678357809782,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 388.2291717529297,
+      "epoch": 0.0705,
+      "grad_norm": 2.6297348721129463,
+      "kl": 0.0208740234375,
+      "learning_rate": 7.049999999999999e-07,
+      "loss": -0.0319,
+      "reward": 1.8632826209068298,
+      "reward_std": 0.35653068125247955,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.4652777910232544,
+      "rewards/repetition_penalty_reward": -0.10199519246816635,
+      "rewards/tag_count_reward": 1.0,
       "step": 141
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 631.0625152587891,
-      "epoch": 0.142,
-      "grad_norm": 3.9726646160122283,
-      "kl": 0.0572509765625,
-      "learning_rate": 9.951725498333448e-07,
-      "loss": 0.0009,
-      "reward": 2.697226047515869,
-      "reward_std": 0.40920713543891907,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.03888525255024433,
+      "completion_length": 405.81251525878906,
+      "epoch": 0.071,
+      "grad_norm": 3.135261611976366,
+      "kl": 0.02117919921875,
+      "learning_rate": 7.1e-07,
+      "loss": -0.0326,
+      "reward": 2.321570873260498,
+      "reward_std": 0.41436049342155457,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.5277777910232544,
+      "rewards/repetition_penalty_reward": -0.06037369184195995,
       "rewards/tag_count_reward": 1.0,
       "step": 142
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 608.5416870117188,
-      "epoch": 0.143,
-      "grad_norm": 2.415561560430353,
-      "kl": 0.052734375,
-      "learning_rate": 9.949403714711526e-07,
-      "loss": 0.0716,
-      "reward": 2.7793259620666504,
-      "reward_std": 0.3527311235666275,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.045327022671699524,
-      "rewards/tag_count_reward": 0.9843750298023224,
+      "completion_length": 367.7083435058594,
+      "epoch": 0.0715,
+      "grad_norm": 2.8965793866155183,
+      "kl": 0.02557373046875,
+      "learning_rate": 7.149999999999999e-07,
+      "loss": 0.003,
+      "reward": 2.2131210565567017,
+      "reward_std": 0.3542313575744629,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.541666716337204,
+      "rewards/repetition_penalty_reward": -0.05771246552467346,
+      "rewards/tag_count_reward": 1.0,
       "step": 143
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.875,
-      "epoch": 0.144,
-      "grad_norm": 2.9100333446880744,
-      "kl": 0.063232421875,
-      "learning_rate": 9.947027716509488e-07,
-      "loss": 0.0006,
-      "reward": 2.563896417617798,
-      "reward_std": 0.4373708665370941,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.043742526322603226,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 401.12501525878906,
+      "epoch": 0.072,
+      "grad_norm": 3.04746917441956,
+      "kl": 0.02337646484375,
+      "learning_rate": 7.2e-07,
+      "loss": 0.0068,
+      "reward": 2.1876049041748047,
+      "reward_std": 0.4434930086135864,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.43055559694767,
+      "rewards/repetition_penalty_reward": -0.0762840211391449,
+      "rewards/tag_count_reward": 1.0,
       "step": 144
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 566.2916870117188,
-      "epoch": 0.145,
-      "grad_norm": 2.4477067800919987,
-      "kl": 0.0556640625,
-      "learning_rate": 9.944597532678119e-07,
-      "loss": -0.0456,
-      "reward": 2.466762065887451,
-      "reward_std": 0.48124293982982635,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.040182387456297874,
+      "completion_length": 423.3333435058594,
+      "epoch": 0.0725,
+      "grad_norm": 2.7345052559782155,
+      "kl": 0.02398681640625,
+      "learning_rate": 7.249999999999999e-07,
+      "loss": -0.0068,
+      "reward": 2.092087507247925,
+      "reward_std": 0.38775935769081116,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.548611119389534,
+      "rewards/repetition_penalty_reward": -0.06069031357765198,
       "rewards/tag_count_reward": 1.0,
       "step": 145
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.2708740234375,
-      "epoch": 0.146,
-      "grad_norm": 2.735483077486635,
-      "kl": 0.06005859375,
-      "learning_rate": 9.942113192828444e-07,
-      "loss": 0.0496,
-      "reward": 2.605388879776001,
-      "reward_std": 0.47325506806373596,
-      "rewards/accuracy_reward": 0.6458333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.0335000604391098,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 423.06251525878906,
+      "epoch": 0.073,
+      "grad_norm": 2.9633251544566503,
+      "kl": 0.02386474609375,
+      "learning_rate": 7.3e-07,
+      "loss": 0.0508,
+      "reward": 1.913641095161438,
+      "reward_std": 0.47340986132621765,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.5138888955116272,
+      "rewards/repetition_penalty_reward": -0.07420631498098373,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 146
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 508.66668701171875,
-      "epoch": 0.147,
-      "grad_norm": 2.5279842832232133,
-      "kl": 0.0618896484375,
-      "learning_rate": 9.939574727231362e-07,
-      "loss": -0.0023,
-      "reward": 2.484343409538269,
-      "reward_std": 0.5071098208427429,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.04343440383672714,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 466.04168701171875,
+      "epoch": 0.0735,
+      "grad_norm": 2.73697900758412,
+      "kl": 0.0244140625,
+      "learning_rate": 7.35e-07,
+      "loss": -0.0118,
+      "reward": 2.186763048171997,
+      "reward_std": 0.4524783492088318,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.5625000596046448,
+      "rewards/repetition_penalty_reward": -0.09969542920589447,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 147
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 525.5000305175781,
-      "epoch": 0.148,
-      "grad_norm": 2.2840152072866364,
-      "kl": 0.057373046875,
-      "learning_rate": 9.93698216681727e-07,
-      "loss": 0.0172,
-      "reward": 2.810342311859131,
-      "reward_std": 0.24958577752113342,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.029935498721897602,
+      "completion_length": 450.1041717529297,
+      "epoch": 0.074,
+      "grad_norm": 2.7745954642868256,
+      "kl": 0.02392578125,
+      "learning_rate": 7.4e-07,
+      "loss": 0.0112,
+      "reward": 1.895469844341278,
+      "reward_std": 0.42417220771312714,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.5069444626569748,
+      "rewards/repetition_penalty_reward": -0.09064128994941711,
       "rewards/tag_count_reward": 1.0,
       "step": 148
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 587.7083435058594,
-      "epoch": 0.149,
-      "grad_norm": 2.495321625160956,
-      "kl": 0.055908203125,
-      "learning_rate": 9.934335543175705e-07,
-      "loss": 0.0162,
-      "reward": 2.811280369758606,
-      "reward_std": 0.40344125032424927,
-      "rewards/accuracy_reward": 0.875,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.03941434063017368,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 445.29168701171875,
+      "epoch": 0.0745,
+      "grad_norm": 2.6014655366522357,
+      "kl": 0.02288818359375,
+      "learning_rate": 7.45e-07,
+      "loss": 0.015,
+      "reward": 2.204192638397217,
+      "reward_std": 0.38898123800754547,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.451388880610466,
+      "rewards/repetition_penalty_reward": -0.08052962273359299,
+      "rewards/tag_count_reward": 1.0,
       "step": 149
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 634.9166870117188,
-      "epoch": 0.15,
-      "grad_norm": 2.2429641351216834,
-      "kl": 0.0501708984375,
-      "learning_rate": 9.931634888554935e-07,
-      "loss": 0.0582,
-      "reward": 2.6869924068450928,
-      "reward_std": 0.39581531286239624,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.056063249707221985,
+      "completion_length": 455.0208435058594,
+      "epoch": 0.075,
+      "grad_norm": 2.5737888201247574,
+      "kl": 0.02252197265625,
+      "learning_rate": 7.5e-07,
+      "loss": -0.0142,
+      "reward": 2.2747018337249756,
+      "reward_std": 0.37661734223365784,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.5277777910232544,
+      "rewards/repetition_penalty_reward": -0.08640927076339722,
       "rewards/tag_count_reward": 1.0,
       "step": 150
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 500.54168701171875,
-      "epoch": 0.151,
-      "grad_norm": 2.468490484200265,
-      "kl": 0.0621337890625,
-      "learning_rate": 9.928880235861588e-07,
-      "loss": 0.1059,
-      "reward": 2.7397245168685913,
-      "reward_std": 0.32457810640335083,
-      "rewards/accuracy_reward": 0.8125,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.03805328160524368,
+      "completion_length": 380.35418701171875,
+      "epoch": 0.0755,
+      "grad_norm": 3.109877896725919,
+      "kl": 0.030029296875,
+      "learning_rate": 7.55e-07,
+      "loss": 0.0968,
+      "reward": 2.1878127455711365,
+      "reward_std": 0.4361240267753601,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.5208334028720856,
+      "rewards/repetition_penalty_reward": -0.0830206349492073,
       "rewards/tag_count_reward": 1.0,
       "step": 151
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 566.6875,
-      "epoch": 0.152,
-      "grad_norm": 2.5723071707401126,
-      "kl": 0.05517578125,
-      "learning_rate": 9.926071618660237e-07,
-      "loss": 0.0205,
-      "reward": 2.7304086685180664,
-      "reward_std": 0.34727388620376587,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.04042468871921301,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 405.3541717529297,
+      "epoch": 0.076,
+      "grad_norm": 2.883829245260866,
+      "kl": 0.02972412109375,
+      "learning_rate": 7.599999999999999e-07,
+      "loss": 0.0348,
+      "reward": 2.174973964691162,
+      "reward_std": 0.47633519768714905,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.5416667461395264,
+      "rewards/repetition_penalty_reward": -0.06460951268672943,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 152
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 591.1875,
-      "epoch": 0.153,
-      "grad_norm": 2.2620129683268595,
-      "kl": 0.058349609375,
-      "learning_rate": 9.923209071172994e-07,
-      "loss": 0.0377,
-      "reward": 2.5389167070388794,
-      "reward_std": 0.3387536555528641,
-      "rewards/accuracy_reward": 0.6041666716337204,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.044416630640625954,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 400.0833435058594,
+      "epoch": 0.0765,
+      "grad_norm": 11.818146755937764,
+      "kl": 0.3184814453125,
+      "learning_rate": 7.65e-07,
+      "loss": -0.0353,
+      "reward": 1.938881516456604,
+      "reward_std": 0.4211771637201309,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.4027778059244156,
+      "rewards/repetition_penalty_reward": -0.11493790149688721,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 153
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 565.8958435058594,
-      "epoch": 0.154,
-      "grad_norm": 2.7014878630289547,
-      "kl": 0.064453125,
-      "learning_rate": 9.9202926282791e-07,
-      "loss": 0.0875,
-      "reward": 2.7602481842041016,
-      "reward_std": 0.4075208753347397,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.04357131011784077,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 350.4791793823242,
+      "epoch": 0.077,
+      "grad_norm": 9.783552011785373,
+      "kl": 0.1412353515625,
+      "learning_rate": 7.699999999999999e-07,
+      "loss": 0.0569,
+      "reward": 1.8291372656822205,
+      "reward_std": 0.3712882995605469,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.4722222685813904,
+      "rewards/repetition_penalty_reward": -0.08579343557357788,
+      "rewards/tag_count_reward": 0.8593750298023224,
       "step": 154
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.3541870117188,
-      "epoch": 0.155,
-      "grad_norm": 2.6542156680058486,
-      "kl": 0.060791015625,
-      "learning_rate": 9.917322325514487e-07,
-      "loss": 0.0558,
-      "reward": 2.7091113328933716,
-      "reward_std": 0.3156071752309799,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03568043000996113,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 389.3333435058594,
+      "epoch": 0.0775,
+      "grad_norm": 2.8193374221463054,
+      "kl": 0.027099609375,
+      "learning_rate": 7.75e-07,
+      "loss": 0.0898,
+      "reward": 2.1005048751831055,
+      "reward_std": 0.42864808440208435,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.4930555820465088,
+      "rewards/repetition_penalty_reward": -0.0696341022849083,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 155
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 547.8333435058594,
-      "epoch": 0.156,
-      "grad_norm": 2.6130648886630397,
-      "kl": 0.0601806640625,
-      "learning_rate": 9.91429819907136e-07,
-      "loss": -0.002,
-      "reward": 2.6609745025634766,
-      "reward_std": 0.2819754481315613,
-      "rewards/accuracy_reward": 0.708333358168602,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.04041440971195698,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 389.79168701171875,
+      "epoch": 0.078,
+      "grad_norm": 3.00650124802611,
+      "kl": 0.0260009765625,
+      "learning_rate": 7.799999999999999e-07,
+      "loss": 0.0294,
+      "reward": 2.472437024116516,
+      "reward_std": 0.3745059221982956,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.6319445073604584,
+      "rewards/repetition_penalty_reward": -0.0501323863863945,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 156
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 535.5208435058594,
-      "epoch": 0.157,
-      "grad_norm": 2.4079001126290325,
-      "kl": 0.064208984375,
-      "learning_rate": 9.911220285797748e-07,
-      "loss": 0.0085,
-      "reward": 2.513691782951355,
-      "reward_std": 0.3074871450662613,
-      "rewards/accuracy_reward": 0.5625000149011612,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04880848899483681,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 443.6458435058594,
+      "epoch": 0.0785,
+      "grad_norm": 2.698570132718695,
+      "kl": 0.02716064453125,
+      "learning_rate": 7.85e-07,
+      "loss": -0.0863,
+      "reward": 2.4511717557907104,
+      "reward_std": 0.2767828330397606,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.5763888955116272,
+      "rewards/repetition_penalty_reward": -0.08875882998108864,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 157
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 539.7500305175781,
-      "epoch": 0.158,
-      "grad_norm": 2.772614283171138,
-      "kl": 0.0655517578125,
-      "learning_rate": 9.908088623197048e-07,
-      "loss": -0.0064,
-      "reward": 2.664715528488159,
-      "reward_std": 0.265642948448658,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.029728862456977367,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 453.8333435058594,
+      "epoch": 0.079,
+      "grad_norm": 2.9560567179248194,
+      "kl": 0.0284423828125,
+      "learning_rate": 7.9e-07,
+      "loss": 0.0109,
+      "reward": 2.1405357122421265,
+      "reward_std": 0.46409404277801514,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.4444444626569748,
+      "rewards/repetition_penalty_reward": -0.06432541459798813,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 158
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 541.2291870117188,
-      "epoch": 0.159,
-      "grad_norm": 2.4163026682291426,
-      "kl": 0.0654296875,
-      "learning_rate": 9.904903249427582e-07,
-      "loss": -0.0569,
-      "reward": 2.6886651515960693,
-      "reward_std": 0.3855375796556473,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.03355725575238466,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 391.3958435058594,
+      "epoch": 0.0795,
+      "grad_norm": 4.542085974407862,
+      "kl": 0.0609130859375,
+      "learning_rate": 7.95e-07,
+      "loss": -0.1684,
+      "reward": 2.112655282020569,
+      "reward_std": 0.5330419987440109,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.5000000298023224,
+      "rewards/repetition_penalty_reward": -0.0800531879067421,
+      "rewards/tag_count_reward": 0.984375,
       "step": 159
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 576.7916870117188,
-      "epoch": 0.16,
-      "grad_norm": 2.4138081235417306,
-      "kl": 0.06298828125,
-      "learning_rate": 9.901664203302124e-07,
-      "loss": 0.0256,
-      "reward": 2.8020124435424805,
-      "reward_std": 0.28147071599960327,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.03305710572749376,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 450.1875,
+      "epoch": 0.08,
+      "grad_norm": 3.8700369269616512,
+      "kl": 0.0513916015625,
+      "learning_rate": 8e-07,
+      "loss": -0.088,
+      "reward": 1.9016221165657043,
+      "reward_std": 0.26658795773983,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.3819444626569748,
+      "rewards/repetition_penalty_reward": -0.13657239079475403,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 160
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 538.8125,
-      "epoch": 0.161,
-      "grad_norm": 2.3732077198637116,
-      "kl": 0.0616455078125,
-      "learning_rate": 9.89837152428743e-07,
-      "loss": -0.0005,
-      "reward": 2.4948108196258545,
-      "reward_std": 0.3342844396829605,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.04338358715176582,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 388.31251525878906,
+      "epoch": 0.0805,
+      "grad_norm": 3.247547106584298,
+      "kl": 0.0306396484375,
+      "learning_rate": 8.05e-07,
+      "loss": -0.0041,
+      "reward": 1.9749692678451538,
+      "reward_std": 0.468127503991127,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.631944477558136,
+      "rewards/repetition_penalty_reward": -0.08926697075366974,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 161
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 593.4166870117188,
-      "epoch": 0.162,
-      "grad_norm": 2.1912046412336528,
-      "kl": 0.06591796875,
-      "learning_rate": 9.895025252503755e-07,
-      "loss": -0.0436,
-      "reward": 2.7147765159606934,
-      "reward_std": 0.40104810893535614,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.05084844306111336,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 434.5416717529297,
+      "epoch": 0.081,
+      "grad_norm": 3.0024455965281724,
+      "kl": 0.0406494140625,
+      "learning_rate": 8.1e-07,
+      "loss": -0.1179,
+      "reward": 2.2072755098342896,
+      "reward_std": 0.5785205662250519,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.5694444626569748,
+      "rewards/repetition_penalty_reward": -0.07571066915988922,
+      "rewards/tag_count_reward": 0.984375,
       "step": 162
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 604.3750305175781,
-      "epoch": 0.163,
-      "grad_norm": 2.276675908100595,
-      "kl": 0.06494140625,
-      "learning_rate": 9.891625428724364e-07,
-      "loss": -0.0477,
-      "reward": 2.567950963973999,
-      "reward_std": 0.3306031674146652,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.036215804517269135,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 257.0833435058594,
+      "epoch": 0.0815,
+      "grad_norm": 5.717018782517202,
+      "kl": 0.0693359375,
+      "learning_rate": 8.149999999999999e-07,
+      "loss": -0.1785,
+      "reward": 1.6311047077178955,
+      "reward_std": 0.5291544497013092,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.3750000223517418,
+      "rewards/repetition_penalty_reward": -0.061603715643286705,
+      "rewards/tag_count_reward": 0.859375,
       "step": 163
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 522.5416870117188,
-      "epoch": 0.164,
-      "grad_norm": 3.181028204076106,
-      "kl": 0.071044921875,
-      "learning_rate": 9.888172094375033e-07,
-      "loss": 0.1641,
-      "reward": 2.7617541551589966,
-      "reward_std": 0.19565748795866966,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02991252951323986,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 402.7708435058594,
+      "epoch": 0.082,
+      "grad_norm": 2.9816895384452073,
+      "kl": 0.03515625,
+      "learning_rate": 8.199999999999999e-07,
+      "loss": -0.0196,
+      "reward": 2.12343430519104,
+      "reward_std": 0.3986600935459137,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.3888889253139496,
+      "rewards/repetition_penalty_reward": -0.08837117999792099,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 164
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 604.3333435058594,
-      "epoch": 0.165,
-      "grad_norm": 2.3335259823634105,
-      "kl": 0.06494140625,
-      "learning_rate": 9.88466529153356e-07,
-      "loss": -0.1023,
-      "reward": 2.599223256111145,
-      "reward_std": 0.4700702428817749,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.055290715768933296,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 459.5,
+      "epoch": 0.0825,
+      "grad_norm": 2.7857777262413346,
+      "kl": 0.0350341796875,
+      "learning_rate": 8.249999999999999e-07,
+      "loss": -0.0107,
+      "reward": 2.1470755338668823,
+      "reward_std": 0.4360540807247162,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.5000000149011612,
+      "rewards/repetition_penalty_reward": -0.08209127560257912,
+      "rewards/tag_count_reward": 1.0,
       "step": 165
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 583.4791870117188,
-      "epoch": 0.166,
-      "grad_norm": 2.4208829670581853,
-      "kl": 0.070556640625,
-      "learning_rate": 9.881105062929221e-07,
-      "loss": 0.0434,
-      "reward": 2.332135558128357,
-      "reward_std": 0.36515676975250244,
-      "rewards/accuracy_reward": 0.3958333544433117,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.039392316713929176,
-      "rewards/tag_count_reward": 0.9895833730697632,
+      "completion_length": 453.5208435058594,
+      "epoch": 0.083,
+      "grad_norm": 2.904120641790527,
+      "kl": 0.03466796875,
+      "learning_rate": 8.299999999999999e-07,
+      "loss": 0.0829,
+      "reward": 2.2425050735473633,
+      "reward_std": 0.48431138694286346,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.6111111640930176,
+      "rewards/repetition_penalty_reward": -0.0769394002854824,
+      "rewards/tag_count_reward": 1.0,
       "step": 166
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 630.8125,
-      "epoch": 0.167,
-      "grad_norm": 2.4360050121887475,
-      "kl": 0.06396484375,
-      "learning_rate": 9.877491451942284e-07,
-      "loss": -0.0026,
-      "reward": 2.669437289237976,
-      "reward_std": 0.23492664098739624,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9791667461395264,
-      "rewards/repetition_penalty_reward": -0.03889598324894905,
+      "completion_length": 420.43751525878906,
+      "epoch": 0.0835,
+      "grad_norm": 2.7716468429377215,
+      "kl": 0.0323486328125,
+      "learning_rate": 8.349999999999999e-07,
+      "loss": -0.0462,
+      "reward": 2.455272912979126,
+      "reward_std": 0.31839539110660553,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.652777761220932,
+      "rewards/repetition_penalty_reward": -0.07250505313277245,
       "rewards/tag_count_reward": 1.0,
       "step": 167
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 653.8750305175781,
-      "epoch": 0.168,
-      "grad_norm": 2.299477713264774,
-      "kl": 0.066162109375,
-      "learning_rate": 9.873824502603459e-07,
-      "loss": 0.0313,
-      "reward": 2.6158487796783447,
-      "reward_std": 0.346173420548439,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.04387335851788521,
+      "completion_length": 398.2291717529297,
+      "epoch": 0.084,
+      "grad_norm": 2.8774923504646184,
+      "kl": 0.032958984375,
+      "learning_rate": 8.399999999999999e-07,
+      "loss": -0.0118,
+      "reward": 2.3479169607162476,
+      "reward_std": 0.284304603934288,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.5069444477558136,
+      "rewards/repetition_penalty_reward": -0.07569434866309166,
       "rewards/tag_count_reward": 1.0,
       "step": 168
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 625.8541870117188,
-      "epoch": 0.169,
-      "grad_norm": 2.4011318956925503,
-      "kl": 0.0640869140625,
-      "learning_rate": 9.870104259593362e-07,
-      "loss": 0.0797,
-      "reward": 2.6212562322616577,
-      "reward_std": 0.1528831347823143,
-      "rewards/accuracy_reward": 0.6666666716337204,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.034993914887309074,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 376.43751525878906,
+      "epoch": 0.0845,
+      "grad_norm": 3.0955160270611013,
+      "kl": 0.0306396484375,
+      "learning_rate": 8.45e-07,
+      "loss": 0.0031,
+      "reward": 2.163801431655884,
+      "reward_std": 0.33351075649261475,
+      "rewards/accuracy_reward": 0.5000000111758709,
+      "rewards/reasoning_steps_reward": 0.7083333432674408,
+      "rewards/repetition_penalty_reward": -0.04453178122639656,
+      "rewards/tag_count_reward": 1.0,
       "step": 169
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 579.7083435058594,
-      "epoch": 0.17,
-      "grad_norm": 2.5134518850907948,
-      "kl": 0.069580078125,
-      "learning_rate": 9.866330768241983e-07,
-      "loss": 0.0129,
-      "reward": 2.715871572494507,
-      "reward_std": 0.325920432806015,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.03760070167481899,
-      "rewards/tag_count_reward": 0.9895833730697632,
+      "completion_length": 439.2708435058594,
+      "epoch": 0.085,
+      "grad_norm": 2.8041778591178126,
+      "kl": 0.0338134765625,
+      "learning_rate": 8.499999999999999e-07,
+      "loss": 0.0152,
+      "reward": 2.1758224964141846,
+      "reward_std": 0.48637865483760834,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.6180555820465088,
+      "rewards/repetition_penalty_reward": -0.08285807259380817,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 170
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 633.0625305175781,
-      "epoch": 0.171,
-      "grad_norm": 2.4224007332086424,
-      "kl": 0.0693359375,
-      "learning_rate": 9.862504074528126e-07,
-      "loss": 0.0557,
-      "reward": 2.5395129919052124,
-      "reward_std": 0.4677208960056305,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04382044076919556,
+      "completion_length": 368.125,
+      "epoch": 0.0855,
+      "grad_norm": 3.047194462178732,
+      "kl": 0.0386962890625,
+      "learning_rate": 8.55e-07,
+      "loss": -0.1328,
+      "reward": 2.137127637863159,
+      "reward_std": 0.3549790009856224,
+      "rewards/accuracy_reward": 0.5833333358168602,
+      "rewards/reasoning_steps_reward": 0.6250000447034836,
+      "rewards/repetition_penalty_reward": -0.07120569795370102,
       "rewards/tag_count_reward": 1.0,
       "step": 171
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 621.0416870117188,
-      "epoch": 0.172,
-      "grad_norm": 2.1277905450282595,
-      "kl": 0.0667724609375,
-      "learning_rate": 9.85862422507884e-07,
-      "loss": 0.0193,
-      "reward": 2.8295018672943115,
-      "reward_std": 0.13751935493201017,
-      "rewards/accuracy_reward": 0.875,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04549824073910713,
+      "completion_length": 424.5833435058594,
+      "epoch": 0.086,
+      "grad_norm": 2.8616005778229785,
+      "kl": 0.03955078125,
+      "learning_rate": 8.599999999999999e-07,
+      "loss": -0.0308,
+      "reward": 2.2606923580169678,
+      "reward_std": 0.5883138179779053,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.659722238779068,
+      "rewards/repetition_penalty_reward": -0.08653007447719574,
       "rewards/tag_count_reward": 1.0,
       "step": 172
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 648.7916870117188,
-      "epoch": 0.173,
-      "grad_norm": 2.324532047104547,
-      "kl": 0.069091796875,
-      "learning_rate": 9.854691267168871e-07,
-      "loss": 0.1049,
-      "reward": 2.722783327102661,
-      "reward_std": 0.41735316812992096,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.0376333836466074,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 448.9166717529297,
+      "epoch": 0.0865,
+      "grad_norm": 2.649883532273222,
+      "kl": 0.0318603515625,
+      "learning_rate": 8.65e-07,
+      "loss": -0.0174,
+      "reward": 2.2526204586029053,
+      "reward_std": 0.45084331929683685,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.5625000596046448,
+      "rewards/repetition_penalty_reward": -0.08071288466453552,
+      "rewards/tag_count_reward": 1.0,
       "step": 173
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 671.7083435058594,
-      "epoch": 0.174,
-      "grad_norm": 1.9577431816288606,
-      "kl": 0.066650390625,
-      "learning_rate": 9.850705248720068e-07,
-      "loss": 0.0282,
-      "reward": 2.7458845376968384,
-      "reward_std": 0.3250259757041931,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04578226990997791,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 366.625,
+      "epoch": 0.087,
+      "grad_norm": 3.025435169465401,
+      "kl": 0.0360107421875,
+      "learning_rate": 8.699999999999999e-07,
+      "loss": 0.0487,
+      "reward": 2.397473096847534,
+      "reward_std": 0.47506286203861237,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.7222222685813904,
+      "rewards/repetition_penalty_reward": -0.04349912703037262,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 174
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 700.7916870117188,
-      "epoch": 0.175,
-      "grad_norm": 1.8987257817841927,
-      "kl": 0.0645751953125,
-      "learning_rate": 9.846666218300807e-07,
-      "loss": 0.0104,
-      "reward": 2.6111772060394287,
-      "reward_std": 0.39906148612499237,
-      "rewards/accuracy_reward": 0.708333358168602,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.06243372894823551,
+      "completion_length": 431.31251525878906,
+      "epoch": 0.0875,
+      "grad_norm": 2.9618305547728543,
+      "kl": 0.0430908203125,
+      "learning_rate": 8.75e-07,
+      "loss": -0.0493,
+      "reward": 2.137214183807373,
+      "reward_std": 0.5185305774211884,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.6458333730697632,
+      "rewards/repetition_penalty_reward": -0.09195243567228317,
       "rewards/tag_count_reward": 1.0,
       "step": 175
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 628.1041870117188,
-      "epoch": 0.176,
-      "grad_norm": 2.404386166811554,
-      "kl": 0.071533203125,
-      "learning_rate": 9.8425742251254e-07,
-      "loss": 0.0773,
-      "reward": 2.6240488290786743,
-      "reward_std": 0.37613917887210846,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.06692333333194256,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 326.56251525878906,
+      "epoch": 0.088,
+      "grad_norm": 2.9727047540243623,
+      "kl": 0.0399169921875,
+      "learning_rate": 8.799999999999999e-07,
+      "loss": 0.0354,
+      "reward": 2.605048894882202,
+      "reward_std": 0.31301962584257126,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.6875,
+      "rewards/repetition_penalty_reward": -0.04078466631472111,
+      "rewards/tag_count_reward": 1.0,
       "step": 176
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 587.6458435058594,
-      "epoch": 0.177,
-      "grad_norm": 2.4043287387592125,
-      "kl": 0.075439453125,
-      "learning_rate": 9.838429319053495e-07,
-      "loss": 0.0096,
-      "reward": 2.5300588607788086,
-      "reward_std": 0.3319057375192642,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.04112182557582855,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 473.66668701171875,
+      "epoch": 0.0885,
+      "grad_norm": 2.522294184113134,
+      "kl": 0.0426025390625,
+      "learning_rate": 8.85e-07,
+      "loss": -0.0891,
+      "reward": 2.436373710632324,
+      "reward_std": 0.4591614753007889,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.8125000298023224,
+      "rewards/repetition_penalty_reward": -0.06362627819180489,
+      "rewards/tag_count_reward": 1.0,
       "step": 177
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 698.8125305175781,
-      "epoch": 0.178,
-      "grad_norm": 2.028048720044857,
-      "kl": 0.072265625,
-      "learning_rate": 9.83423155058946e-07,
-      "loss": -0.0443,
-      "reward": 2.4296261072158813,
-      "reward_std": 0.34993553161621094,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04954059235751629,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 398.0,
+      "epoch": 0.089,
+      "grad_norm": 2.952883589913921,
+      "kl": 0.0462646484375,
+      "learning_rate": 8.9e-07,
+      "loss": 0.007,
+      "reward": 2.2954567670822144,
+      "reward_std": 0.47059088945388794,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.6944444477558136,
+      "rewards/repetition_penalty_reward": -0.06044617295265198,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 178
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 643.4166870117188,
-      "epoch": 0.179,
-      "grad_norm": 2.5526900934607544,
-      "kl": 0.07763671875,
-      "learning_rate": 9.829980970881784e-07,
-      "loss": 0.1619,
-      "reward": 2.847709059715271,
-      "reward_std": 0.24474234879016876,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04291607812047005,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 377.7916717529297,
+      "epoch": 0.0895,
+      "grad_norm": 2.8614701370754245,
+      "kl": 0.04736328125,
+      "learning_rate": 8.95e-07,
+      "loss": -0.0243,
+      "reward": 2.436295747756958,
+      "reward_std": 0.4291905164718628,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.701388955116272,
+      "rewards/repetition_penalty_reward": -0.06717650964856148,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 179
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 617.2291870117188,
-      "epoch": 0.18,
-      "grad_norm": 1.9749460293522159,
-      "kl": 0.082275390625,
-      "learning_rate": 9.825677631722435e-07,
-      "loss": 0.0669,
-      "reward": 2.691727042198181,
-      "reward_std": 0.3086921200156212,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03743956796824932,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 485.9583435058594,
+      "epoch": 0.09,
+      "grad_norm": 3.026265114082148,
+      "kl": 0.0382080078125,
+      "learning_rate": 9e-07,
+      "loss": 0.0719,
+      "reward": 2.1689553260803223,
+      "reward_std": 0.30075494945049286,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.7361111640930176,
+      "rewards/repetition_penalty_reward": -0.06194741278886795,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 180
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 619.3541870117188,
-      "epoch": 0.181,
-      "grad_norm": 2.115751260793685,
-      "kl": 0.0771484375,
-      "learning_rate": 9.821321585546243e-07,
-      "loss": 0.0106,
-      "reward": 2.788450598716736,
-      "reward_std": 0.2660830020904541,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.05529944226145744,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 448.3125,
+      "epoch": 0.0905,
+      "grad_norm": 2.810163992973923,
+      "kl": 0.04541015625,
+      "learning_rate": 9.05e-07,
+      "loss": 0.0161,
+      "reward": 2.3727446794509888,
+      "reward_std": 0.5085368752479553,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.7083333730697632,
+      "rewards/repetition_penalty_reward": -0.07517208904027939,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 181
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 631.7500305175781,
-      "epoch": 0.182,
-      "grad_norm": 2.195356667493195,
-      "kl": 0.084716796875,
-      "learning_rate": 9.816912885430258e-07,
-      "loss": 0.0269,
-      "reward": 2.6126253604888916,
-      "reward_std": 0.32306814193725586,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.054041286930441856,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 569.9583435058594,
+      "epoch": 0.091,
+      "grad_norm": 2.6212790656780056,
+      "kl": 0.0643310546875,
+      "learning_rate": 9.1e-07,
+      "loss": -0.0303,
+      "reward": 2.1286094188690186,
+      "reward_std": 0.5249549150466919,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.79861119389534,
+      "rewards/repetition_penalty_reward": -0.09187676757574081,
+      "rewards/tag_count_reward": 0.984375,
       "step": 182
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 634.8125305175781,
-      "epoch": 0.183,
-      "grad_norm": 2.1090909584715587,
-      "kl": 0.0791015625,
-      "learning_rate": 9.812451585093098e-07,
-      "loss": 0.0252,
-      "reward": 2.7393864393234253,
-      "reward_std": 0.3596974164247513,
-      "rewards/accuracy_reward": 0.8125,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.052280182018876076,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 571.0208435058594,
+      "epoch": 0.0915,
+      "grad_norm": 2.7275256259539025,
+      "kl": 0.045166015625,
+      "learning_rate": 9.15e-07,
+      "loss": 0.0061,
+      "reward": 2.457062244415283,
+      "reward_std": 0.44156837463378906,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.8611111640930176,
+      "rewards/repetition_penalty_reward": -0.09154891595244408,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 183
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 661.7083435058594,
-      "epoch": 0.184,
-      "grad_norm": 2.077351011146557,
-      "kl": 0.0771484375,
-      "learning_rate": 9.807937738894303e-07,
-      "loss": 0.0566,
-      "reward": 2.869002103805542,
-      "reward_std": 0.24853621423244476,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04766450449824333,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 568.0416870117188,
+      "epoch": 0.092,
+      "grad_norm": 2.5794652768185697,
+      "kl": 0.0457763671875,
+      "learning_rate": 9.2e-07,
+      "loss": -0.0542,
+      "reward": 2.1548627614974976,
+      "reward_std": 0.41270676255226135,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.784722238779068,
+      "rewards/repetition_penalty_reward": -0.10381785407662392,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 184
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 603.0208435058594,
-      "epoch": 0.185,
-      "grad_norm": 2.1637455121074223,
-      "kl": 0.087158203125,
-      "learning_rate": 9.80337140183366e-07,
-      "loss": 0.0752,
-      "reward": 2.8117637634277344,
-      "reward_std": 0.24315628595650196,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.035458519123494625,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 550.5833435058594,
+      "epoch": 0.0925,
+      "grad_norm": 2.6922728125371544,
+      "kl": 0.0513916015625,
+      "learning_rate": 9.25e-07,
+      "loss": -0.0506,
+      "reward": 2.1982321739196777,
+      "reward_std": 0.47882315516471863,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.909722238779068,
+      "rewards/repetition_penalty_reward": -0.09690695255994797,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 185
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 583.1458435058594,
-      "epoch": 0.186,
-      "grad_norm": 2.275135225584201,
-      "kl": 0.09326171875,
-      "learning_rate": 9.798752629550546e-07,
-      "loss": 0.0655,
-      "reward": 2.8143444061279297,
-      "reward_std": 0.21083202213048935,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.039822228252887726,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 552.75,
+      "epoch": 0.093,
+      "grad_norm": 2.34743025577209,
+      "kl": 0.0548095703125,
+      "learning_rate": 9.3e-07,
+      "loss": -0.017,
+      "reward": 2.2561983466148376,
+      "reward_std": 0.37413595616817474,
+      "rewards/accuracy_reward": 0.4583333358168602,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.0944962427020073,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 186
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 583.2916870117188,
-      "epoch": 0.187,
-      "grad_norm": 2.233756675170646,
-      "kl": 0.089599609375,
-      "learning_rate": 9.794081478323245e-07,
-      "loss": 0.0115,
-      "reward": 2.55490505695343,
-      "reward_std": 0.20559479296207428,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.04231713339686394,
+      "completion_length": 445.5208435058594,
+      "epoch": 0.0935,
+      "grad_norm": 3.017831161763003,
+      "kl": 0.061767578125,
+      "learning_rate": 9.35e-07,
+      "loss": 0.0359,
+      "reward": 2.617498517036438,
+      "reward_std": 0.2746337354183197,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.86111119389534,
+      "rewards/repetition_penalty_reward": -0.07694609090685844,
       "rewards/tag_count_reward": 1.0,
       "step": 187
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 577.3958435058594,
-      "epoch": 0.188,
-      "grad_norm": 2.312702195384097,
-      "kl": 0.09033203125,
-      "learning_rate": 9.78935800506826e-07,
-      "loss": 0.0262,
-      "reward": 2.8108856678009033,
-      "reward_std": 0.29880291223526,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04328102432191372,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 516.4166870117188,
+      "epoch": 0.094,
+      "grad_norm": 10.120993359942497,
+      "kl": 0.0875244140625,
+      "learning_rate": 9.399999999999999e-07,
+      "loss": -0.0016,
+      "reward": 2.3637200593948364,
+      "reward_std": 0.35263994336128235,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.92361119389534,
+      "rewards/repetition_penalty_reward": -0.09114107862114906,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 188
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 621.6458435058594,
-      "epoch": 0.189,
-      "grad_norm": 2.450467067468715,
-      "kl": 0.095458984375,
-      "learning_rate": 9.784582267339622e-07,
-      "loss": 0.1076,
-      "reward": 2.691537618637085,
-      "reward_std": 0.46115170419216156,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03762921877205372,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 578.9166717529297,
+      "epoch": 0.0945,
+      "grad_norm": 2.4810269448429314,
+      "kl": 0.060546875,
+      "learning_rate": 9.45e-07,
+      "loss": 0.135,
+      "reward": 2.5320863723754883,
+      "reward_std": 0.4612206518650055,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.12763602659106255,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 189
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 552.2291870117188,
-      "epoch": 0.19,
-      "grad_norm": 2.33663470812737,
-      "kl": 0.096923828125,
-      "learning_rate": 9.779754323328192e-07,
-      "loss": 0.0829,
-      "reward": 2.556964159011841,
-      "reward_std": 0.2390810027718544,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02636928204447031,
+      "completion_length": 536.8333435058594,
+      "epoch": 0.095,
+      "grad_norm": 2.42276363492381,
+      "kl": 0.0567626953125,
+      "learning_rate": 9.499999999999999e-07,
+      "loss": -0.0277,
+      "reward": 2.275315523147583,
+      "reward_std": 0.42472073435783386,
+      "rewards/accuracy_reward": 0.4375000149011612,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.09274015948176384,
       "rewards/tag_count_reward": 1.0,
       "step": 190
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.0416870117188,
-      "epoch": 0.191,
-      "grad_norm": 2.145945974879737,
-      "kl": 0.094482421875,
-      "learning_rate": 9.774874231860935e-07,
-      "loss": -0.0278,
-      "reward": 2.621474266052246,
-      "reward_std": 0.42711225152015686,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04519243072718382,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 515.4166870117188,
+      "epoch": 0.0955,
+      "grad_norm": 3.0986857990959504,
+      "kl": 0.102783203125,
+      "learning_rate": 9.55e-07,
+      "loss": -0.069,
+      "reward": 2.445477247238159,
+      "reward_std": 0.40403106808662415,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.10487010702490807,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 191
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 539.6875305175781,
-      "epoch": 0.192,
-      "grad_norm": 2.147974112089912,
-      "kl": 0.090087890625,
-      "learning_rate": 9.769942052400235e-07,
-      "loss": -0.0237,
-      "reward": 2.76838481426239,
-      "reward_std": 0.310782328248024,
-      "rewards/accuracy_reward": 0.8125,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04411514103412628,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 576.6041717529297,
+      "epoch": 0.096,
+      "grad_norm": 2.527746035134317,
+      "kl": 0.0672607421875,
+      "learning_rate": 9.6e-07,
+      "loss": 0.0669,
+      "reward": 2.4357703924179077,
+      "reward_std": 0.40775516629219055,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.10763245820999146,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 192
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 525.2916870117188,
-      "epoch": 0.193,
-      "grad_norm": 2.1483746204852827,
-      "kl": 0.096923828125,
-      "learning_rate": 9.764957845043135e-07,
-      "loss": 0.0031,
-      "reward": 2.842332124710083,
-      "reward_std": 0.13549592159688473,
-      "rewards/accuracy_reward": 0.875,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.032667966559529305,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 561.1250152587891,
+      "epoch": 0.0965,
+      "grad_norm": 2.0944240845048028,
+      "kl": 0.0623779296875,
+      "learning_rate": 9.649999999999999e-07,
+      "loss": -0.012,
+      "reward": 2.684821605682373,
+      "reward_std": 0.2746804505586624,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.11726166307926178,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 193
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 595.1875305175781,
-      "epoch": 0.194,
-      "grad_norm": 1.981195616724217,
-      "kl": 0.09228515625,
-      "learning_rate": 9.759921670520634e-07,
-      "loss": -0.0784,
-      "reward": 2.706938862800598,
-      "reward_std": 0.286833293735981,
+      "completion_length": 545.1875305175781,
+      "epoch": 0.097,
+      "grad_norm": 2.1965210412413896,
+      "kl": 0.059814453125,
+      "learning_rate": 9.7e-07,
+      "loss": -0.0042,
+      "reward": 2.5165964365005493,
+      "reward_std": 0.41372165083885193,
       "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.03611667454242706,
-      "rewards/tag_count_reward": 1.0,
+      "rewards/reasoning_steps_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.09277866780757904,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 194
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 576.2708587646484,
-      "epoch": 0.195,
-      "grad_norm": 2.648586034553415,
-      "kl": 0.09375,
-      "learning_rate": 9.754833590196926e-07,
-      "loss": 0.198,
-      "reward": 2.8884334564208984,
-      "reward_std": 0.24114538729190826,
-      "rewards/accuracy_reward": 0.9375000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.035177784971892834,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 519.5416870117188,
+      "epoch": 0.0975,
+      "grad_norm": 2.4964745244416346,
+      "kl": 0.052978515625,
+      "learning_rate": 9.75e-07,
+      "loss": 0.065,
+      "reward": 2.522639751434326,
+      "reward_std": 0.43553659319877625,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.11277706176042557,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 195
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 593.3750305175781,
-      "epoch": 0.196,
-      "grad_norm": 2.403156236141252,
-      "kl": 0.10107421875,
-      "learning_rate": 9.749693666068663e-07,
-      "loss": -0.0295,
-      "reward": 2.310402512550354,
-      "reward_std": 0.2717669606208801,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.03681975603103638,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 564.7708435058594,
+      "epoch": 0.098,
+      "grad_norm": 2.2610161792561527,
+      "kl": 0.061767578125,
+      "learning_rate": 9.8e-07,
+      "loss": -0.0127,
+      "reward": 2.3144911527633667,
+      "reward_std": 0.4604138135910034,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.10912004858255386,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 196
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 466.60418701171875,
-      "epoch": 0.197,
-      "grad_norm": 5.530819309602341,
-      "kl": 0.12060546875,
-      "learning_rate": 9.744501960764203e-07,
-      "loss": 0.0515,
-      "reward": 2.7814574241638184,
-      "reward_std": 0.3644670993089676,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02062598057091236,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 471.6666717529297,
+      "epoch": 0.0985,
+      "grad_norm": 2.720385116064672,
+      "kl": 0.061767578125,
+      "learning_rate": 9.849999999999999e-07,
+      "loss": 0.0132,
+      "reward": 2.4773154258728027,
+      "reward_std": 0.4094041436910629,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.06955961138010025,
+      "rewards/tag_count_reward": 0.984375,
       "step": 197
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 528.1666870117188,
-      "epoch": 0.198,
-      "grad_norm": 2.484071909086178,
-      "kl": 0.1015625,
-      "learning_rate": 9.739258537542835e-07,
-      "loss": 0.0689,
-      "reward": 2.845056176185608,
-      "reward_std": 0.29101284593343735,
+      "completion_length": 537.2500305175781,
+      "epoch": 0.099,
+      "grad_norm": 2.7986049436758194,
+      "kl": 0.06494140625,
+      "learning_rate": 9.9e-07,
+      "loss": 0.0464,
+      "reward": 2.6958858966827393,
+      "reward_std": 0.3497622162103653,
       "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.0229995958507061,
-      "rewards/tag_count_reward": 1.0,
+      "rewards/reasoning_steps_reward": 0.958333432674408,
+      "rewards/repetition_penalty_reward": -0.11140592023730278,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 198
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 472.9583435058594,
-      "epoch": 0.199,
-      "grad_norm": 2.111890817265467,
-      "kl": 0.10791015625,
-      "learning_rate": 9.733963460294015e-07,
-      "loss": -0.0139,
-      "reward": 2.8303216695785522,
-      "reward_std": 0.34869830310344696,
-      "rewards/accuracy_reward": 0.8958333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.034261688590049744,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 525.125,
+      "epoch": 0.0995,
+      "grad_norm": 2.5376762252956166,
+      "kl": 0.057861328125,
+      "learning_rate": 9.95e-07,
+      "loss": 0.0873,
+      "reward": 2.4644105434417725,
+      "reward_std": 0.5261338353157043,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.888888955116272,
+      "rewards/repetition_penalty_reward": -0.09635350480675697,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 199
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 515.3333587646484,
-      "epoch": 0.2,
-      "grad_norm": 2.557342178020337,
-      "kl": 0.1005859375,
-      "learning_rate": 9.728616793536587e-07,
-      "loss": -0.0301,
-      "reward": 2.5907087326049805,
-      "reward_std": 0.3476664125919342,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.027346878312528133,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 479.56251525878906,
+      "epoch": 0.1,
+      "grad_norm": 2.51025292795749,
+      "kl": 0.0599365234375,
+      "learning_rate": 1e-06,
+      "loss": -0.0162,
+      "reward": 2.544384002685547,
+      "reward_std": 0.447608157992363,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9166667461395264,
+      "rewards/repetition_penalty_reward": -0.10144944489002228,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 200
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 531.8333435058594,
-      "epoch": 0.201,
-      "grad_norm": 2.699850984290719,
-      "kl": 0.107177734375,
-      "learning_rate": 9.723218602418e-07,
-      "loss": 0.026,
-      "reward": 2.548583984375,
-      "reward_std": 0.4529786705970764,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.027805176563560963,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 569.0625,
+      "epoch": 0.1005,
+      "grad_norm": 2.1126300771805564,
+      "kl": 0.0582275390625,
+      "learning_rate": 9.999993146109795e-07,
+      "loss": 0.0202,
+      "reward": 2.5578041076660156,
+      "reward_std": 0.334363654255867,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.92361119389534,
+      "rewards/repetition_penalty_reward": -0.10539035871624947,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 201
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 506.7708435058594,
-      "epoch": 0.202,
-      "grad_norm": 2.3216160000213297,
-      "kl": 0.088623046875,
-      "learning_rate": 9.717768952713511e-07,
-      "loss": -0.0301,
-      "reward": 2.6107916831970215,
-      "reward_std": 0.3744069039821625,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.03851393796503544,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 487.8333435058594,
+      "epoch": 0.101,
+      "grad_norm": 2.537230361987624,
+      "kl": 0.068603515625,
+      "learning_rate": 9.999972584460056e-07,
+      "loss": -0.0015,
+      "reward": 2.552291989326477,
+      "reward_std": 0.5273123234510422,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.8750000596046448,
+      "rewards/repetition_penalty_reward": -0.09874986857175827,
+      "rewards/tag_count_reward": 0.984375,
       "step": 202
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 530.9375152587891,
-      "epoch": 0.203,
-      "grad_norm": 3.0552289765907665,
-      "kl": 0.105712890625,
-      "learning_rate": 9.71226791082538e-07,
-      "loss": 0.0276,
-      "reward": 2.899757981300354,
-      "reward_std": 0.19229509681463242,
-      "rewards/accuracy_reward": 0.9583333432674408,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.032533735036849976,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 460.3333435058594,
+      "epoch": 0.1015,
+      "grad_norm": 2.663772208431935,
+      "kl": 0.0540771484375,
+      "learning_rate": 9.99993831511342e-07,
+      "loss": 0.0486,
+      "reward": 2.662340521812439,
+      "reward_std": 0.3466527909040451,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.8263889253139496,
+      "rewards/repetition_penalty_reward": -0.08071524277329445,
+      "rewards/tag_count_reward": 1.0,
       "step": 203
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 613.8333435058594,
-      "epoch": 0.204,
-      "grad_norm": 2.25614862930677,
-      "kl": 0.094482421875,
-      "learning_rate": 9.706715543782064e-07,
-      "loss": -0.0024,
-      "reward": 2.8068939447402954,
-      "reward_std": 0.3218880593776703,
-      "rewards/accuracy_reward": 0.8958333730697632,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.0438005393370986,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 551.3750305175781,
+      "epoch": 0.102,
+      "grad_norm": 2.3037155314735895,
+      "kl": 0.0634765625,
+      "learning_rate": 9.999890338174275e-07,
+      "loss": 0.116,
+      "reward": 2.5760300159454346,
+      "reward_std": 0.3985503613948822,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.07674792408943176,
+      "rewards/tag_count_reward": 1.0,
       "step": 204
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 548.3333740234375,
-      "epoch": 0.205,
-      "grad_norm": 2.1739212203171405,
-      "kl": 0.100830078125,
-      "learning_rate": 9.701111919237408e-07,
-      "loss": -0.0353,
-      "reward": 2.628522038459778,
-      "reward_std": 0.37829773128032684,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.03120046854019165,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 460.6458435058594,
+      "epoch": 0.1025,
+      "grad_norm": 2.659875652962376,
+      "kl": 0.064453125,
+      "learning_rate": 9.99982865378877e-07,
+      "loss": -0.0381,
+      "reward": 2.600967049598694,
+      "reward_std": 0.34831008315086365,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.8680555820465088,
+      "rewards/repetition_penalty_reward": -0.0691719576716423,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 205
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 575.6666870117188,
-      "epoch": 0.206,
-      "grad_norm": 2.142090873135409,
-      "kl": 0.081298828125,
-      "learning_rate": 9.695457105469804e-07,
-      "loss": 0.0283,
-      "reward": 2.6884310245513916,
-      "reward_std": 0.1302720569074154,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.042471904307603836,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 473.8333435058594,
+      "epoch": 0.103,
+      "grad_norm": 2.672719916018964,
+      "kl": 0.0626220703125,
+      "learning_rate": 9.999753262144804e-07,
+      "loss": 0.1019,
+      "reward": 2.5358208417892456,
+      "reward_std": 0.42981427907943726,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.888888955116272,
+      "rewards/repetition_penalty_reward": -0.061401575803756714,
+      "rewards/tag_count_reward": 1.0,
       "step": 206
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 537.5833587646484,
-      "epoch": 0.207,
-      "grad_norm": 2.5678922668456563,
-      "kl": 0.092529296875,
-      "learning_rate": 9.689751171381377e-07,
-      "loss": 0.0451,
-      "reward": 2.6688740253448486,
-      "reward_std": 0.40736155211925507,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.025570510886609554,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 469.6458435058594,
+      "epoch": 0.1035,
+      "grad_norm": 2.5741947297183385,
+      "kl": 0.0634765625,
+      "learning_rate": 9.999664163472034e-07,
+      "loss": -0.0236,
+      "reward": 2.332731008529663,
+      "reward_std": 0.4647013247013092,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.8611111342906952,
+      "rewards/repetition_penalty_reward": -0.08567183464765549,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 207
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 585.875,
-      "epoch": 0.208,
-      "grad_norm": 2.5924187225102355,
-      "kl": 0.098388671875,
-      "learning_rate": 9.683994186497132e-07,
-      "loss": 0.169,
-      "reward": 2.3460363149642944,
-      "reward_std": 0.4289597123861313,
-      "rewards/accuracy_reward": 0.4375000111758709,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.034172071143984795,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 546.7916870117188,
+      "epoch": 0.104,
+      "grad_norm": 2.7033266921870123,
+      "kl": 0.0594482421875,
+      "learning_rate": 9.999561358041868e-07,
+      "loss": -0.0888,
+      "reward": 2.297361373901367,
+      "reward_std": 0.5272791683673859,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.8750001192092896,
+      "rewards/repetition_penalty_reward": -0.07763872668147087,
+      "rewards/tag_count_reward": 1.0,
       "step": 208
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 488.35418701171875,
-      "epoch": 0.209,
-      "grad_norm": 2.6979608618101913,
-      "kl": 0.099609375,
-      "learning_rate": 9.67818622096411e-07,
-      "loss": 0.0553,
-      "reward": 2.830100178718567,
-      "reward_std": 0.3162241727113724,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.017122075892984867,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 443.62501525878906,
+      "epoch": 0.1045,
+      "grad_norm": 3.90101302775015,
+      "kl": 0.072265625,
+      "learning_rate": 9.99944484616747e-07,
+      "loss": 0.0252,
+      "reward": 2.6521321535110474,
+      "reward_std": 0.33744025230407715,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.8541667461395264,
+      "rewards/repetition_penalty_reward": -0.07703462615609169,
+      "rewards/tag_count_reward": 1.0,
       "step": 209
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 535.6041870117188,
-      "epoch": 0.21,
-      "grad_norm": 2.6398456904077494,
-      "kl": 0.09814453125,
-      "learning_rate": 9.672327345550543e-07,
-      "loss": 0.0618,
-      "reward": 2.509260654449463,
-      "reward_std": 0.28831499069929123,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.023725682869553566,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 540.4375,
+      "epoch": 0.105,
+      "grad_norm": 2.31473784045809,
+      "kl": 0.0623779296875,
+      "learning_rate": 9.99931462820376e-07,
+      "loss": 0.0167,
+      "reward": 2.5439374446868896,
+      "reward_std": 0.2827417850494385,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.10884056985378265,
+      "rewards/tag_count_reward": 1.0,
       "step": 210
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 599.6041870117188,
-      "epoch": 0.211,
-      "grad_norm": 4.517806643959562,
-      "kl": 0.109130859375,
-      "learning_rate": 9.666417631644976e-07,
-      "loss": 0.0023,
-      "reward": 2.5767656564712524,
-      "reward_std": 0.3708747327327728,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.03608160652220249,
+      "completion_length": 518.5625,
+      "epoch": 0.1055,
+      "grad_norm": 2.4382433122696763,
+      "kl": 0.074462890625,
+      "learning_rate": 9.999170704547398e-07,
+      "loss": 0.0511,
+      "reward": 2.5632206201553345,
+      "reward_std": 0.31003791093826294,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.0912933386862278,
       "rewards/tag_count_reward": 0.9947916865348816,
       "step": 211
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.4166870117188,
-      "epoch": 0.212,
-      "grad_norm": 2.261259945602269,
-      "kl": 0.10205078125,
-      "learning_rate": 9.66045715125541e-07,
-      "loss": 0.097,
-      "reward": 2.5739282369613647,
-      "reward_std": 0.3682664930820465,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.040655218064785004,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 503.8125305175781,
+      "epoch": 0.106,
+      "grad_norm": 2.370275119306627,
+      "kl": 0.0654296875,
+      "learning_rate": 9.999013075636804e-07,
+      "loss": -0.008,
+      "reward": 2.4429709911346436,
+      "reward_std": 0.4311106353998184,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.09175140410661697,
+      "rewards/tag_count_reward": 1.0,
       "step": 212
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 515.7291870117188,
-      "epoch": 0.213,
-      "grad_norm": 2.397926932727844,
-      "kl": 0.1044921875,
-      "learning_rate": 9.654445977008414e-07,
-      "loss": 0.0394,
-      "reward": 2.682453751564026,
-      "reward_std": 0.4673650562763214,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.024143685586750507,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 588.3958740234375,
+      "epoch": 0.1065,
+      "grad_norm": 2.5342323653428696,
+      "kl": 0.0633544921875,
+      "learning_rate": 9.998841741952141e-07,
+      "loss": 0.1784,
+      "reward": 2.6273250579833984,
+      "reward_std": 0.28149472177028656,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.09489719569683075,
+      "rewards/tag_count_reward": 1.0,
       "step": 213
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 691.2916870117188,
-      "epoch": 0.214,
-      "grad_norm": 8.712060550440611,
-      "kl": 0.124755859375,
-      "learning_rate": 9.648384182148252e-07,
-      "loss": 0.0044,
-      "reward": 2.719285011291504,
-      "reward_std": 0.3266633450984955,
-      "rewards/accuracy_reward": 0.8125,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.04807615838944912,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 525.9375305175781,
+      "epoch": 0.107,
+      "grad_norm": 2.1916917773556763,
+      "kl": 0.06689453125,
+      "learning_rate": 9.998656704015323e-07,
+      "loss": -0.0292,
+      "reward": 2.3184261322021484,
+      "reward_std": 0.42756618559360504,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.09303238615393639,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 214
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 609.5416870117188,
-      "epoch": 0.215,
-      "grad_norm": 2.215014758114719,
-      "kl": 0.09326171875,
-      "learning_rate": 9.64227184053598e-07,
-      "loss": 0.0437,
-      "reward": 2.835343837738037,
-      "reward_std": 0.26147839426994324,
-      "rewards/accuracy_reward": 0.9375000298023224,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.04312858823686838,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 622.7708435058594,
+      "epoch": 0.1075,
+      "grad_norm": 2.279834707710596,
+      "kl": 0.065673828125,
+      "learning_rate": 9.998457962390008e-07,
+      "loss": 0.0573,
+      "reward": 2.3788540363311768,
+      "reward_std": 0.44249793887138367,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.09510444849729538,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 215
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 600.7291870117188,
-      "epoch": 0.216,
-      "grad_norm": 2.0090732793867403,
-      "kl": 0.124267578125,
-      "learning_rate": 9.636109026648554e-07,
-      "loss": -0.091,
-      "reward": 2.805617928504944,
-      "reward_std": 0.33824414014816284,
-      "rewards/accuracy_reward": 0.8958333730697632,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.04160444252192974,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 561.1458435058594,
+      "epoch": 0.108,
+      "grad_norm": 2.4104772332041033,
+      "kl": 0.068603515625,
+      "learning_rate": 9.998245517681593e-07,
+      "loss": 0.0665,
+      "reward": 2.5858339071273804,
+      "reward_std": 0.3917628526687622,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.07562451809644699,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 216
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 645.2083740234375,
-      "epoch": 0.217,
-      "grad_norm": 13.509335743253168,
-      "kl": 0.146484375,
-      "learning_rate": 9.629895815577915e-07,
-      "loss": -0.0366,
-      "reward": 2.7113317251205444,
-      "reward_std": 0.34692464768886566,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.045612769201397896,
+      "completion_length": 494.2291717529297,
+      "epoch": 0.1085,
+      "grad_norm": 2.3453403680398917,
+      "kl": 0.07568359375,
+      "learning_rate": 9.998019370537227e-07,
+      "loss": -0.0702,
+      "reward": 2.553897261619568,
+      "reward_std": 0.29422348737716675,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.06415844522416592,
       "rewards/tag_count_reward": 1.0,
       "step": 217
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 644.5208435058594,
-      "epoch": 0.218,
-      "grad_norm": 2.1194540845072596,
-      "kl": 0.09033203125,
-      "learning_rate": 9.623632283030077e-07,
-      "loss": 0.0715,
-      "reward": 2.900136113166809,
-      "reward_std": 0.12626729905605316,
-      "rewards/accuracy_reward": 0.9791666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.04430842399597168,
-      "rewards/tag_count_reward": 1.0,
-      "step": 218
-    },
+      "completion_length": 535.3541870117188,
+      "epoch": 0.109,
+      "grad_norm": 2.203319463752206,
+      "kl": 0.0732421875,
+      "learning_rate": 9.997779521645791e-07,
+      "loss": -0.0537,
+      "reward": 2.600339412689209,
+      "reward_std": 0.34052593261003494,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.07500774413347244,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 218
+    },
     {
       "clip_ratio": 0.0,
-      "completion_length": 621.1041870117188,
-      "epoch": 0.219,
-      "grad_norm": 2.2781852426998395,
-      "kl": 0.1044921875,
-      "learning_rate": 9.617318505324212e-07,
-      "loss": -0.0177,
-      "reward": 2.647742986679077,
-      "reward_std": 0.42986422777175903,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.0328126884996891,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 511.16668701171875,
+      "epoch": 0.1095,
+      "grad_norm": 2.6145352347525184,
+      "kl": 0.075927734375,
+      "learning_rate": 9.997525971737909e-07,
+      "loss": 0.1401,
+      "reward": 2.736908793449402,
+      "reward_std": 0.2718254253268242,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.09121626242995262,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 219
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 605.4791870117188,
-      "epoch": 0.22,
-      "grad_norm": 2.061772640741915,
-      "kl": 0.096435546875,
-      "learning_rate": 9.610954559391704e-07,
-      "loss": 0.041,
-      "reward": 2.9348647594451904,
-      "reward_std": 0.0986992521211505,
-      "rewards/accuracy_reward": 0.9791666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.03735771402716637,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 524.2500305175781,
+      "epoch": 0.11,
+      "grad_norm": 2.5498578386752473,
+      "kl": 0.07666015625,
+      "learning_rate": 9.997258721585931e-07,
+      "loss": 0.0166,
+      "reward": 2.643193244934082,
+      "reward_std": 0.3977600634098053,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.08770973607897758,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 220
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 548.5416870117188,
-      "epoch": 0.221,
-      "grad_norm": 2.174295218510334,
-      "kl": 0.10888671875,
-      "learning_rate": 9.604540522775227e-07,
-      "loss": 0.0495,
-      "reward": 2.896275758743286,
-      "reward_std": 0.1799733191728592,
-      "rewards/accuracy_reward": 0.9375,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02386310324072838,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 534.625,
+      "epoch": 0.1105,
+      "grad_norm": 2.4145494510921766,
+      "kl": 0.08544921875,
+      "learning_rate": 9.99697777200395e-07,
+      "loss": 0.0173,
+      "reward": 2.4541492462158203,
+      "reward_std": 0.46001937985420227,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.09446194767951965,
+      "rewards/tag_count_reward": 1.0,
       "step": 221
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 544.0,
-      "epoch": 0.222,
-      "grad_norm": 2.1466512297504834,
-      "kl": 0.10986328125,
-      "learning_rate": 9.598076473627796e-07,
-      "loss": 0.0578,
-      "reward": 2.740749478340149,
-      "reward_std": 0.31613367795944214,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.03182023763656616,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 586.8541870117188,
+      "epoch": 0.111,
+      "grad_norm": 2.465480916047294,
+      "kl": 0.069580078125,
+      "learning_rate": 9.996683123847795e-07,
+      "loss": 0.0974,
+      "reward": 2.7027587890625,
+      "reward_std": 0.3243703097105026,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.09238022565841675,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 222
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 535.7708435058594,
-      "epoch": 0.223,
-      "grad_norm": 2.280033428835585,
-      "kl": 0.10107421875,
-      "learning_rate": 9.59156249071181e-07,
-      "loss": 0.0484,
-      "reward": 2.7902063131332397,
-      "reward_std": 0.30508676171302795,
+      "completion_length": 557.25,
+      "epoch": 0.1115,
+      "grad_norm": 2.363152779089099,
+      "kl": 0.077880859375,
+      "learning_rate": 9.996374778015007e-07,
+      "loss": 0.075,
+      "reward": 2.7129660844802856,
+      "reward_std": 0.39234504103660583,
       "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.03618272300809622,
-      "rewards/tag_count_reward": 1.0,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.07522866874933243,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 223
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 611.3958435058594,
-      "epoch": 0.224,
-      "grad_norm": 2.78636229997098,
-      "kl": 0.103515625,
-      "learning_rate": 9.58499865339809e-07,
-      "loss": -0.0115,
-      "reward": 2.8599945306777954,
-      "reward_std": 0.2055322751402855,
-      "rewards/accuracy_reward": 0.8958333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03583900257945061,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 606.8958435058594,
+      "epoch": 0.112,
+      "grad_norm": 2.153335337523184,
+      "kl": 0.073486328125,
+      "learning_rate": 9.996052735444862e-07,
+      "loss": 0.088,
+      "reward": 2.5669835805892944,
+      "reward_std": 0.45203205943107605,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.08579418063163757,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 224
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 539.375,
-      "epoch": 0.225,
-      "grad_norm": 2.3136617005859286,
-      "kl": 0.100341796875,
-      "learning_rate": 9.578385041664925e-07,
-      "loss": 0.0279,
-      "reward": 2.5980384349823,
-      "reward_std": 0.35135801136493683,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.026961631141602993,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 568.1875305175781,
+      "epoch": 0.1125,
+      "grad_norm": 2.3030578231246324,
+      "kl": 0.069580078125,
+      "learning_rate": 9.99571699711836e-07,
+      "loss": 0.0667,
+      "reward": 2.5682613849639893,
+      "reward_std": 0.37949907779693604,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.09319691359996796,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 225
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 599.7291870117188,
-      "epoch": 0.226,
-      "grad_norm": 2.7312479125410287,
-      "kl": 0.11376953125,
-      "learning_rate": 9.571721736097088e-07,
-      "loss": -0.0362,
-      "reward": 2.461398482322693,
-      "reward_std": 0.46244145929813385,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.024712713435292244,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 600.0833435058594,
+      "epoch": 0.113,
+      "grad_norm": 2.6766438894868294,
+      "kl": 0.070556640625,
+      "learning_rate": 9.995367564058216e-07,
+      "loss": 0.2762,
+      "reward": 2.503145933151245,
+      "reward_std": 0.5531609952449799,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.12011810764670372,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 226
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 503.6458435058594,
-      "epoch": 0.227,
-      "grad_norm": 2.9850488724293416,
-      "kl": 0.11083984375,
-      "learning_rate": 9.565008817884854e-07,
-      "loss": 0.1147,
-      "reward": 2.619432210922241,
-      "reward_std": 0.47723488509655,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.028137334622442722,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 579.0833740234375,
+      "epoch": 0.1135,
+      "grad_norm": 2.2069984250119963,
+      "kl": 0.074462890625,
+      "learning_rate": 9.995004437328865e-07,
+      "loss": 0.0125,
+      "reward": 2.4455236196517944,
+      "reward_std": 0.4911371320486069,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.08919859677553177,
+      "rewards/tag_count_reward": 1.0,
       "step": 227
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 523.9583435058594,
-      "epoch": 0.228,
-      "grad_norm": 2.171102325634445,
-      "kl": 0.107421875,
-      "learning_rate": 9.55824636882301e-07,
-      "loss": -0.0038,
-      "reward": 2.581455111503601,
-      "reward_std": 0.357488214969635,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.036600593477487564,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 562.5000305175781,
+      "epoch": 0.114,
+      "grad_norm": 2.219775988144525,
+      "kl": 0.07666015625,
+      "learning_rate": 9.994627618036452e-07,
+      "loss": 0.0123,
+      "reward": 2.5935659408569336,
+      "reward_std": 0.34948965907096863,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.08351738005876541,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 228
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 552.9375152587891,
-      "epoch": 0.229,
-      "grad_norm": 2.3604731804980457,
-      "kl": 0.115234375,
-      "learning_rate": 9.55143447130987e-07,
-      "loss": 0.044,
-      "reward": 2.7417712211608887,
-      "reward_std": 0.29566267877817154,
+      "completion_length": 591.1666870117188,
+      "epoch": 0.1145,
+      "grad_norm": 2.952466021730348,
+      "kl": 0.080078125,
+      "learning_rate": 9.994237107328838e-07,
+      "loss": 0.0722,
+      "reward": 2.6423157453536987,
+      "reward_std": 0.2888629548251629,
       "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02906225249171257,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.08685098960995674,
       "rewards/tag_count_reward": 1.0,
       "step": 229
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 599.6666870117188,
-      "epoch": 0.23,
-      "grad_norm": 2.247209523128909,
-      "kl": 0.1123046875,
-      "learning_rate": 9.54457320834625e-07,
-      "loss": 0.0325,
-      "reward": 2.77982234954834,
-      "reward_std": 0.38850660622119904,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03267781715840101,
+      "completion_length": 591.8958435058594,
+      "epoch": 0.115,
+      "grad_norm": 2.1182653417835424,
+      "kl": 0.07373046875,
+      "learning_rate": 9.993832906395582e-07,
+      "loss": 0.0002,
+      "reward": 2.5170371532440186,
+      "reward_std": 0.468481108546257,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.08712967112660408,
       "rewards/tag_count_reward": 1.0,
       "step": 230
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 498.91668701171875,
-      "epoch": 0.231,
-      "grad_norm": 2.1665114909176872,
-      "kl": 0.103515625,
-      "learning_rate": 9.537662663534477e-07,
-      "loss": 0.01,
-      "reward": 2.9319541454315186,
-      "reward_std": 0.12954077869653702,
-      "rewards/accuracy_reward": 0.9791666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.03332359157502651,
+      "completion_length": 546.125,
+      "epoch": 0.1155,
+      "grad_norm": 2.0941520101521065,
+      "kl": 0.072509765625,
+      "learning_rate": 9.993415016467952e-07,
+      "loss": -0.01,
+      "reward": 2.8371798992156982,
+      "reward_std": 0.18849333748221397,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.07254217192530632,
       "rewards/tag_count_reward": 1.0,
       "step": 231
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 565.1458435058594,
-      "epoch": 0.232,
-      "grad_norm": 2.397445136474842,
-      "kl": 0.1044921875,
-      "learning_rate": 9.530702921077358e-07,
-      "loss": 0.0176,
-      "reward": 2.7926957607269287,
-      "reward_std": 0.3122672885656357,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.04063776135444641,
+      "completion_length": 600.3333587646484,
+      "epoch": 0.116,
+      "grad_norm": 2.070220926978154,
+      "kl": 0.0732421875,
+      "learning_rate": 9.992983438818915e-07,
+      "loss": 0.0136,
+      "reward": 2.483844041824341,
+      "reward_std": 0.35293829441070557,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889849185944,
+      "rewards/repetition_penalty_reward": -0.09254487603902817,
       "rewards/tag_count_reward": 1.0,
       "step": 232
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 515.5625152587891,
-      "epoch": 0.233,
-      "grad_norm": 2.3670675203551528,
-      "kl": 0.113525390625,
-      "learning_rate": 9.523694065777156e-07,
-      "loss": 0.0494,
-      "reward": 2.6802080869674683,
-      "reward_std": 0.28352178633213043,
+      "completion_length": 499.60418701171875,
+      "epoch": 0.1165,
+      "grad_norm": 3.1702925639839306,
+      "kl": 0.08349609375,
+      "learning_rate": 9.992538174763127e-07,
+      "loss": -0.0305,
+      "reward": 2.5292654037475586,
+      "reward_std": 0.334541991353035,
       "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02812536619603634,
-      "rewards/tag_count_reward": 1.0,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.09226257354021072,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 233
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 524.875,
-      "epoch": 0.234,
-      "grad_norm": 2.1646989735165527,
-      "kl": 0.121826171875,
-      "learning_rate": 9.516636183034564e-07,
-      "loss": 0.0408,
-      "reward": 2.791554570198059,
-      "reward_std": 0.38019663095474243,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.03309844061732292,
+      "completion_length": 564.8125152587891,
+      "epoch": 0.117,
+      "grad_norm": 2.3960583772955424,
+      "kl": 0.07763671875,
+      "learning_rate": 9.992079225656944e-07,
+      "loss": 0.029,
+      "reward": 2.762247085571289,
+      "reward_std": 0.279547318816185,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.09018350392580032,
       "rewards/tag_count_reward": 0.984375,
       "step": 234
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 560.0208435058594,
-      "epoch": 0.235,
-      "grad_norm": 2.2756987535611404,
-      "kl": 0.10693359375,
-      "learning_rate": 9.509529358847654e-07,
-      "loss": 0.0085,
-      "reward": 2.8194351196289062,
-      "reward_std": 0.3406111150979996,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03473140113055706,
+      "completion_length": 607.2083435058594,
+      "epoch": 0.1175,
+      "grad_norm": 2.042913157935333,
+      "kl": 0.068115234375,
+      "learning_rate": 9.9916065928984e-07,
+      "loss": -0.0039,
+      "reward": 2.4307457208633423,
+      "reward_std": 0.44265393912792206,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.902777761220932,
+      "rewards/repetition_penalty_reward": -0.11786547303199768,
       "rewards/tag_count_reward": 1.0,
       "step": 235
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 575.375,
-      "epoch": 0.236,
-      "grad_norm": 2.174744939943205,
-      "kl": 0.100830078125,
-      "learning_rate": 9.502373679810839e-07,
-      "loss": -0.0023,
-      "reward": 2.7065417766571045,
-      "reward_std": 0.15310228383168578,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.03651383891701698,
+      "completion_length": 533.5208435058594,
+      "epoch": 0.118,
+      "grad_norm": 2.2072960786944313,
+      "kl": 0.073974609375,
+      "learning_rate": 9.991120277927223e-07,
+      "loss": 0.0585,
+      "reward": 2.7634425163269043,
+      "reward_std": 0.2200346365571022,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.10461306571960449,
       "rewards/tag_count_reward": 1.0,
       "step": 236
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 484.9583435058594,
-      "epoch": 0.237,
-      "grad_norm": 2.6379983576818242,
-      "kl": 0.119140625,
-      "learning_rate": 9.495169233113806e-07,
-      "loss": 0.0566,
-      "reward": 2.538248300552368,
-      "reward_std": 0.3112593740224838,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.031196235679090023,
+      "completion_length": 564.6458435058594,
+      "epoch": 0.1185,
+      "grad_norm": 2.2413703328479935,
+      "kl": 0.076171875,
+      "learning_rate": 9.990620282224806e-07,
+      "loss": -0.024,
+      "reward": 2.489652156829834,
+      "reward_std": 0.32129333913326263,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.09368134289979935,
       "rewards/tag_count_reward": 1.0,
       "step": 237
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 565.7916870117188,
-      "epoch": 0.238,
-      "grad_norm": 2.052140048933374,
-      "kl": 0.114990234375,
-      "learning_rate": 9.487916106540465e-07,
-      "loss": -0.0027,
-      "reward": 2.778996706008911,
-      "reward_std": 0.32464583218097687,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03350352682173252,
+      "completion_length": 509.1041717529297,
+      "epoch": 0.119,
+      "grad_norm": 2.255281998818256,
+      "kl": 0.071044921875,
+      "learning_rate": 9.990106607314225e-07,
+      "loss": 0.0002,
+      "reward": 2.6565502882003784,
+      "reward_std": 0.40016523003578186,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.07261638343334198,
       "rewards/tag_count_reward": 1.0,
       "step": 238
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 507.2291717529297,
-      "epoch": 0.239,
-      "grad_norm": 2.3870284690729915,
-      "kl": 0.103515625,
-      "learning_rate": 9.480614388467877e-07,
-      "loss": 0.0266,
-      "reward": 2.865510582923889,
-      "reward_std": 0.2045399323105812,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.037267398089170456,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 528.0833587646484,
+      "epoch": 0.1195,
+      "grad_norm": 2.188877162746338,
+      "kl": 0.07568359375,
+      "learning_rate": 9.989579254760224e-07,
+      "loss": -0.0314,
+      "reward": 2.826788544654846,
+      "reward_std": 0.20467501878738403,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.05862821638584137,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 239
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 473.0833435058594,
-      "epoch": 0.24,
-      "grad_norm": 2.340901673290699,
-      "kl": 0.120849609375,
-      "learning_rate": 9.473264167865171e-07,
-      "loss": -0.0045,
-      "reward": 2.849257707595825,
-      "reward_std": 0.26365862786769867,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.018797898665070534,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 568.8125305175781,
+      "epoch": 0.12,
+      "grad_norm": 2.434441145634029,
+      "kl": 0.074951171875,
+      "learning_rate": 9.989038226169207e-07,
+      "loss": 0.1175,
+      "reward": 2.318480372428894,
+      "reward_std": 0.2832699418067932,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.07908925786614418,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 240
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 502.3125,
-      "epoch": 0.241,
-      "grad_norm": 1.9771145576165876,
-      "kl": 0.10791015625,
-      "learning_rate": 9.465865534292464e-07,
-      "loss": 0.0245,
-      "reward": 2.665071487426758,
-      "reward_std": 0.25069355964660645,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.036317549645900726,
+      "completion_length": 533.5416870117188,
+      "epoch": 0.1205,
+      "grad_norm": 2.378465168474232,
+      "kl": 0.081787109375,
+      "learning_rate": 9.988483523189248e-07,
+      "loss": 0.0519,
+      "reward": 2.758083462715149,
+      "reward_std": 0.313043013215065,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.07524998486042023,
       "rewards/tag_count_reward": 1.0,
       "step": 241
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 469.0625,
-      "epoch": 0.242,
-      "grad_norm": 2.429463773148552,
-      "kl": 0.12890625,
-      "learning_rate": 9.458418577899774e-07,
-      "loss": 0.0219,
-      "reward": 2.832092523574829,
-      "reward_std": 0.3335033804178238,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.015129742678254843,
+      "completion_length": 646.6041870117188,
+      "epoch": 0.121,
+      "grad_norm": 2.0129363099493394,
+      "kl": 0.070068359375,
+      "learning_rate": 9.98791514751006e-07,
+      "loss": -0.0271,
+      "reward": 2.4740262031555176,
+      "reward_std": 0.3102487027645111,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.08847374841570854,
       "rewards/tag_count_reward": 1.0,
       "step": 242
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 509.1041717529297,
-      "epoch": 0.243,
-      "grad_norm": 2.2888490947191453,
-      "kl": 0.11669921875,
-      "learning_rate": 9.450923389425911e-07,
-      "loss": 0.0679,
-      "reward": 2.8057637214660645,
-      "reward_std": 0.3620642125606537,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.027569908648729324,
+      "completion_length": 495.85418701171875,
+      "epoch": 0.1215,
+      "grad_norm": 2.3579635378621226,
+      "kl": 0.080810546875,
+      "learning_rate": 9.98733310086302e-07,
+      "loss": 0.0068,
+      "reward": 2.6361684799194336,
+      "reward_std": 0.22484752535820007,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.0860537700355053,
       "rewards/tag_count_reward": 1.0,
       "step": 243
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 462.3958435058594,
-      "epoch": 0.244,
-      "grad_norm": 2.7695341753126383,
-      "kl": 0.12841796875,
-      "learning_rate": 9.443380060197385e-07,
-      "loss": 0.028,
-      "reward": 2.687015175819397,
-      "reward_std": 0.19807551801204681,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.02305438881739974,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 570.1250305175781,
+      "epoch": 0.122,
+      "grad_norm": 2.2032148417869277,
+      "kl": 0.0771484375,
+      "learning_rate": 9.98673738502114e-07,
+      "loss": -0.0345,
+      "reward": 2.1713971495628357,
+      "reward_std": 0.3244406059384346,
+      "rewards/accuracy_reward": 0.27083333395421505,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.07860295474529266,
+      "rewards/tag_count_reward": 1.0,
       "step": 244
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 562.0625152587891,
-      "epoch": 0.245,
-      "grad_norm": 2.0280976947569536,
-      "kl": 0.12060546875,
-      "learning_rate": 9.43578868212728e-07,
-      "loss": -0.044,
-      "reward": 2.79758882522583,
-      "reward_std": 0.2576555460691452,
-      "rewards/accuracy_reward": 0.875,
-      "rewards/reasoning_steps_reward": 0.9513889849185944,
-      "rewards/repetition_penalty_reward": -0.028800049796700478,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 607.0833435058594,
+      "epoch": 0.1225,
+      "grad_norm": 2.343457795616331,
+      "kl": 0.07861328125,
+      "learning_rate": 9.986128001799076e-07,
+      "loss": -0.0081,
+      "reward": 2.586448907852173,
+      "reward_std": 0.3658638447523117,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.08889832720160484,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 245
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 529.3333435058594,
-      "epoch": 0.246,
-      "grad_norm": 2.3299542232310477,
-      "kl": 0.119873046875,
-      "learning_rate": 9.428149347714143e-07,
-      "loss": -0.0061,
-      "reward": 2.660530924797058,
-      "reward_std": 0.16930758208036423,
-      "rewards/accuracy_reward": 0.6875000149011612,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026969049125909805,
-      "rewards/tag_count_reward": 1.0,
-      "step": 246
+      "completion_length": 528.2708435058594,
+      "epoch": 0.123,
+      "grad_norm": 2.1457635532124315,
+      "kl": 0.09033203125,
+      "learning_rate": 9.985504953053113e-07,
+      "loss": 0.0243,
+      "reward": 2.761418104171753,
+      "reward_std": 0.19703956693410873,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.07365139201283455,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 246
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 488.72918701171875,
-      "epoch": 0.247,
-      "grad_norm": 2.300003946060058,
-      "kl": 0.1318359375,
-      "learning_rate": 9.420462150040852e-07,
-      "loss": 0.0401,
-      "reward": 2.5491377115249634,
-      "reward_std": 0.3199358731508255,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.027251286432147026,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 606.6458435058594,
+      "epoch": 0.1235,
+      "grad_norm": 2.3768865639458063,
+      "kl": 0.083740234375,
+      "learning_rate": 9.984868240681164e-07,
+      "loss": 0.061,
+      "reward": 2.5193967819213867,
+      "reward_std": 0.4161098003387451,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.08129774034023285,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 247
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 528.2291870117188,
-      "epoch": 0.248,
-      "grad_norm": 2.242455600453007,
-      "kl": 0.119384765625,
-      "learning_rate": 9.412727182773486e-07,
-      "loss": 0.0059,
-      "reward": 2.7635743618011475,
-      "reward_std": 0.3389218971133232,
-      "rewards/accuracy_reward": 0.8125,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02809229213744402,
+      "completion_length": 626.9375305175781,
+      "epoch": 0.124,
+      "grad_norm": 2.4645417816762683,
+      "kl": 0.077392578125,
+      "learning_rate": 9.98421786662277e-07,
+      "loss": 0.148,
+      "reward": 2.6498433351516724,
+      "reward_std": 0.4323410838842392,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.08626773580908775,
       "rewards/tag_count_reward": 1.0,
       "step": 248
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 639.5625305175781,
-      "epoch": 0.249,
-      "grad_norm": 1.9733887624956274,
-      "kl": 0.110595703125,
-      "learning_rate": 9.404944540160177e-07,
-      "loss": -0.0684,
-      "reward": 2.6618722677230835,
-      "reward_std": 0.2736282553523779,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.03951651602983475,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 543.25,
+      "epoch": 0.1245,
+      "grad_norm": 2.3866168405234474,
+      "kl": 0.081298828125,
+      "learning_rate": 9.983553832859078e-07,
+      "loss": 0.0302,
+      "reward": 2.5024802684783936,
+      "reward_std": 0.43941864371299744,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.07043641060590744,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 249
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 501.12501525878906,
-      "epoch": 0.25,
-      "grad_norm": 2.167863484844858,
-      "kl": 0.13525390625,
-      "learning_rate": 9.397114317029974e-07,
-      "loss": 0.004,
-      "reward": 2.526219606399536,
-      "reward_std": 0.380461186170578,
-      "rewards/accuracy_reward": 0.5833333730697632,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.0293360473588109,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 571.1875305175781,
+      "epoch": 0.125,
+      "grad_norm": 2.425419283542172,
+      "kl": 0.112060546875,
+      "learning_rate": 9.982876141412855e-07,
+      "loss": 0.034,
+      "reward": 2.78093945980072,
+      "reward_std": 0.2996261715888977,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.07496321201324463,
+      "rewards/tag_count_reward": 0.9739583730697632,
       "step": 250
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 592.2083435058594,
-      "epoch": 0.251,
-      "grad_norm": 2.441314951742952,
-      "kl": 0.11767578125,
-      "learning_rate": 9.38923660879167e-07,
-      "loss": 0.1345,
-      "reward": 2.7307329177856445,
-      "reward_std": 0.2994953393936157,
-      "rewards/accuracy_reward": 0.7708333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.033156177029013634,
+      "completion_length": 620.5,
+      "epoch": 0.1255,
+      "grad_norm": 2.129504256319349,
+      "kl": 0.080322265625,
+      "learning_rate": 9.982184794348462e-07,
+      "loss": 0.0166,
+      "reward": 2.693773031234741,
+      "reward_std": 0.40914659202098846,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.08400484919548035,
       "rewards/tag_count_reward": 1.0,
       "step": 251
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 564.2708435058594,
-      "epoch": 0.252,
-      "grad_norm": 2.49677783518446,
-      "kl": 0.122802734375,
-      "learning_rate": 9.381311511432658e-07,
-      "loss": 0.1067,
-      "reward": 2.4708348512649536,
-      "reward_std": 0.36502179503440857,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.036109643056988716,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 534.3958435058594,
+      "epoch": 0.126,
+      "grad_norm": 1.8697440201638618,
+      "kl": 0.087890625,
+      "learning_rate": 9.981479793771866e-07,
+      "loss": 0.0042,
+      "reward": 2.5290188789367676,
+      "reward_std": 0.21817893348634243,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.10466165468096733,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 252
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 601.3125305175781,
-      "epoch": 0.253,
-      "grad_norm": 2.031276195067717,
-      "kl": 0.119384765625,
-      "learning_rate": 9.373339121517746e-07,
-      "loss": 0.0435,
-      "reward": 2.9231903553009033,
-      "reward_std": 0.10614164918661118,
-      "rewards/accuracy_reward": 0.9791666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.04382369481027126,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 609.8958587646484,
+      "epoch": 0.1265,
+      "grad_norm": 2.3009060679085156,
+      "kl": 0.078369140625,
+      "learning_rate": 9.98076114183062e-07,
+      "loss": 0.1217,
+      "reward": 2.6637524366378784,
+      "reward_std": 0.40659724175930023,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06194208189845085,
+      "rewards/tag_count_reward": 0.96875,
       "step": 253
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 572.4791870117188,
-      "epoch": 0.254,
-      "grad_norm": 2.1020100515227313,
-      "kl": 0.12060546875,
-      "learning_rate": 9.36531953618799e-07,
-      "loss": 0.0657,
-      "reward": 2.7917503118515015,
-      "reward_std": 0.26876559667289257,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.029430264607071877,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 720.5416870117188,
+      "epoch": 0.127,
+      "grad_norm": 2.2567238155019322,
+      "kl": 0.080078125,
+      "learning_rate": 9.98002884071386e-07,
+      "loss": 0.0781,
+      "reward": 2.4784340858459473,
+      "reward_std": 0.4698047339916229,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.11531602591276169,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 254
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 553.8333740234375,
-      "epoch": 0.255,
-      "grad_norm": 3.5327461787319776,
-      "kl": 0.13232421875,
-      "learning_rate": 9.357252853159505e-07,
-      "loss": 0.0351,
-      "reward": 2.589944839477539,
-      "reward_std": 0.22619716823101044,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02811075560748577,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 644.8125305175781,
+      "epoch": 0.1275,
+      "grad_norm": 1.972516265850252,
+      "kl": 0.081298828125,
+      "learning_rate": 9.979282892652304e-07,
+      "loss": 0.0858,
+      "reward": 2.5695180892944336,
+      "reward_std": 0.34912221878767014,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.11277362704277039,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 255
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 548.0416870117188,
-      "epoch": 0.256,
-      "grad_norm": 2.183159179082376,
-      "kl": 0.13525390625,
-      "learning_rate": 9.34913917072228e-07,
-      "loss": -0.0696,
-      "reward": 2.6049808263778687,
-      "reward_std": 0.3286217898130417,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.028699966147542,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 641.7708435058594,
+      "epoch": 0.128,
+      "grad_norm": 2.042073924733626,
+      "kl": 0.08740234375,
+      "learning_rate": 9.97852329991824e-07,
+      "loss": 0.0308,
+      "reward": 2.818283200263977,
+      "reward_std": 0.27264343202114105,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06886980310082436,
+      "rewards/tag_count_reward": 0.984375,
       "step": 256
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 580.6041870117188,
-      "epoch": 0.257,
-      "grad_norm": 2.401471424731539,
-      "kl": 0.12744140625,
-      "learning_rate": 9.340978587738972e-07,
-      "loss": 0.0462,
-      "reward": 2.8021459579467773,
-      "reward_std": 0.3006982207298279,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.031187265180051327,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 572.0625305175781,
+      "epoch": 0.1285,
+      "grad_norm": 2.101899385079832,
+      "kl": 0.08544921875,
+      "learning_rate": 9.977750064825519e-07,
+      "loss": -0.024,
+      "reward": 2.627909779548645,
+      "reward_std": 0.3649473935365677,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.08215974271297455,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 257
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 570.6041870117188,
-      "epoch": 0.258,
-      "grad_norm": 7.315971327716766,
-      "kl": 0.12353515625,
-      "learning_rate": 9.332771203643714e-07,
-      "loss": 0.1692,
-      "reward": 2.6945351362228394,
-      "reward_std": 0.3729802221059799,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.0398398581892252,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 498.64585876464844,
+      "epoch": 0.129,
+      "grad_norm": 2.2345487850590944,
+      "kl": 0.088134765625,
+      "learning_rate": 9.976963189729547e-07,
+      "loss": -0.0191,
+      "reward": 2.7989070415496826,
+      "reward_std": 0.24870866537094116,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.06220416724681854,
+      "rewards/tag_count_reward": 1.0,
       "step": 258
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 462.18751525878906,
-      "epoch": 0.259,
-      "grad_norm": 2.249715244876303,
-      "kl": 0.1328125,
-      "learning_rate": 9.324517118440888e-07,
-      "loss": 0.0048,
-      "reward": 2.757077693939209,
-      "reward_std": 0.31304194778203964,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.020700284279882908,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 715.6250305175781,
+      "epoch": 0.1295,
+      "grad_norm": 1.8695009477650302,
+      "kl": 0.07861328125,
+      "learning_rate": 9.976162677027284e-07,
+      "loss": 0.0655,
+      "reward": 2.5826069116592407,
+      "reward_std": 0.4399617910385132,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.11010143533349037,
+      "rewards/tag_count_reward": 0.984375,
       "step": 259
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 508.2083435058594,
-      "epoch": 0.26,
-      "grad_norm": 5.193269569172354,
-      "kl": 0.1416015625,
-      "learning_rate": 9.316216432703916e-07,
-      "loss": 0.1727,
-      "reward": 2.8865363597869873,
-      "reward_std": 0.19043887220323086,
-      "rewards/accuracy_reward": 0.9583333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.04228321462869644,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 501.1458435058594,
+      "epoch": 0.13,
+      "grad_norm": 2.1459813731795068,
+      "kl": 0.08251953125,
+      "learning_rate": 9.975348529157229e-07,
+      "loss": 0.0961,
+      "reward": 2.8910681009292603,
+      "reward_std": 0.10162430070340633,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0776820220053196,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 260
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 496.58335876464844,
-      "epoch": 0.261,
-      "grad_norm": 2.2509057041996443,
-      "kl": 0.13623046875,
-      "learning_rate": 9.307869247574038e-07,
-      "loss": 0.0359,
-      "reward": 2.8467074632644653,
-      "reward_std": 0.2996975928544998,
-      "rewards/accuracy_reward": 0.8958333730697632,
-      "rewards/reasoning_steps_reward": 0.9791667461395264,
-      "rewards/repetition_penalty_reward": -0.02829269878566265,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 555.25,
+      "epoch": 0.1305,
+      "grad_norm": 2.0994299368541287,
+      "kl": 0.0859375,
+      "learning_rate": 9.974520748599421e-07,
+      "loss": 0.0963,
+      "reward": 2.3302276134490967,
+      "reward_std": 0.17742525041103363,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.07775864750146866,
+      "rewards/tag_count_reward": 0.984375,
       "step": 261
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 628.9375,
-      "epoch": 0.262,
-      "grad_norm": 4.668495262967337,
-      "kl": 0.13623046875,
-      "learning_rate": 9.299475664759068e-07,
-      "loss": 0.1609,
-      "reward": 2.720739483833313,
-      "reward_std": 0.4649975746870041,
-      "rewards/accuracy_reward": 0.8125,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.04141336679458618,
-      "rewards/tag_count_reward": 0.9843750298023224,
+      "completion_length": 707.6041870117188,
+      "epoch": 0.131,
+      "grad_norm": 2.179306110235832,
+      "kl": 0.09033203125,
+      "learning_rate": 9.973679337875418e-07,
+      "loss": 0.0756,
+      "reward": 2.457505941390991,
+      "reward_std": 0.49539755284786224,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.09978591650724411,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 262
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 519.0,
-      "epoch": 0.263,
-      "grad_norm": 2.3541332969660975,
-      "kl": 0.13671875,
-      "learning_rate": 9.291035786532163e-07,
-      "loss": 0.0638,
-      "reward": 2.4355462789535522,
-      "reward_std": 0.21204119874164462,
-      "rewards/accuracy_reward": 0.4791666716337204,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.033203769475221634,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 616.2916870117188,
+      "epoch": 0.1315,
+      "grad_norm": 2.0703888587416466,
+      "kl": 0.08544921875,
+      "learning_rate": 9.972824299548309e-07,
+      "loss": 0.0048,
+      "reward": 2.7762060165405273,
+      "reward_std": 0.2107193972915411,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.07796090468764305,
+      "rewards/tag_count_reward": 1.0,
       "step": 263
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 546.3125,
-      "epoch": 0.264,
-      "grad_norm": 2.1456751308727213,
-      "kl": 0.1416015625,
-      "learning_rate": 9.282549715730579e-07,
-      "loss": -0.0358,
-      "reward": 2.4012598991394043,
-      "reward_std": 0.44198279082775116,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.03971245139837265,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 475.9166717529297,
+      "epoch": 0.132,
+      "grad_norm": 2.170217179415964,
+      "kl": 0.090087890625,
+      "learning_rate": 9.971955636222684e-07,
+      "loss": 0.0146,
+      "reward": 2.930613875389099,
+      "reward_std": 0.0612574927508831,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.05549720861017704,
+      "rewards/tag_count_reward": 1.0,
       "step": 264
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 460.12501525878906,
-      "epoch": 0.265,
-      "grad_norm": 2.3449066935567697,
-      "kl": 0.15087890625,
-      "learning_rate": 9.274017555754407e-07,
-      "loss": 0.0136,
-      "reward": 2.7137099504470825,
-      "reward_std": 0.326103575527668,
-      "rewards/accuracy_reward": 0.7708333730697632,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.02240123925730586,
+      "completion_length": 510.8958435058594,
+      "epoch": 0.1325,
+      "grad_norm": 2.0247597136853077,
+      "kl": 0.093994140625,
+      "learning_rate": 9.971073350544644e-07,
+      "loss": 0.0141,
+      "reward": 2.8639657497406006,
+      "reward_std": 0.18884775042533875,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.06658982113003731,
       "rewards/tag_count_reward": 1.0,
       "step": 265
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 526.7291870117188,
-      "epoch": 0.266,
-      "grad_norm": 2.521194502817804,
-      "kl": 0.1572265625,
-      "learning_rate": 9.265439410565328e-07,
-      "loss": 0.1172,
-      "reward": 2.733197331428528,
-      "reward_std": 0.43402716517448425,
-      "rewards/accuracy_reward": 0.8125,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.028955545276403427,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 613.6041870117188,
+      "epoch": 0.133,
+      "grad_norm": 2.819432158137642,
+      "kl": 0.124267578125,
+      "learning_rate": 9.970177445201783e-07,
+      "loss": 0.0593,
+      "reward": 2.438739776611328,
+      "reward_std": 0.35922619700431824,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.07688540033996105,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 266
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 580.3541870117188,
-      "epoch": 0.267,
-      "grad_norm": 2.2571936139784374,
-      "kl": 0.15185546875,
-      "learning_rate": 9.256815384685328e-07,
-      "loss": 0.0205,
-      "reward": 2.665441632270813,
-      "reward_std": 0.2782168686389923,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.029003008268773556,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 521.1458587646484,
+      "epoch": 0.1335,
+      "grad_norm": 2.046082487346144,
+      "kl": 0.101806640625,
+      "learning_rate": 9.969267922923188e-07,
+      "loss": 0.0155,
+      "reward": 2.68787944316864,
+      "reward_std": 0.290618859231472,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.09163457155227661,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 267
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 528.8541870117188,
-      "epoch": 0.268,
-      "grad_norm": 2.527114445757883,
-      "kl": 0.15478515625,
-      "learning_rate": 9.248145583195447e-07,
-      "loss": 0.0642,
-      "reward": 2.634427309036255,
-      "reward_std": 0.35257890820503235,
-      "rewards/accuracy_reward": 0.6666666716337204,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.025295117869973183,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 531.6041870117188,
+      "epoch": 0.134,
+      "grad_norm": 2.415289797288351,
+      "kl": 0.10546875,
+      "learning_rate": 9.968344786479415e-07,
+      "loss": 0.0305,
+      "reward": 2.6710007190704346,
+      "reward_std": 0.35774578154087067,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.05469386838376522,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 268
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 480.4791717529297,
-      "epoch": 0.269,
-      "grad_norm": 2.565433677851082,
-      "kl": 0.15185546875,
-      "learning_rate": 9.239430111734476e-07,
-      "loss": -0.0129,
-      "reward": 2.6094166040420532,
-      "reward_std": 0.32510536164045334,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.02252791728824377,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 486.10418701171875,
+      "epoch": 0.1345,
+      "grad_norm": 2.0309227835825108,
+      "kl": 0.10546875,
+      "learning_rate": 9.967408038682505e-07,
+      "loss": 0.0142,
+      "reward": 2.8688745498657227,
+      "reward_std": 0.1512543261051178,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.06688930839300156,
+      "rewards/tag_count_reward": 0.984375,
       "step": 269
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 535.3333435058594,
-      "epoch": 0.27,
-      "grad_norm": 7.328177927840423,
-      "kl": 0.15185546875,
-      "learning_rate": 9.230669076497687e-07,
-      "loss": 0.1255,
-      "reward": 2.5093064308166504,
-      "reward_std": 0.3799414336681366,
-      "rewards/accuracy_reward": 0.5625000149011612,
+      "completion_length": 502.2708435058594,
+      "epoch": 0.135,
+      "grad_norm": 2.2773691497710242,
+      "kl": 0.1025390625,
+      "learning_rate": 9.96645768238595e-07,
+      "loss": 0.0202,
+      "reward": 2.5804989337921143,
+      "reward_std": 0.20569386333227158,
+      "rewards/accuracy_reward": 0.708333358168602,
       "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.02367980219423771,
-      "rewards/tag_count_reward": 0.984375,
+      "rewards/repetition_penalty_reward": -0.10873730108141899,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 270
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 509.85418701171875,
-      "epoch": 0.271,
-      "grad_norm": 2.718362036973218,
-      "kl": 0.15576171875,
-      "learning_rate": 9.221862584235526e-07,
-      "loss": 0.0344,
-      "reward": 2.775928258895874,
-      "reward_std": 0.3645378649234772,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.02962728962302208,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 519.25,
+      "epoch": 0.1355,
+      "grad_norm": 2.1749339879371896,
+      "kl": 0.106689453125,
+      "learning_rate": 9.965493720484698e-07,
+      "loss": 0.0813,
+      "reward": 2.627415657043457,
+      "reward_std": 0.32304753363132477,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.06702888198196888,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 271
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 511.50001525878906,
-      "epoch": 0.272,
-      "grad_norm": 2.272195946333135,
-      "kl": 0.1591796875,
-      "learning_rate": 9.213010742252327e-07,
-      "loss": -0.0293,
-      "reward": 2.742081642150879,
-      "reward_std": 0.2952452003955841,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.03048785123974085,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 632.9375305175781,
+      "epoch": 0.136,
+      "grad_norm": 1.908928580484598,
+      "kl": 0.12060546875,
+      "learning_rate": 9.964516155915151e-07,
+      "loss": 0.061,
+      "reward": 2.5479973554611206,
+      "reward_std": 0.32585373520851135,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.0770026333630085,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 272
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 585.25,
-      "epoch": 0.273,
-      "grad_norm": 2.32139747171617,
-      "kl": 0.14794921875,
-      "learning_rate": 9.204113658404989e-07,
-      "loss": -0.0799,
-      "reward": 2.521125078201294,
-      "reward_std": 0.2736772522330284,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.04137510620057583,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 619.9791870117188,
+      "epoch": 0.1365,
+      "grad_norm": 1.9025926132999502,
+      "kl": 0.1201171875,
+      "learning_rate": 9.963524991655133e-07,
+      "loss": -0.02,
+      "reward": 2.6221734285354614,
+      "reward_std": 0.2964826971292496,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.07053492963314056,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 273
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 559.2708435058594,
-      "epoch": 0.274,
-      "grad_norm": 3.852408203301354,
-      "kl": 0.1982421875,
-      "learning_rate": 9.195171441101668e-07,
-      "loss": -0.0098,
-      "reward": 2.5533807277679443,
-      "reward_std": 0.35553011298179626,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.019536098465323448,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 577.2291870117188,
+      "epoch": 0.137,
+      "grad_norm": 2.695248913656631,
+      "kl": 0.11669921875,
+      "learning_rate": 9.962520230723906e-07,
+      "loss": -0.0972,
+      "reward": 2.581347942352295,
+      "reward_std": 0.22622781991958618,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.08184656128287315,
+      "rewards/tag_count_reward": 0.96875,
       "step": 274
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 633.4583435058594,
-      "epoch": 0.275,
-      "grad_norm": 2.0627187344345166,
-      "kl": 0.1513671875,
-      "learning_rate": 9.186184199300463e-07,
-      "loss": -0.0293,
-      "reward": 2.7091498374938965,
-      "reward_std": 0.1623903214931488,
+      "completion_length": 623.5208435058594,
+      "epoch": 0.1375,
+      "grad_norm": 3.342402691090167,
+      "kl": 0.12939453125,
+      "learning_rate": 9.961501876182148e-07,
+      "loss": 0.2172,
+      "reward": 2.620313286781311,
+      "reward_std": 0.4897034168243408,
       "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.04432236962020397,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.0689227245748043,
+      "rewards/tag_count_reward": 0.953125,
       "step": 275
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 532.0416717529297,
-      "epoch": 0.276,
-      "grad_norm": 2.401556778445366,
-      "kl": 0.1552734375,
-      "learning_rate": 9.177152042508077e-07,
-      "loss": 0.0535,
-      "reward": 2.728344440460205,
-      "reward_std": 0.3177375793457031,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.025128038600087166,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 522.1666717529297,
+      "epoch": 0.138,
+      "grad_norm": 2.3942345495108803,
+      "kl": 0.12939453125,
+      "learning_rate": 9.960469931131936e-07,
+      "loss": 0.0718,
+      "reward": 2.6490787267684937,
+      "reward_std": 0.3025604486465454,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.05925481766462326,
+      "rewards/tag_count_reward": 1.0,
       "step": 276
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 497.7500305175781,
-      "epoch": 0.277,
-      "grad_norm": 2.544090680740168,
-      "kl": 0.17333984375,
-      "learning_rate": 9.168075080778494e-07,
-      "loss": 0.0797,
-      "reward": 2.892575979232788,
-      "reward_std": 0.2266939841210842,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.01714644394814968,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 506.5,
+      "epoch": 0.1385,
+      "grad_norm": 2.3474305496678625,
+      "kl": 0.1181640625,
+      "learning_rate": 9.959424398716763e-07,
+      "loss": 0.0796,
+      "reward": 2.911013603210449,
+      "reward_std": 0.09207849018275738,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.06294489465653896,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 277
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 496.6666717529297,
-      "epoch": 0.278,
-      "grad_norm": 2.35427197118846,
-      "kl": 0.15576171875,
-      "learning_rate": 9.158953424711624e-07,
-      "loss": 0.0443,
-      "reward": 2.8280975818634033,
-      "reward_std": 0.15652123093605042,
-      "rewards/accuracy_reward": 0.875,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.029541232623159885,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 627.6666870117188,
+      "epoch": 0.139,
+      "grad_norm": 2.6905854255736252,
+      "kl": 0.15625,
+      "learning_rate": 9.958365282121496e-07,
+      "loss": 0.0858,
+      "reward": 2.7694180011749268,
+      "reward_std": 0.29491981118917465,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.06738761439919472,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 278
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 505.7708435058594,
-      "epoch": 0.279,
-      "grad_norm": 2.8854860164238882,
-      "kl": 0.1611328125,
-      "learning_rate": 9.149787185451969e-07,
-      "loss": 0.1633,
-      "reward": 2.6974780559539795,
-      "reward_std": 0.26681460440158844,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.019535831175744534,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 510.91668701171875,
+      "epoch": 0.1395,
+      "grad_norm": 2.175289440432667,
+      "kl": 0.16796875,
+      "learning_rate": 9.95729258457239e-07,
+      "loss": 0.0046,
+      "reward": 2.836862325668335,
+      "reward_std": 0.21854999661445618,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.08674902468919754,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 279
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 544.1458587646484,
-      "epoch": 0.28,
-      "grad_norm": 2.1507469824718592,
-      "kl": 0.15576171875,
-      "learning_rate": 9.140576474687263e-07,
-      "loss": 0.0214,
-      "reward": 2.660393714904785,
-      "reward_std": 0.35248108208179474,
-      "rewards/accuracy_reward": 0.708333358168602,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02884255349636078,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 577.4583435058594,
+      "epoch": 0.14,
+      "grad_norm": 5.672542060786621,
+      "kl": 0.24072265625,
+      "learning_rate": 9.956206309337066e-07,
+      "loss": -0.005,
+      "reward": 2.5967882871627808,
+      "reward_std": 0.48753632605075836,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.075086809694767,
+      "rewards/tag_count_reward": 0.984375,
       "step": 280
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 582.9791870117188,
-      "epoch": 0.281,
-      "grad_norm": 2.2081213981330974,
-      "kl": 0.1484375,
-      "learning_rate": 9.131321404647109e-07,
-      "loss": 0.0259,
-      "reward": 2.749055504798889,
-      "reward_std": 0.3296816051006317,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.035666871815919876,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 589.9791870117188,
+      "epoch": 0.1405,
+      "grad_norm": 5.053359884020266,
+      "kl": 0.23046875,
+      "learning_rate": 9.955106459724508e-07,
+      "loss": 0.1177,
+      "reward": 2.557823896408081,
+      "reward_std": 0.23683376610279083,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.0741206556558609,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 281
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 495.8958435058594,
-      "epoch": 0.282,
-      "grad_norm": 2.6311621368175055,
-      "kl": 0.185546875,
-      "learning_rate": 9.122022088101613e-07,
-      "loss": 0.0722,
-      "reward": 2.718705654144287,
-      "reward_std": 0.37944258749485016,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.01740554254502058,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 647.2291870117188,
+      "epoch": 0.141,
+      "grad_norm": 3.63123333573302,
+      "kl": 0.2919921875,
+      "learning_rate": 9.953993039085048e-07,
+      "loss": 0.094,
+      "reward": 2.555752992630005,
+      "reward_std": 0.38118627667427063,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.06751113198697567,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 282
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 585.125,
-      "epoch": 0.283,
-      "grad_norm": 2.435559416337875,
-      "kl": 0.16943359375,
-      "learning_rate": 9.112678638360015e-07,
-      "loss": 0.0748,
-      "reward": 2.535888671875,
-      "reward_std": 0.3460448384284973,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.04918085224926472,
-      "rewards/tag_count_reward": 0.9739583730697632,
+      "completion_length": 848.5000305175781,
+      "epoch": 0.1415,
+      "grad_norm": 12.94083486478269,
+      "kl": 0.4365234375,
+      "learning_rate": 9.952866050810363e-07,
+      "loss": 0.2525,
+      "reward": 2.34884250164032,
+      "reward_std": 0.5896148979663849,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.07476872578263283,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 283
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 568.4583740234375,
-      "epoch": 0.284,
-      "grad_norm": 2.3346177606876206,
-      "kl": 0.1689453125,
-      "learning_rate": 9.103291169269299e-07,
-      "loss": 0.0921,
-      "reward": 2.8603535890579224,
-      "reward_std": 0.25171563029289246,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.0389520637691021,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 537.2500305175781,
+      "epoch": 0.142,
+      "grad_norm": 921.7663615274216,
+      "kl": 3.4921875,
+      "learning_rate": 9.951725498333448e-07,
+      "loss": 0.2981,
+      "reward": 2.666219711303711,
+      "reward_std": 0.38224542140960693,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.06121090054512024,
+      "rewards/tag_count_reward": 0.984375,
       "step": 284
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 663.0416870117188,
-      "epoch": 0.285,
-      "grad_norm": 2.193370842368294,
-      "kl": 0.15673828125,
-      "learning_rate": 9.093859795212817e-07,
-      "loss": 0.1456,
-      "reward": 2.744846820831299,
-      "reward_std": 0.27857429534196854,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.04508392512798309,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 840.1458740234375,
+      "epoch": 0.1425,
+      "grad_norm": 1120.8208786375085,
+      "kl": 8.21875,
+      "learning_rate": 9.950571385128625e-07,
+      "loss": 0.488,
+      "reward": 2.1636295914649963,
+      "reward_std": 0.5795368552207947,
+      "rewards/accuracy_reward": 0.3333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.07768993638455868,
+      "rewards/tag_count_reward": 0.921875,
       "step": 285
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 563.3125305175781,
-      "epoch": 0.286,
-      "grad_norm": 2.4347017189444737,
-      "kl": 0.20166015625,
-      "learning_rate": 9.084384631108882e-07,
-      "loss": 0.0819,
-      "reward": 2.745976209640503,
-      "reward_std": 0.29341720789670944,
-      "rewards/accuracy_reward": 0.7708333730697632,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.01964883040636778,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 779.0000305175781,
+      "epoch": 0.143,
+      "grad_norm": 13.757784684112176,
+      "kl": 0.984375,
+      "learning_rate": 9.949403714711526e-07,
+      "loss": 0.2595,
+      "reward": 2.427113175392151,
+      "reward_std": 0.5199109017848969,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.06420639157295227,
+      "rewards/tag_count_reward": 0.8593750298023224,
       "step": 286
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 530.8333435058594,
-      "epoch": 0.287,
-      "grad_norm": 2.5821740187900097,
-      "kl": 0.1962890625,
-      "learning_rate": 9.074865792409381e-07,
-      "loss": 0.0427,
-      "reward": 2.4354602098464966,
-      "reward_std": 0.31608445942401886,
-      "rewards/accuracy_reward": 0.458333358168602,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.015928767155855894,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 830.5208435058594,
+      "epoch": 0.1435,
+      "grad_norm": 24.88573937996162,
+      "kl": 0.892578125,
+      "learning_rate": 9.948222490639075e-07,
+      "loss": 0.3276,
+      "reward": 2.313633441925049,
+      "reward_std": 0.5377289652824402,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.07699155062437057,
+      "rewards/tag_count_reward": 0.8489583730697632,
       "step": 287
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 516.6875,
-      "epoch": 0.288,
-      "grad_norm": 2.4879515053528927,
-      "kl": 0.1796875,
-      "learning_rate": 9.065303395098358e-07,
-      "loss": 0.0885,
-      "reward": 2.7357919216156006,
-      "reward_std": 0.21385888010263443,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.022888831794261932,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 682.3750305175781,
+      "epoch": 0.144,
+      "grad_norm": 108.97734325463617,
+      "kl": 1.734375,
+      "learning_rate": 9.947027716509488e-07,
+      "loss": 0.2837,
+      "reward": 2.294344663619995,
+      "reward_std": 0.6158420443534851,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.07718315534293652,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 288
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 610.2291870117188,
-      "epoch": 0.289,
-      "grad_norm": 10.258023745667739,
-      "kl": 0.24951171875,
-      "learning_rate": 9.055697555690607e-07,
-      "loss": 0.2087,
-      "reward": 2.447946548461914,
-      "reward_std": 0.3903558999300003,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.0399008896201849,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 729.6875,
+      "epoch": 0.1445,
+      "grad_norm": 20.647159290728784,
+      "kl": 0.93359375,
+      "learning_rate": 9.94581939596225e-07,
+      "loss": 0.1069,
+      "reward": 2.5923370122909546,
+      "reward_std": 0.4817170798778534,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.058704692870378494,
+      "rewards/tag_count_reward": 0.8802083432674408,
       "step": 289
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 488.2500305175781,
-      "epoch": 0.29,
-      "grad_norm": 2.72857938643871,
-      "kl": 0.2431640625,
-      "learning_rate": 9.046048391230247e-07,
-      "loss": -0.0029,
-      "reward": 2.764933466911316,
-      "reward_std": 0.31763769686222076,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.019788892939686775,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 632.1666870117188,
+      "epoch": 0.145,
+      "grad_norm": 23.007824810794826,
+      "kl": 0.37890625,
+      "learning_rate": 9.944597532678119e-07,
+      "loss": 0.1326,
+      "reward": 2.522615909576416,
+      "reward_std": 0.5047437995672226,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.07460657134652138,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 290
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 548.4166717529297,
-      "epoch": 0.291,
-      "grad_norm": 13.446114403763556,
-      "kl": 0.3310546875,
-      "learning_rate": 9.036356019289309e-07,
-      "loss": 0.1304,
-      "reward": 2.442963719367981,
-      "reward_std": 0.4253944456577301,
-      "rewards/accuracy_reward": 0.4791666865348816,
+      "completion_length": 613.8125,
+      "epoch": 0.1455,
+      "grad_norm": 17.27026907734221,
+      "kl": 0.34375,
+      "learning_rate": 9.943362130379101e-07,
+      "loss": 0.1261,
+      "reward": 2.5630831718444824,
+      "reward_std": 0.35013893246650696,
+      "rewards/accuracy_reward": 0.6666666865348816,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.020578143652528524,
-      "rewards/tag_count_reward": 0.984375,
+      "rewards/repetition_penalty_reward": -0.061916783452034,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 291
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.5000305175781,
-      "epoch": 0.292,
-      "grad_norm": 2.399625776751827,
-      "kl": 0.275390625,
-      "learning_rate": 9.026620557966279e-07,
-      "loss": 0.03,
-      "reward": 2.5269211530685425,
-      "reward_std": 0.2096565067768097,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.01474552322179079,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 596.3125152587891,
+      "epoch": 0.146,
+      "grad_norm": 24.305363731366423,
+      "kl": 0.4072265625,
+      "learning_rate": 9.942113192828444e-07,
+      "loss": 0.0284,
+      "reward": 2.6639617681503296,
+      "reward_std": 0.3908410295844078,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.05826057866215706,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 292
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 557.6458435058594,
-      "epoch": 0.293,
-      "grad_norm": 2.6309659598581225,
-      "kl": 0.28955078125,
-      "learning_rate": 9.016842125884684e-07,
-      "loss": 0.0651,
-      "reward": 2.6259918212890625,
-      "reward_std": 0.3636874854564667,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.018105541355907917,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 454.125,
+      "epoch": 0.1465,
+      "grad_norm": 7.404474348588299,
+      "kl": 0.22705078125,
+      "learning_rate": 9.940850723830632e-07,
+      "loss": 0.0509,
+      "reward": 2.862213373184204,
+      "reward_std": 0.18731126189231873,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.0613978561013937,
+      "rewards/tag_count_reward": 1.0,
       "step": 293
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 591.3958435058594,
-      "epoch": 0.294,
-      "grad_norm": 2.594070997434929,
-      "kl": 0.2685546875,
-      "learning_rate": 9.007020842191634e-07,
-      "loss": 0.0742,
-      "reward": 2.6963618993759155,
-      "reward_std": 0.27481937408447266,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.0258603785187006,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 621.3125305175781,
+      "epoch": 0.147,
+      "grad_norm": 18.874856559368098,
+      "kl": 0.2001953125,
+      "learning_rate": 9.939574727231362e-07,
+      "loss": 0.0696,
+      "reward": 2.7581335306167603,
+      "reward_std": 0.3472418487071991,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.06999156251549721,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 294
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 503.7916717529297,
-      "epoch": 0.295,
-      "grad_norm": 2.797055325097868,
-      "kl": 0.3544921875,
-      "learning_rate": 8.997156826556369e-07,
-      "loss": 0.0827,
-      "reward": 2.522792100906372,
-      "reward_std": 0.4541844576597214,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.01887462381273508,
+      "completion_length": 549.125,
+      "epoch": 0.1475,
+      "grad_norm": 3.2744531346829038,
+      "kl": 0.14990234375,
+      "learning_rate": 9.93828520691754e-07,
+      "loss": 0.0335,
+      "reward": 2.661821484565735,
+      "reward_std": 0.22525277733802795,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.08123417943716049,
       "rewards/tag_count_reward": 1.0,
       "step": 295
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 572.1875305175781,
-      "epoch": 0.296,
-      "grad_norm": 3.656654339941009,
-      "kl": 0.388671875,
-      "learning_rate": 8.987250199168808e-07,
-      "loss": 0.0611,
-      "reward": 2.7958298921585083,
-      "reward_std": 0.3831692487001419,
-      "rewards/accuracy_reward": 0.8541666865348816,
+      "completion_length": 630.3958435058594,
+      "epoch": 0.148,
+      "grad_norm": 3.3512313017643782,
+      "kl": 0.212890625,
+      "learning_rate": 9.93698216681727e-07,
+      "loss": 0.0249,
+      "reward": 2.5591964721679688,
+      "reward_std": 0.309579461812973,
+      "rewards/accuracy_reward": 0.6458333730697632,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03750366624444723,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.086636982858181,
+      "rewards/tag_count_reward": 1.0,
       "step": 296
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 679.375,
-      "epoch": 0.297,
-      "grad_norm": 5.294668622093825,
-      "kl": 0.4609375,
-      "learning_rate": 8.977301080738079e-07,
-      "loss": 0.1776,
-      "reward": 2.7792227268218994,
-      "reward_std": 0.2825077772140503,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.04369393363595009,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 570.3958740234375,
+      "epoch": 0.1485,
+      "grad_norm": 19.392543852050114,
+      "kl": 0.19921875,
+      "learning_rate": 9.93566561089984e-07,
+      "loss": 0.1167,
+      "reward": 2.6523066759109497,
+      "reward_std": 0.39907996356487274,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.05429063364863396,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 297
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 777.9375,
-      "epoch": 0.298,
-      "grad_norm": 9.442864310512464,
-      "kl": 0.658203125,
-      "learning_rate": 8.967309592491052e-07,
-      "loss": 0.3089,
-      "reward": 2.554570198059082,
-      "reward_std": 0.5709795355796814,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.04265192709863186,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 632.3125305175781,
+      "epoch": 0.149,
+      "grad_norm": 4.2670800629296775,
+      "kl": 0.22900390625,
+      "learning_rate": 9.934335543175705e-07,
+      "loss": 0.022,
+      "reward": 2.6927382946014404,
+      "reward_std": 0.3433649092912674,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.05899788439273834,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 298
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 774.7708435058594,
-      "epoch": 0.299,
-      "grad_norm": 4.044252023077153,
-      "kl": 1.21484375,
-      "learning_rate": 8.957275856170855e-07,
-      "loss": 0.1666,
-      "reward": 2.4033294916152954,
-      "reward_std": 0.4724307656288147,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.028962312266230583,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 618.9583435058594,
+      "epoch": 0.1495,
+      "grad_norm": 15.195924787873745,
+      "kl": 0.28076171875,
+      "learning_rate": 9.932991967696482e-07,
+      "loss": 0.1031,
+      "reward": 2.660310745239258,
+      "reward_std": 0.16687491163611412,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.041078392416238785,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 299
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 775.9166870117188,
-      "epoch": 0.3,
-      "grad_norm": 13.271553563881938,
-      "kl": 1.8515625,
-      "learning_rate": 8.9471999940354e-07,
-      "loss": 0.2894,
-      "reward": 2.3227447271347046,
-      "reward_std": 0.5490213930606842,
-      "rewards/accuracy_reward": 0.4583333432674408,
+      "completion_length": 555.4375305175781,
+      "epoch": 0.15,
+      "grad_norm": 23.003706484527033,
+      "kl": 0.33984375,
+      "learning_rate": 9.931634888554935e-07,
+      "loss": 0.1633,
+      "reward": 2.595449686050415,
+      "reward_std": 0.35070881247520447,
+      "rewards/accuracy_reward": 0.6666666865348816,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03663041442632675,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "rewards/repetition_penalty_reward": -0.039967115968465805,
+      "rewards/tag_count_reward": 0.96875,
       "step": 300
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 862.0625305175781,
-      "epoch": 0.301,
-      "grad_norm": 11.329717412687812,
-      "kl": 2.6328125,
-      "learning_rate": 8.937082128855891e-07,
-      "loss": 0.303,
-      "reward": 2.497220277786255,
-      "reward_std": 0.5546972751617432,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.025349291041493416,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 591.4791870117188,
+      "epoch": 0.1505,
+      "grad_norm": 29.883708582664426,
+      "kl": 1.10546875,
+      "learning_rate": 9.930264309884964e-07,
+      "loss": 0.1903,
+      "reward": 2.266157388687134,
+      "reward_std": 0.4054761230945587,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.05155107006430626,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 301
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 718.0000305175781,
-      "epoch": 0.302,
-      "grad_norm": 11.620481772542528,
-      "kl": 1.78125,
-      "learning_rate": 8.926922383915315e-07,
-      "loss": 0.1349,
-      "reward": 2.820877194404602,
-      "reward_std": 0.2959998771548271,
-      "rewards/accuracy_reward": 0.8958333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.026345071382820606,
-      "rewards/tag_count_reward": 0.9583333432674408,
+      "completion_length": 559.6875305175781,
+      "epoch": 0.151,
+      "grad_norm": 53.32541543626509,
+      "kl": 1.33203125,
+      "learning_rate": 9.928880235861588e-07,
+      "loss": 0.0845,
+      "reward": 2.680440068244934,
+      "reward_std": 0.240939699113369,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.033101567067205906,
+      "rewards/tag_count_reward": 0.984375,
       "step": 302
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 699.8958740234375,
-      "epoch": 0.303,
-      "grad_norm": 18.84955689848848,
-      "kl": 0.779296875,
-      "learning_rate": 8.916720883006963e-07,
-      "loss": 0.2289,
-      "reward": 2.662257194519043,
-      "reward_std": 0.4918932765722275,
+      "completion_length": 569.5833587646484,
+      "epoch": 0.1515,
+      "grad_norm": 64.08652020758656,
+      "kl": 1.0302734375,
+      "learning_rate": 9.927482670700936e-07,
+      "loss": 0.1684,
+      "reward": 2.6656733751296997,
+      "reward_std": 0.3831760287284851,
       "rewards/accuracy_reward": 0.7500000298023224,
       "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02697897609323263,
-      "rewards/tag_count_reward": 0.9531250298023224,
+      "rewards/repetition_penalty_reward": -0.054812896996736526,
+      "rewards/tag_count_reward": 0.984375,
       "step": 303
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 854.3333435058594,
-      "epoch": 0.304,
-      "grad_norm": 16.06283643508135,
-      "kl": 0.9453125,
-      "learning_rate": 8.906477750432903e-07,
-      "loss": 0.2394,
-      "reward": 2.463517665863037,
-      "reward_std": 0.5290980041027069,
-      "rewards/accuracy_reward": 0.6041666865348816,
+      "completion_length": 505.60418701171875,
+      "epoch": 0.152,
+      "grad_norm": 4.124901631085503,
+      "kl": 0.162109375,
+      "learning_rate": 9.926071618660237e-07,
+      "loss": -0.017,
+      "reward": 2.7977336645126343,
+      "reward_std": 0.2716551870107651,
+      "rewards/accuracy_reward": 0.8541666865348816,
       "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.034746479243040085,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "rewards/repetition_penalty_reward": -0.04948872700333595,
+      "rewards/tag_count_reward": 1.0,
       "step": 304
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 795.7291870117188,
-      "epoch": 0.305,
-      "grad_norm": 7.368811609366542,
-      "kl": 1.072265625,
-      "learning_rate": 8.896193111002475e-07,
-      "loss": 0.1781,
-      "reward": 2.7576704025268555,
-      "reward_std": 0.3518691807985306,
-      "rewards/accuracy_reward": 0.8958333730697632,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03399639390408993,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 535.5000305175781,
+      "epoch": 0.1525,
+      "grad_norm": 2.3959309961624955,
+      "kl": 0.169921875,
+      "learning_rate": 9.924647084037797e-07,
+      "loss": 0.028,
+      "reward": 2.503369092941284,
+      "reward_std": 0.22217638790607452,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.057394931092858315,
+      "rewards/tag_count_reward": 0.984375,
       "step": 305
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 646.9583435058594,
-      "epoch": 0.306,
-      "grad_norm": 5.727482347866686,
-      "kl": 1.28125,
-      "learning_rate": 8.88586709003076e-07,
-      "loss": 0.1126,
-      "reward": 2.540480852127075,
-      "reward_std": 0.4421972632408142,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03243602532893419,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 533.0208435058594,
+      "epoch": 0.153,
+      "grad_norm": 2.4185306667509985,
+      "kl": 0.149658203125,
+      "learning_rate": 9.923209071172994e-07,
+      "loss": 0.0305,
+      "reward": 2.573015809059143,
+      "reward_std": 0.3675818666815758,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06934532895684242,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 306
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 876.6041870117188,
-      "epoch": 0.307,
-      "grad_norm": 6.1244435821869905,
-      "kl": 1.9296875,
-      "learning_rate": 8.875499813337067e-07,
-      "loss": 0.2147,
-      "reward": 2.5565717220306396,
-      "reward_std": 0.5118266344070435,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026761652901768684,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 533.2500305175781,
+      "epoch": 0.1535,
+      "grad_norm": 3.4451894822132405,
+      "kl": 0.146484375,
+      "learning_rate": 9.921757584446268e-07,
+      "loss": 0.038,
+      "reward": 2.809865355491638,
+      "reward_std": 0.27805351465940475,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.07902374118566513,
+      "rewards/tag_count_reward": 1.0,
       "step": 307
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 735.1041870117188,
-      "epoch": 0.308,
-      "grad_norm": 6.2567586393756685,
-      "kl": 1.25390625,
-      "learning_rate": 8.865091407243394e-07,
-      "loss": 0.1521,
-      "reward": 2.8010218143463135,
-      "reward_std": 0.3402135968208313,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.03057546727359295,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 626.2708740234375,
+      "epoch": 0.154,
+      "grad_norm": 4.282437820486171,
+      "kl": 0.18994140625,
+      "learning_rate": 9.9202926282791e-07,
+      "loss": 0.0255,
+      "reward": 2.4780622720718384,
+      "reward_std": 0.4932084083557129,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.051451750099658966,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 308
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 884.6666870117188,
-      "epoch": 0.309,
-      "grad_norm": 11.171414148004054,
-      "kl": 2.3984375,
-      "learning_rate": 8.85464199857288e-07,
-      "loss": 0.2805,
-      "reward": 2.442511558532715,
-      "reward_std": 0.5785179138183594,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026238556019961834,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 544.6875305175781,
+      "epoch": 0.1545,
+      "grad_norm": 3.8575808680226182,
+      "kl": 0.16064453125,
+      "learning_rate": 9.918814207133997e-07,
+      "loss": -0.0068,
+      "reward": 2.1725984811782837,
+      "reward_std": 0.3830345869064331,
+      "rewards/accuracy_reward": 0.2916666716337204,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.08434605225920677,
+      "rewards/tag_count_reward": 1.0,
       "step": 309
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 982.625,
-      "epoch": 0.31,
-      "grad_norm": 5.999562164727096,
-      "kl": 2.0234375,
-      "learning_rate": 8.844151714648274e-07,
-      "loss": 0.2776,
-      "reward": 2.3847498893737793,
-      "reward_std": 0.6050747036933899,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026708428747951984,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 556.2083740234375,
+      "epoch": 0.155,
+      "grad_norm": 2.30299380218455,
+      "kl": 0.1865234375,
+      "learning_rate": 9.917322325514487e-07,
+      "loss": 0.0359,
+      "reward": 2.8288975954055786,
+      "reward_std": 0.22805538028478622,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.05999153107404709,
+      "rewards/tag_count_reward": 1.0,
       "step": 310
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 750.9166870117188,
-      "epoch": 0.311,
-      "grad_norm": 10.25459242080789,
-      "kl": 1.228515625,
-      "learning_rate": 8.833620683290375e-07,
-      "loss": 0.2226,
-      "reward": 2.489295244216919,
-      "reward_std": 0.3828739821910858,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.031538188457489014,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 571.8333435058594,
+      "epoch": 0.1555,
+      "grad_norm": 13.544980888404456,
+      "kl": 0.1787109375,
+      "learning_rate": 9.915816987965102e-07,
+      "loss": 0.1167,
+      "reward": 2.6313341856002808,
+      "reward_std": 0.3306514471769333,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.06137407571077347,
+      "rewards/tag_count_reward": 0.984375,
       "step": 311
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 920.7083435058594,
-      "epoch": 0.312,
-      "grad_norm": 11.341166194035626,
-      "kl": 3.0390625,
-      "learning_rate": 8.823049032816478e-07,
-      "loss": 0.3465,
-      "reward": 2.3931429386138916,
-      "reward_std": 0.5464552640914917,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.033940425142645836,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 601.6041870117188,
+      "epoch": 0.156,
+      "grad_norm": 7.684331357658625,
+      "kl": 0.29296875,
+      "learning_rate": 9.91429819907136e-07,
+      "loss": 0.1471,
+      "reward": 2.8133013248443604,
+      "reward_std": 0.28423790633678436,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.06690708734095097,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 312
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 824.3541870117188,
-      "epoch": 0.313,
-      "grad_norm": 14.499216833551465,
-      "kl": 2.5703125,
-      "learning_rate": 8.812436892038805e-07,
-      "loss": 0.4531,
-      "reward": 2.4514354467391968,
-      "reward_std": 0.4615107327699661,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.03120344504714012,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 568.6666870117188,
+      "epoch": 0.1565,
+      "grad_norm": 16.451352502057013,
+      "kl": 0.5234375,
+      "learning_rate": 9.912765963459756e-07,
+      "loss": -0.0008,
+      "reward": 2.7371630668640137,
+      "reward_std": 0.3937046229839325,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.05103146657347679,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 313
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 752.2500305175781,
-      "epoch": 0.314,
-      "grad_norm": 9.160833779249737,
-      "kl": 2.52734375,
-      "learning_rate": 8.801784390262943e-07,
-      "loss": 0.4578,
-      "reward": 2.459816098213196,
-      "reward_std": 0.5396545231342316,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02629512920975685,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 538.75,
+      "epoch": 0.157,
+      "grad_norm": 5.827769683111703,
+      "kl": 0.4130859375,
+      "learning_rate": 9.911220285797748e-07,
+      "loss": 0.0415,
+      "reward": 2.58489727973938,
+      "reward_std": 0.38202086091041565,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.06614442355930805,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 314
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 666.5625305175781,
-      "epoch": 0.315,
-      "grad_norm": 16.821527260355506,
-      "kl": 1.90625,
-      "learning_rate": 8.791091657286267e-07,
-      "loss": 0.1424,
-      "reward": 2.7689753770828247,
-      "reward_std": 0.27851930260658264,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.03137190733104944,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 527.9166870117188,
+      "epoch": 0.1575,
+      "grad_norm": 7.880587052881426,
+      "kl": 0.20703125,
+      "learning_rate": 9.909661170793733e-07,
+      "loss": 0.1088,
+      "reward": 2.4540834426879883,
+      "reward_std": 0.39223696291446686,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.059805694967508316,
+      "rewards/tag_count_reward": 1.0,
       "step": 315
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 749.3958435058594,
-      "epoch": 0.316,
-      "grad_norm": 15.034225144072645,
-      "kl": 2.33984375,
-      "learning_rate": 8.780358823396352e-07,
-      "loss": 0.2368,
-      "reward": 2.8358311653137207,
-      "reward_std": 0.2934058606624603,
-      "rewards/accuracy_reward": 0.9375,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03396068140864372,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 550.7291717529297,
+      "epoch": 0.158,
+      "grad_norm": 6.6106566025225515,
+      "kl": 0.314453125,
+      "learning_rate": 9.908088623197048e-07,
+      "loss": 0.013,
+      "reward": 2.4705549478530884,
+      "reward_std": 0.46846726536750793,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.058959029614925385,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 316
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 740.875,
-      "epoch": 0.317,
-      "grad_norm": 8.197253327861047,
-      "kl": 2.1796875,
-      "learning_rate": 8.769586019369391e-07,
-      "loss": 0.2262,
-      "reward": 2.721475839614868,
-      "reward_std": 0.3922436535358429,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.023315943777561188,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 567.0208435058594,
+      "epoch": 0.1585,
+      "grad_norm": 2.7832594906032084,
+      "kl": 0.24462890625,
+      "learning_rate": 9.906502647797945e-07,
+      "loss": -0.0046,
+      "reward": 2.705116033554077,
+      "reward_std": 0.33963292837142944,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.058772869408130646,
+      "rewards/tag_count_reward": 1.0,
       "step": 317
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 731.8125,
-      "epoch": 0.318,
-      "grad_norm": 17.02162446986935,
-      "kl": 0.9296875,
-      "learning_rate": 8.758773376468604e-07,
-      "loss": 0.2494,
-      "reward": 2.5680298805236816,
-      "reward_std": 0.4028843492269516,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.030928438529372215,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 566.4583435058594,
+      "epoch": 0.159,
+      "grad_norm": 6.607393683777124,
+      "kl": 0.19873046875,
+      "learning_rate": 9.904903249427582e-07,
+      "loss": 0.0072,
+      "reward": 2.756907343864441,
+      "reward_std": 0.3417035788297653,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.074690081179142,
+      "rewards/tag_count_reward": 0.984375,
       "step": 318
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 646.0625305175781,
-      "epoch": 0.319,
-      "grad_norm": 8.403943914356434,
-      "kl": 0.6796875,
-      "learning_rate": 8.747921026442629e-07,
-      "loss": 0.1368,
-      "reward": 2.6740156412124634,
-      "reward_std": 0.3067634850740433,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.018692771438509226,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 525.0833587646484,
+      "epoch": 0.1595,
+      "grad_norm": 12.632626492261867,
+      "kl": 0.22802734375,
+      "learning_rate": 9.903290432958003e-07,
+      "loss": 0.0901,
+      "reward": 2.739717960357666,
+      "reward_std": 0.4267748296260834,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03979583643376827,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 319
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 612.8125,
-      "epoch": 0.32,
-      "grad_norm": 3.055603498237129,
-      "kl": 0.384765625,
-      "learning_rate": 8.737029101523929e-07,
-      "loss": 0.0287,
-      "reward": 2.7557495832443237,
-      "reward_std": 0.21633769571781158,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.028972744941711426,
+      "completion_length": 434.00001525878906,
+      "epoch": 0.16,
+      "grad_norm": 3.2432048734417296,
+      "kl": 0.169921875,
+      "learning_rate": 9.901664203302124e-07,
+      "loss": 0.0588,
+      "reward": 2.945529341697693,
+      "reward_std": 0.02130332589149475,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.054470714181661606,
       "rewards/tag_count_reward": 1.0,
       "step": 320
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 619.375,
-      "epoch": 0.321,
-      "grad_norm": 3.8882839855022113,
-      "kl": 0.3203125,
-      "learning_rate": 8.726097734427172e-07,
-      "loss": 0.0372,
-      "reward": 2.877520799636841,
-      "reward_std": 0.2280464619398117,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.028729302808642387,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 511.7708435058594,
+      "epoch": 0.1605,
+      "grad_norm": 13.83323285922156,
+      "kl": 0.26025390625,
+      "learning_rate": 9.900024565413727e-07,
+      "loss": 0.075,
+      "reward": 2.590381622314453,
+      "reward_std": 0.46982041001319885,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.06760460883378983,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 321
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 642.2291870117188,
-      "epoch": 0.322,
-      "grad_norm": 5.934518493796481,
-      "kl": 0.5390625,
-      "learning_rate": 8.715127058347614e-07,
-      "loss": 0.1573,
-      "reward": 2.8018211126327515,
-      "reward_std": 0.2535254070535302,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.029776128008961678,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 568.3958435058594,
+      "epoch": 0.161,
+      "grad_norm": 15.041988051632119,
+      "kl": 0.603515625,
+      "learning_rate": 9.89837152428743e-07,
+      "loss": 0.0217,
+      "reward": 2.7006497383117676,
+      "reward_std": 0.3142661973834038,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.04761423170566559,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 322
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 584.3958435058594,
-      "epoch": 0.323,
-      "grad_norm": 12.30823020424776,
-      "kl": 0.6533203125,
-      "learning_rate": 8.704117206959484e-07,
-      "loss": -0.0165,
-      "reward": 2.274023652076721,
-      "reward_std": 0.3736593574285507,
-      "rewards/accuracy_reward": 0.3125000111758709,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.022851460613310337,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 498.0833435058594,
+      "epoch": 0.1615,
+      "grad_norm": 8.377756641490475,
+      "kl": 0.564453125,
+      "learning_rate": 9.896705084958687e-07,
+      "loss": 0.0812,
+      "reward": 2.581056237220764,
+      "reward_std": 0.424957811832428,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04741601645946503,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 323
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 616.7083435058594,
-      "epoch": 0.324,
-      "grad_norm": 11.03962394142098,
+      "completion_length": 561.6875,
+      "epoch": 0.162,
+      "grad_norm": 24.710240478593196,
       "kl": 0.708984375,
-      "learning_rate": 8.693068314414344e-07,
-      "loss": 0.2153,
-      "reward": 2.9491621255874634,
-      "reward_std": 0.0671940129250288,
-      "rewards/accuracy_reward": 1.0,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.030004790984094143,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "learning_rate": 9.895025252503755e-07,
+      "loss": 0.2563,
+      "reward": 2.6197162866592407,
+      "reward_std": 0.4363028407096863,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.05736706592142582,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 324
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 586.2916870117188,
-      "epoch": 0.325,
-      "grad_norm": 21.7519973984484,
-      "kl": 0.9091796875,
-      "learning_rate": 8.681980515339463e-07,
-      "loss": 0.1511,
-      "reward": 2.7508959770202637,
-      "reward_std": 0.3181898444890976,
-      "rewards/accuracy_reward": 0.7916666865348816,
+      "completion_length": 469.5625,
+      "epoch": 0.1625,
+      "grad_norm": 9.070483774061584,
+      "kl": 0.58203125,
+      "learning_rate": 9.8933320320397e-07,
+      "loss": 0.0186,
+      "reward": 2.583137035369873,
+      "reward_std": 0.3166651949286461,
+      "rewards/accuracy_reward": 0.6458333432674408,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.025145714171230793,
+      "rewards/repetition_penalty_reward": -0.04707140102982521,
       "rewards/tag_count_reward": 0.9843750298023224,
       "step": 325
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 609.3125305175781,
-      "epoch": 0.326,
-      "grad_norm": 7.475624041841047,
-      "kl": 1.08984375,
-      "learning_rate": 8.670853944836176e-07,
-      "loss": 0.1854,
-      "reward": 2.718400478363037,
-      "reward_std": 0.3192872703075409,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.024655278772115707,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 480.0416717529297,
+      "epoch": 0.163,
+      "grad_norm": 4.3751653300427655,
+      "kl": 0.4638671875,
+      "learning_rate": 9.891625428724364e-07,
+      "loss": 0.0585,
+      "reward": 2.6971195936203003,
+      "reward_std": 0.33363979309797287,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.04072774015367031,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 326
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 537.0833435058594,
-      "epoch": 0.327,
-      "grad_norm": 3.934696523513004,
-      "kl": 0.4482421875,
-      "learning_rate": 8.659688738478231e-07,
-      "loss": 0.0237,
-      "reward": 2.5368727445602417,
-      "reward_std": 0.2816409021615982,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.029099617153406143,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 469.9791717529297,
+      "epoch": 0.1635,
+      "grad_norm": 8.17049817832142,
+      "kl": 0.4169921875,
+      "learning_rate": 9.889905447756355e-07,
+      "loss": -0.0211,
+      "reward": 2.66062331199646,
+      "reward_std": 0.09492377191781998,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.05639079958200455,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 327
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 561.1875,
-      "epoch": 0.328,
-      "grad_norm": 4.769237705857611,
-      "kl": 0.576171875,
-      "learning_rate": 8.648485032310144e-07,
-      "loss": 0.0438,
-      "reward": 2.7067657709121704,
-      "reward_std": 0.287298321723938,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.029345519840717316,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 553.0833435058594,
+      "epoch": 0.164,
+      "grad_norm": 11.036171582992713,
+      "kl": 0.36328125,
+      "learning_rate": 9.888172094375033e-07,
+      "loss": -0.0311,
+      "reward": 2.6875771284103394,
+      "reward_std": 0.2717094421386719,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.053742485120892525,
+      "rewards/tag_count_reward": 0.984375,
       "step": 328
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 630.4375,
-      "epoch": 0.329,
-      "grad_norm": 10.245397889363776,
-      "kl": 2.02734375,
-      "learning_rate": 8.63724296284554e-07,
-      "loss": 0.2262,
-      "reward": 2.599774479866028,
-      "reward_std": 0.4478468745946884,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.023489387705922127,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 475.93751525878906,
+      "epoch": 0.1645,
+      "grad_norm": 2.6290357341375863,
+      "kl": 0.166015625,
+      "learning_rate": 9.886425373860496e-07,
+      "loss": -0.0091,
+      "reward": 2.445455312728882,
+      "reward_std": 0.3463872969150543,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.04239194467663765,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 329
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 571.2916870117188,
-      "epoch": 0.33,
-      "grad_norm": 22.88975372356627,
-      "kl": 1.796875,
-      "learning_rate": 8.625962667065487e-07,
-      "loss": 0.206,
-      "reward": 2.5127453804016113,
-      "reward_std": 0.4001367390155792,
-      "rewards/accuracy_reward": 0.5625000149011612,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.028921468183398247,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 549.9791870117188,
+      "epoch": 0.165,
+      "grad_norm": 2.8947233993694232,
+      "kl": 0.15380859375,
+      "learning_rate": 9.88466529153356e-07,
+      "loss": 0.0504,
+      "reward": 2.676490306854248,
+      "reward_std": 0.26038385927677155,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.07350974902510643,
+      "rewards/tag_count_reward": 1.0,
       "step": 330
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 524.7083740234375,
-      "epoch": 0.331,
-      "grad_norm": 5.9516087463505025,
-      "kl": 0.42626953125,
-      "learning_rate": 8.614644282416831e-07,
-      "loss": 0.0881,
-      "reward": 2.731919050216675,
-      "reward_std": 0.31574640423059464,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.040650567039847374,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 442.12501525878906,
+      "epoch": 0.1655,
+      "grad_norm": 2.304611040563405,
+      "kl": 0.1357421875,
+      "learning_rate": 9.882891852755732e-07,
+      "loss": 0.0241,
+      "reward": 2.857384443283081,
+      "reward_std": 0.18227218464016914,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06622675433754921,
+      "rewards/tag_count_reward": 1.0,
       "step": 331
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.9583740234375,
-      "epoch": 0.332,
-      "grad_norm": 5.083242340459772,
-      "kl": 0.27294921875,
-      "learning_rate": 8.603287946810513e-07,
-      "loss": 0.05,
-      "reward": 2.7102267742156982,
-      "reward_std": 0.2585765942931175,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03456501290202141,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 457.7708435058594,
+      "epoch": 0.166,
+      "grad_norm": 2.2368549675600944,
+      "kl": 0.13671875,
+      "learning_rate": 9.881105062929221e-07,
+      "loss": -0.0458,
+      "reward": 2.5673773288726807,
+      "reward_std": 0.4164246767759323,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.040261661633849144,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 332
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 618.1041870117188,
-      "epoch": 0.333,
-      "grad_norm": 6.1273870030040065,
-      "kl": 1.0595703125,
-      "learning_rate": 8.591893798619903e-07,
-      "loss": 0.2075,
-      "reward": 2.7363253831863403,
-      "reward_std": 0.37927868962287903,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.018883120268583298,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 583.4375152587891,
+      "epoch": 0.1665,
+      "grad_norm": 3.020419228903663,
+      "kl": 0.12939453125,
+      "learning_rate": 9.879304927496896e-07,
+      "loss": 0.1167,
+      "reward": 2.7053964138031006,
+      "reward_std": 0.3210095912218094,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06196486949920654,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 333
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 603.9375,
-      "epoch": 0.334,
-      "grad_norm": 5.897033604873569,
-      "kl": 1.12890625,
-      "learning_rate": 8.580461976679099e-07,
-      "loss": 0.2075,
-      "reward": 2.6124191284179688,
-      "reward_std": 0.4683973491191864,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.017789172008633614,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 529.2708587646484,
+      "epoch": 0.167,
+      "grad_norm": 2.375485512826028,
+      "kl": 0.142578125,
+      "learning_rate": 9.877491451942284e-07,
+      "loss": 0.0589,
+      "reward": 2.5137102603912354,
+      "reward_std": 0.3925721198320389,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.05747038125991821,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 334
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 515.75,
-      "epoch": 0.335,
-      "grad_norm": 7.88708659023299,
-      "kl": 0.53515625,
-      "learning_rate": 8.568992620281243e-07,
-      "loss": 0.0447,
-      "reward": 2.5799560546875,
-      "reward_std": 0.4440341293811798,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.019002487882971764,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 510.5,
+      "epoch": 0.1675,
+      "grad_norm": 4.018502569391315,
+      "kl": 0.146484375,
+      "learning_rate": 9.875664641789543e-07,
+      "loss": 0.1097,
+      "reward": 2.4702670574188232,
+      "reward_std": 0.21461456269025803,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.050566382706165314,
+      "rewards/tag_count_reward": 1.0,
       "step": 335
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 618.2291870117188,
-      "epoch": 0.336,
-      "grad_norm": 13.855588768970593,
-      "kl": 1.8125,
-      "learning_rate": 8.557485869176825e-07,
-      "loss": 0.1842,
-      "reward": 2.804866313934326,
-      "reward_std": 0.27131783962249756,
-      "rewards/accuracy_reward": 0.8958333730697632,
+      "completion_length": 563.6875,
+      "epoch": 0.168,
+      "grad_norm": 2.9361596954162597,
+      "kl": 0.14208984375,
+      "learning_rate": 9.873824502603459e-07,
+      "loss": -0.0059,
+      "reward": 2.709442138671875,
+      "reward_std": 0.3301195055246353,
+      "rewards/accuracy_reward": 0.7708333730697632,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.023258699104189873,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "rewards/repetition_penalty_reward": -0.06139139086008072,
+      "rewards/tag_count_reward": 1.0,
       "step": 336
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 544.0416870117188,
-      "epoch": 0.337,
-      "grad_norm": 10.22565696968507,
-      "kl": 0.556640625,
-      "learning_rate": 8.545941863571973e-07,
-      "loss": 0.0697,
-      "reward": 2.7841432094573975,
-      "reward_std": 0.22680224478244781,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.028356771916151047,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 590.7916870117188,
+      "epoch": 0.1685,
+      "grad_norm": 4.349944868969375,
+      "kl": 0.1650390625,
+      "learning_rate": 9.871971039989407e-07,
+      "loss": 0.0989,
+      "reward": 2.6142622232437134,
+      "reward_std": 0.44250747561454773,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.07844629138708115,
+      "rewards/tag_count_reward": 0.984375,
       "step": 337
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 660.1875305175781,
-      "epoch": 0.338,
-      "grad_norm": 13.275393254857,
-      "kl": 1.57421875,
-      "learning_rate": 8.534360744126753e-07,
-      "loss": 0.3522,
-      "reward": 2.626600503921509,
-      "reward_std": 0.5416260808706284,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.026177333667874336,
-      "rewards/tag_count_reward": 0.9375000298023224,
+      "completion_length": 454.06251525878906,
+      "epoch": 0.169,
+      "grad_norm": 2.19737492930891,
+      "kl": 0.16650390625,
+      "learning_rate": 9.870104259593362e-07,
+      "loss": 0.0282,
+      "reward": 2.822553515434265,
+      "reward_std": 0.14519703015685081,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.04550204798579216,
+      "rewards/tag_count_reward": 1.0,
       "step": 338
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 538.2916870117188,
-      "epoch": 0.339,
-      "grad_norm": 9.284294976639861,
-      "kl": 0.890625,
-      "learning_rate": 8.522742651953456e-07,
-      "loss": 0.2327,
-      "reward": 2.6350467205047607,
-      "reward_std": 0.30000850558280945,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.024675646796822548,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 521.3333435058594,
+      "epoch": 0.1695,
+      "grad_norm": 2.595702991755517,
+      "kl": 0.21337890625,
+      "learning_rate": 9.86822416710186e-07,
+      "loss": -0.0125,
+      "reward": 2.728282928466797,
+      "reward_std": 0.35869090259075165,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.04255035147070885,
+      "rewards/tag_count_reward": 1.0,
       "step": 339
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 572.5208740234375,
-      "epoch": 0.34,
-      "grad_norm": 11.869158855207635,
-      "kl": 1.427734375,
-      "learning_rate": 8.511087728614862e-07,
-      "loss": 0.2899,
-      "reward": 2.693272352218628,
-      "reward_std": 0.4559956192970276,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.023741761222481728,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 487.3333435058594,
+      "epoch": 0.17,
+      "grad_norm": 2.150403718440524,
+      "kl": 0.18505859375,
+      "learning_rate": 9.866330768241983e-07,
+      "loss": 0.0148,
+      "reward": 2.5233949422836304,
+      "reward_std": 0.3504791557788849,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.06688292883336544,
+      "rewards/tag_count_reward": 1.0,
       "step": 340
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 559.7083435058594,
-      "epoch": 0.341,
-      "grad_norm": 16.084642786041737,
-      "kl": 2.78125,
-      "learning_rate": 8.499396116122535e-07,
-      "loss": 0.404,
-      "reward": 2.5177054405212402,
-      "reward_std": 0.358899861574173,
-      "rewards/accuracy_reward": 0.583333358168602,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.018752962350845337,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 499.25,
+      "epoch": 0.1705,
+      "grad_norm": 2.1347353681012384,
+      "kl": 0.1826171875,
+      "learning_rate": 9.86442406878136e-07,
+      "loss": 0.0095,
+      "reward": 2.8164087533950806,
+      "reward_std": 0.22795867174863815,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.06032741814851761,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 341
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 552.3125305175781,
-      "epoch": 0.342,
-      "grad_norm": 22.00620591226436,
-      "kl": 2.71875,
-      "learning_rate": 8.487667956935087e-07,
-      "loss": 0.1789,
-      "reward": 2.8170549869537354,
-      "reward_std": 0.35434219241142273,
-      "rewards/accuracy_reward": 0.8958333432674408,
+      "completion_length": 545.9791870117188,
+      "epoch": 0.171,
+      "grad_norm": 2.0936215044851685,
+      "kl": 0.14453125,
+      "learning_rate": 9.862504074528126e-07,
+      "loss": -0.0212,
+      "reward": 2.819482445716858,
+      "reward_std": 0.2626145929098129,
+      "rewards/accuracy_reward": 0.8958333730697632,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026695125736296177,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "rewards/repetition_penalty_reward": -0.07635113224387169,
+      "rewards/tag_count_reward": 1.0,
       "step": 342
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 522.9791717529297,
-      "epoch": 0.343,
-      "grad_norm": 22.634127347183696,
-      "kl": 3.59375,
-      "learning_rate": 8.475903393956433e-07,
-      "loss": 0.3774,
-      "reward": 2.7711684703826904,
-      "reward_std": 0.3556895852088928,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.018762326799333096,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 541.4375,
+      "epoch": 0.1715,
+      "grad_norm": 2.104187790019279,
+      "kl": 0.15478515625,
+      "learning_rate": 9.860570791330911e-07,
+      "loss": 0.0196,
+      "reward": 2.7721027135849,
+      "reward_std": 0.3143990561366081,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.07164733856916428,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 343
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 594.7291870117188,
-      "epoch": 0.344,
-      "grad_norm": 25.37619859311342,
-      "kl": 2.046875,
-      "learning_rate": 8.464102570534061e-07,
-      "loss": 0.0985,
-      "reward": 2.51139497756958,
-      "reward_std": 0.4084962010383606,
-      "rewards/accuracy_reward": 0.583333358168602,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.03374398034065962,
-      "rewards/tag_count_reward": 0.9687500298023224,
+      "completion_length": 584.7291870117188,
+      "epoch": 0.172,
+      "grad_norm": 3.299557850274513,
+      "kl": 0.16455078125,
+      "learning_rate": 9.85862422507884e-07,
+      "loss": 0.1283,
+      "reward": 2.278268575668335,
+      "reward_std": 0.44967466592788696,
+      "rewards/accuracy_reward": 0.395833358168602,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.04812048375606537,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 344
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 584.375,
-      "epoch": 0.345,
-      "grad_norm": 8.691345932821196,
-      "kl": 2.03515625,
-      "learning_rate": 8.452265630457282e-07,
-      "loss": 0.2564,
-      "reward": 2.603990077972412,
-      "reward_std": 0.5245843231678009,
+      "completion_length": 579.9583435058594,
+      "epoch": 0.1725,
+      "grad_norm": 4.038057650027608,
+      "kl": 0.16552734375,
+      "learning_rate": 9.856664381701483e-07,
+      "loss": 0.0722,
+      "reward": 2.5613759756088257,
+      "reward_std": 0.376383513212204,
       "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02448232937604189,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.06188800558447838,
+      "rewards/tag_count_reward": 0.984375,
       "step": 345
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 533.9166870117188,
-      "epoch": 0.346,
-      "grad_norm": 27.37252896321387,
-      "kl": 1.34375,
-      "learning_rate": 8.440392717955475e-07,
-      "loss": 0.4947,
-      "reward": 2.6430338621139526,
-      "reward_std": 0.4521474093198776,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.014952600467950106,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 485.10418701171875,
+      "epoch": 0.173,
+      "grad_norm": 2.3341365558471643,
+      "kl": 0.15625,
+      "learning_rate": 9.854691267168871e-07,
+      "loss": 0.059,
+      "reward": 2.5120105743408203,
+      "reward_std": 0.23116411548107862,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.057433972135186195,
+      "rewards/tag_count_reward": 1.0,
       "step": 346
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 528.8541870117188,
-      "epoch": 0.347,
-      "grad_norm": 3.9349545954124743,
-      "kl": 0.39453125,
-      "learning_rate": 8.428483977696328e-07,
-      "loss": 0.0493,
-      "reward": 2.7178937196731567,
-      "reward_std": 0.36671870201826096,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026898046024143696,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 540.1875305175781,
+      "epoch": 0.1735,
+      "grad_norm": 2.6154155854500614,
+      "kl": 0.1572265625,
+      "learning_rate": 9.852704887491445e-07,
+      "loss": -0.0386,
+      "reward": 2.24581515789032,
+      "reward_std": 0.4455568790435791,
+      "rewards/accuracy_reward": 0.3333333358168602,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.05974029190838337,
+      "rewards/tag_count_reward": 1.0,
       "step": 347
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 536.8333740234375,
-      "epoch": 0.348,
-      "grad_norm": 5.777524675922901,
-      "kl": 0.3681640625,
-      "learning_rate": 8.416539554784089e-07,
-      "loss": 0.049,
-      "reward": 2.5611852407455444,
-      "reward_std": 0.3319101259112358,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.025620201602578163,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 543.5,
+      "epoch": 0.174,
+      "grad_norm": 2.774589039274609,
+      "kl": 0.16796875,
+      "learning_rate": 9.850705248720068e-07,
+      "loss": 0.0403,
+      "reward": 2.727095603942871,
+      "reward_std": 0.4202606528997421,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.047210052609443665,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 348
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 553.625,
-      "epoch": 0.349,
-      "grad_norm": 5.2707839803098055,
-      "kl": 0.34375,
-      "learning_rate": 8.404559594757777e-07,
-      "loss": 0.0007,
-      "reward": 2.545145034790039,
-      "reward_std": 0.22191336005926132,
-      "rewards/accuracy_reward": 0.5833333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.026035414077341557,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 659.2083435058594,
+      "epoch": 0.1745,
+      "grad_norm": 4.335912070161832,
+      "kl": 0.18359375,
+      "learning_rate": 9.848692356945981e-07,
+      "loss": 0.083,
+      "reward": 2.5149093866348267,
+      "reward_std": 0.48283551633358,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.07884062826633453,
+      "rewards/tag_count_reward": 0.96875,
       "step": 349
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 462.12501525878906,
-      "epoch": 0.35,
-      "grad_norm": 3.643489393286096,
-      "kl": 0.2626953125,
-      "learning_rate": 8.392544243589427e-07,
-      "loss": 0.0118,
-      "reward": 2.5923283100128174,
-      "reward_std": 0.3613494336605072,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.022255297750234604,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 424.60418701171875,
+      "epoch": 0.175,
+      "grad_norm": 2.911920115218028,
+      "kl": 0.1474609375,
+      "learning_rate": 9.846666218300807e-07,
+      "loss": 0.0423,
+      "reward": 2.867781400680542,
+      "reward_std": 0.18726971745491028,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.06798264011740685,
+      "rewards/tag_count_reward": 0.984375,
       "step": 350
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 563.3333740234375,
-      "epoch": 0.351,
-      "grad_norm": 7.73557681949032,
-      "kl": 0.56640625,
-      "learning_rate": 8.3804936476823e-07,
-      "loss": 0.1536,
-      "reward": 2.4618531465530396,
-      "reward_std": 0.3520192950963974,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.024257982149720192,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 625.1875,
+      "epoch": 0.1755,
+      "grad_norm": 4.16432723769119,
+      "kl": 0.263671875,
+      "learning_rate": 9.844626838956513e-07,
+      "loss": 0.0903,
+      "reward": 2.526502013206482,
+      "reward_std": 0.4054105877876282,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.055095236748456955,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 351
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 523.5625305175781,
-      "epoch": 0.352,
-      "grad_norm": 3.134584879525316,
-      "kl": 0.16259765625,
-      "learning_rate": 8.368407953869103e-07,
-      "loss": 0.0164,
-      "reward": 2.8710557222366333,
-      "reward_std": 0.16795307025313377,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.03345821052789688,
-      "rewards/tag_count_reward": 0.9947916865348816,
-      "step": 352
-    },
-    {
-      "clip_ratio": 0.0,
-      "completion_length": 526.0208435058594,
-      "epoch": 0.353,
-      "grad_norm": 4.312118426726596,
-      "kl": 0.224609375,
-      "learning_rate": 8.356287309410204e-07,
-      "loss": 0.0462,
-      "reward": 2.8158966302871704,
-      "reward_std": 0.22574415802955627,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.0226451912894845,
+      "completion_length": 596.2083740234375,
+      "epoch": 0.176,
+      "grad_norm": 4.351440295785368,
+      "kl": 0.3583984375,
+      "learning_rate": 9.8425742251254e-07,
+      "loss": 0.0664,
+      "reward": 2.6749093532562256,
+      "reward_std": 0.30670662224292755,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.03863241523504257,
+      "rewards/tag_count_reward": 0.9843750298023224,
+      "step": 352
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 551.6458587646484,
+      "epoch": 0.1765,
+      "grad_norm": 4.117439533311138,
+      "kl": 0.6171875,
+      "learning_rate": 9.84050838306009e-07,
+      "loss": 0.1052,
+      "reward": 2.7304115295410156,
+      "reward_std": 0.3043531756848097,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.0734081044793129,
       "rewards/tag_count_reward": 0.984375,
       "step": 353
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 490.1666717529297,
-      "epoch": 0.354,
-      "grad_norm": 2.993028075064908,
-      "kl": 0.1904296875,
-      "learning_rate": 8.344131861991828e-07,
-      "loss": -0.0266,
-      "reward": 2.8288527727127075,
-      "reward_std": 0.15067671798169613,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.01836968120187521,
-      "rewards/tag_count_reward": 1.0,
+      "completion_length": 593.0,
+      "epoch": 0.177,
+      "grad_norm": 7.403178869874331,
+      "kl": 1.310546875,
+      "learning_rate": 9.838429319053495e-07,
+      "loss": 0.1308,
+      "reward": 2.756809949874878,
+      "reward_std": 0.3105107471346855,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.08173201233148575,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 354
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 577.9583435058594,
-      "epoch": 0.355,
-      "grad_norm": 4.692169147728827,
-      "kl": 0.69189453125,
-      "learning_rate": 8.331941759724268e-07,
-      "loss": 0.1281,
-      "reward": 2.5701699256896973,
-      "reward_std": 0.3895218074321747,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02705218270421028,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 587.6250305175781,
+      "epoch": 0.1775,
+      "grad_norm": 21.446192643297536,
+      "kl": 2.4765625,
+      "learning_rate": 9.836337039438803e-07,
+      "loss": 0.2692,
+      "reward": 2.3152072429656982,
+      "reward_std": 0.29953232035040855,
+      "rewards/accuracy_reward": 0.4166666716337204,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.04763999953866005,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 355
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 601.1458435058594,
-      "epoch": 0.356,
-      "grad_norm": 10.169896342116992,
-      "kl": 0.734375,
-      "learning_rate": 8.319717151140072e-07,
-      "loss": 0.2155,
-      "reward": 2.395414352416992,
-      "reward_std": 0.416961133480072,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026460560970008373,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 559.5625305175781,
+      "epoch": 0.178,
+      "grad_norm": 28.939767658793734,
+      "kl": 3.359375,
+      "learning_rate": 9.83423155058946e-07,
+      "loss": 0.3177,
+      "reward": 2.3154940605163574,
+      "reward_std": 0.4568801373243332,
+      "rewards/accuracy_reward": 0.4166666716337204,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03693648986518383,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 356
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 546.5416870117188,
-      "epoch": 0.357,
-      "grad_norm": 11.244917008930972,
-      "kl": 0.6279296875,
-      "learning_rate": 8.307458185192238e-07,
-      "loss": 0.1956,
-      "reward": 2.8225170373916626,
-      "reward_std": 0.34238358587026596,
-      "rewards/accuracy_reward": 0.8958333730697632,
+      "completion_length": 534.0,
+      "epoch": 0.1785,
+      "grad_norm": 32.29729153403837,
+      "kl": 2.34375,
+      "learning_rate": 9.832112858919155e-07,
+      "loss": 0.0816,
+      "reward": 2.6441575288772583,
+      "reward_std": 0.3644861727952957,
+      "rewards/accuracy_reward": 0.6875000298023224,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.021233050152659416,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "rewards/repetition_penalty_reward": -0.038134196773171425,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 357
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 601.2916870117188,
-      "epoch": 0.358,
-      "grad_norm": 6.159922849123736,
-      "kl": 0.908203125,
-      "learning_rate": 8.295165011252396e-07,
-      "loss": 0.1542,
-      "reward": 2.7077341079711914,
-      "reward_std": 0.3084152042865753,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03184933587908745,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 563.75,
+      "epoch": 0.179,
+      "grad_norm": 13.6168708257534,
+      "kl": 1.30078125,
+      "learning_rate": 9.829980970881784e-07,
+      "loss": 0.0706,
+      "reward": 2.5547882318496704,
+      "reward_std": 0.30551889538764954,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03548960853368044,
+      "rewards/tag_count_reward": 1.0,
       "step": 358
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 506.85418701171875,
-      "epoch": 0.359,
-      "grad_norm": 4.104290958705583,
-      "kl": 0.3466796875,
-      "learning_rate": 8.282837779108993e-07,
-      "loss": 0.0297,
-      "reward": 2.7414658069610596,
-      "reward_std": 0.21230606734752655,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02415929874405265,
-      "rewards/tag_count_reward": 0.9947916865348816,
+      "completion_length": 569.4791870117188,
+      "epoch": 0.1795,
+      "grad_norm": 9.408634350051774,
+      "kl": 0.673828125,
+      "learning_rate": 9.82783589297145e-07,
+      "loss": 0.0145,
+      "reward": 2.756742000579834,
+      "reward_std": 0.25605448335409164,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.06443855538964272,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 359
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 547.5000305175781,
-      "epoch": 0.36,
-      "grad_norm": 20.57792012790243,
-      "kl": 2.328125,
-      "learning_rate": 8.270476638965461e-07,
-      "loss": 0.502,
-      "reward": 2.5797927379608154,
-      "reward_std": 0.31256987154483795,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.01916569098830223,
-      "rewards/tag_count_reward": 0.9531250298023224,
+      "completion_length": 503.06251525878906,
+      "epoch": 0.18,
+      "grad_norm": 3.6640521264364514,
+      "kl": 0.25537109375,
+      "learning_rate": 9.825677631722435e-07,
+      "loss": -0.0335,
+      "reward": 2.7158700227737427,
+      "reward_std": 0.3013303726911545,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.07926897704601288,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 360
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 729.3541870117188,
-      "epoch": 0.361,
-      "grad_norm": 42.22333450542258,
-      "kl": 6.875,
-      "learning_rate": 8.258081741438394e-07,
-      "loss": 0.8517,
-      "reward": 2.3151514530181885,
-      "reward_std": 0.5804566144943237,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.021654206328094006,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 513.1250305175781,
+      "epoch": 0.1805,
+      "grad_norm": 5.09067870684325,
+      "kl": 0.1796875,
+      "learning_rate": 9.823506193709174e-07,
+      "loss": -0.0324,
+      "reward": 2.6296072006225586,
+      "reward_std": 0.17554676160216331,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.059629036113619804,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 361
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 597.5416870117188,
-      "epoch": 0.362,
-      "grad_norm": 20.91301673590996,
-      "kl": 3.27734375,
-      "learning_rate": 8.245653237555705e-07,
-      "loss": 0.4718,
-      "reward": 2.5699336528778076,
-      "reward_std": 0.4157796800136566,
-      "rewards/accuracy_reward": 0.6666666865348816,
+      "completion_length": 527.1875152587891,
+      "epoch": 0.181,
+      "grad_norm": 2.3488865190438,
+      "kl": 0.1416015625,
+      "learning_rate": 9.821321585546243e-07,
+      "loss": -0.0337,
+      "reward": 2.4458130598068237,
+      "reward_std": 0.36342713236808777,
+      "rewards/accuracy_reward": 0.5416666865348816,
       "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.03076085541397333,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "rewards/repetition_penalty_reward": -0.08196471631526947,
+      "rewards/tag_count_reward": 1.0,
       "step": 362
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 526.9166717529297,
-      "epoch": 0.363,
-      "grad_norm": 15.16442092292686,
-      "kl": 1.78515625,
-      "learning_rate": 8.23319127875479e-07,
-      "loss": 0.2643,
-      "reward": 2.8612154722213745,
-      "reward_std": 0.2645837068557739,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.020729007199406624,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 511.06251525878906,
+      "epoch": 0.1815,
+      "grad_norm": 2.429968776707808,
+      "kl": 0.13134765625,
+      "learning_rate": 9.81912381388834e-07,
+      "loss": 0.0443,
+      "reward": 2.6360349655151367,
+      "reward_std": 0.3865511268377304,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.06535414233803749,
+      "rewards/tag_count_reward": 1.0,
       "step": 363
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 574.6250152587891,
-      "epoch": 0.364,
-      "grad_norm": 27.79134380026425,
-      "kl": 2.84375,
-      "learning_rate": 8.220696016880687e-07,
-      "loss": 0.4106,
-      "reward": 2.6281174421310425,
-      "reward_std": 0.4037089943885803,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.015979719813913107,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 567.7708740234375,
+      "epoch": 0.182,
+      "grad_norm": 2.0606641273830903,
+      "kl": 0.13134765625,
+      "learning_rate": 9.816912885430258e-07,
+      "loss": 0.0193,
+      "reward": 2.468637228012085,
+      "reward_std": 0.42131057381629944,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.07302946597337723,
+      "rewards/tag_count_reward": 1.0,
       "step": 364
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 736.8958435058594,
-      "epoch": 0.365,
-      "grad_norm": 20.741864587098874,
-      "kl": 3.328125,
-      "learning_rate": 8.208167604184217e-07,
-      "loss": 0.5224,
-      "reward": 2.400440216064453,
-      "reward_std": 0.5106681287288666,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.024907216429710388,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 610.7291870117188,
+      "epoch": 0.1825,
+      "grad_norm": 2.0345217984138086,
+      "kl": 0.127197265625,
+      "learning_rate": 9.814688806906868e-07,
+      "loss": 0.0449,
+      "reward": 2.4708261489868164,
+      "reward_std": 0.3200060650706291,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.07778490334749222,
+      "rewards/tag_count_reward": 1.0,
       "step": 365
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 651.7708435058594,
-      "epoch": 0.366,
-      "grad_norm": 14.227390419980614,
-      "kl": 2.10546875,
-      "learning_rate": 8.195606193320136e-07,
-      "loss": 0.3847,
-      "reward": 2.4665273427963257,
-      "reward_std": 0.6216453611850739,
-      "rewards/accuracy_reward": 0.5833333730697632,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03347271308302879,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 524.7083435058594,
+      "epoch": 0.183,
+      "grad_norm": 2.1808261277456205,
+      "kl": 0.12158203125,
+      "learning_rate": 9.812451585093098e-07,
+      "loss": 0.0236,
+      "reward": 2.7193844318389893,
+      "reward_std": 0.34990330785512924,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.05839330144226551,
+      "rewards/tag_count_reward": 1.0,
       "step": 366
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 714.1458740234375,
-      "epoch": 0.367,
-      "grad_norm": 15.698648427320858,
-      "kl": 2.15625,
-      "learning_rate": 8.183011937345271e-07,
-      "loss": 0.3797,
-      "reward": 2.530668020248413,
-      "reward_std": 0.5888173580169678,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02141545619815588,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 483.68751525878906,
+      "epoch": 0.1835,
+      "grad_norm": 2.2192226472108367,
+      "kl": 0.12939453125,
+      "learning_rate": 9.810201226803917e-07,
+      "loss": 0.0069,
+      "reward": 2.4758822917938232,
+      "reward_std": 0.3382147550582886,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.058839818462729454,
+      "rewards/tag_count_reward": 1.0,
       "step": 367
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 646.0000305175781,
-      "epoch": 0.368,
-      "grad_norm": 24.843348556432016,
-      "kl": 1.98828125,
-      "learning_rate": 8.170384989716657e-07,
-      "loss": 0.5184,
-      "reward": 2.5058876276016235,
-      "reward_std": 0.6075495481491089,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.023626457899808884,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 519.3750152587891,
+      "epoch": 0.184,
+      "grad_norm": 2.367140341215985,
+      "kl": 0.128662109375,
+      "learning_rate": 9.807937738894303e-07,
+      "loss": -0.0302,
+      "reward": 2.6772682666778564,
+      "reward_std": 0.2594939023256302,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.051898419857025146,
+      "rewards/tag_count_reward": 1.0,
       "step": 368
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 597.1875,
-      "epoch": 0.369,
-      "grad_norm": 16.338985391404897,
-      "kl": 1.15625,
-      "learning_rate": 8.157725504289664e-07,
-      "loss": 0.3953,
-      "reward": 2.7404117584228516,
-      "reward_std": 0.3872096836566925,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.02347709983587265,
-      "rewards/tag_count_reward": 0.9583333432674408,
+      "completion_length": 513.6875,
+      "epoch": 0.1845,
+      "grad_norm": 2.3819557955341617,
+      "kl": 0.13525390625,
+      "learning_rate": 9.805661128259235e-07,
+      "loss": -0.0399,
+      "reward": 2.5830910205841064,
+      "reward_std": 0.43722137808799744,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.06968686729669571,
+      "rewards/tag_count_reward": 1.0,
       "step": 369
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 596.0833435058594,
-      "epoch": 0.37,
-      "grad_norm": 13.06241235347548,
-      "kl": 1.5,
-      "learning_rate": 8.145033635316128e-07,
-      "loss": 0.3765,
-      "reward": 2.6401225328445435,
-      "reward_std": 0.5106720924377441,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.01959980558604002,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 563.4166870117188,
+      "epoch": 0.185,
+      "grad_norm": 2.4301633780659135,
+      "kl": 0.131103515625,
+      "learning_rate": 9.80337140183366e-07,
+      "loss": 0.0573,
+      "reward": 2.719074010848999,
+      "reward_std": 0.37694530189037323,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.06564832478761673,
+      "rewards/tag_count_reward": 1.0,
       "step": 370
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 726.5208435058594,
-      "epoch": 0.371,
-      "grad_norm": 14.668565421161508,
-      "kl": 2.65625,
-      "learning_rate": 8.13230953744247e-07,
-      "loss": 0.5462,
-      "reward": 2.342048168182373,
-      "reward_std": 0.4522154927253723,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02774341218173504,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 482.1875,
+      "epoch": 0.1855,
+      "grad_norm": 2.4697642245743032,
+      "kl": 0.134521484375,
+      "learning_rate": 9.801068566592483e-07,
+      "loss": 0.0284,
+      "reward": 2.754256248474121,
+      "reward_std": 0.2568470761179924,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.06692435592412949,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 371
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 634.5416870117188,
-      "epoch": 0.372,
-      "grad_norm": 10.126972374008618,
-      "kl": 2.734375,
-      "learning_rate": 8.119553365707802e-07,
-      "loss": 0.5343,
-      "reward": 2.6384775638580322,
-      "reward_std": 0.559327244758606,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.016036429908126593,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 509.1666717529297,
+      "epoch": 0.186,
+      "grad_norm": 2.448475553737196,
+      "kl": 0.116943359375,
+      "learning_rate": 9.798752629550546e-07,
+      "loss": 0.0663,
+      "reward": 2.611233115196228,
+      "reward_std": 0.36806730926036835,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.07800298929214478,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 372
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 620.6041870117188,
-      "epoch": 0.373,
-      "grad_norm": 12.491483118539476,
-      "kl": 3.515625,
-      "learning_rate": 8.106765275542053e-07,
-      "loss": 0.608,
-      "reward": 2.324834704399109,
-      "reward_std": 0.6106734275817871,
-      "rewards/accuracy_reward": 0.4791666716337204,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.022387592121958733,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 533.8125305175781,
+      "epoch": 0.1865,
+      "grad_norm": 2.0486712055274343,
+      "kl": 0.118896484375,
+      "learning_rate": 9.796423597762588e-07,
+      "loss": -0.0294,
+      "reward": 2.593945264816284,
+      "reward_std": 0.43071576952934265,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.09702702611684799,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 373
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 595.9166870117188,
-      "epoch": 0.374,
-      "grad_norm": 34.92698420205304,
-      "kl": 4.8046875,
-      "learning_rate": 8.093945422764069e-07,
-      "loss": 0.7281,
-      "reward": 2.6930631399154663,
-      "reward_std": 0.4513649195432663,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.015270282980054617,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 500.25001525878906,
+      "epoch": 0.187,
+      "grad_norm": 2.3264130204399525,
+      "kl": 0.120849609375,
+      "learning_rate": 9.794081478323245e-07,
+      "loss": 0.0417,
+      "reward": 2.555578351020813,
+      "reward_std": 0.32255755364894867,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.10414392501115799,
+      "rewards/tag_count_reward": 1.0,
       "step": 374
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 555.0000152587891,
-      "epoch": 0.375,
-      "grad_norm": 20.20783740449395,
-      "kl": 3.421875,
-      "learning_rate": 8.081093963579707e-07,
-      "loss": 0.6528,
-      "reward": 2.706920862197876,
-      "reward_std": 0.5360372513532639,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.01703750714659691,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 563.0625305175781,
+      "epoch": 0.1875,
+      "grad_norm": 2.1622259964161303,
+      "kl": 0.12548828125,
+      "learning_rate": 9.791726278367021e-07,
+      "loss": -0.0156,
+      "reward": 2.6402961015701294,
+      "reward_std": 0.40212856233119965,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.07671798765659332,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 375
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 809.8125,
-      "epoch": 0.376,
-      "grad_norm": 42.45590838876723,
-      "kl": 8.359375,
-      "learning_rate": 8.068211054579943e-07,
-      "loss": 0.939,
-      "reward": 2.4243576526641846,
-      "reward_std": 0.6719434857368469,
-      "rewards/accuracy_reward": 0.6041666865348816,
+      "completion_length": 503.00001525878906,
+      "epoch": 0.188,
+      "grad_norm": 1.9354230871668425,
+      "kl": 0.109375,
+      "learning_rate": 9.78935800506826e-07,
+      "loss": 0.0093,
+      "reward": 2.648646354675293,
+      "reward_std": 0.22648237645626068,
+      "rewards/accuracy_reward": 0.7708333730697632,
       "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.033975621685385704,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "rewards/repetition_penalty_reward": -0.10135376825928688,
+      "rewards/tag_count_reward": 1.0,
       "step": 376
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 618.2083435058594,
-      "epoch": 0.377,
-      "grad_norm": 8.055288053209214,
-      "kl": 2.9609375,
-      "learning_rate": 8.055296852738956e-07,
-      "loss": 0.5277,
-      "reward": 2.515875458717346,
-      "reward_std": 0.6309832334518433,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.018846786580979824,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 452.4791717529297,
+      "epoch": 0.1885,
+      "grad_norm": 2.4900006677038897,
+      "kl": 0.1298828125,
+      "learning_rate": 9.786976665641138e-07,
+      "loss": 0.0358,
+      "reward": 2.7121907472610474,
+      "reward_std": 0.29230934381484985,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.07600371912121773,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 377
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 531.5625,
-      "epoch": 0.378,
-      "grad_norm": 10.73266293127411,
-      "kl": 2.0234375,
-      "learning_rate": 8.04235151541222e-07,
-      "loss": 0.5295,
-      "reward": 2.736036777496338,
-      "reward_std": 0.5330559611320496,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.012227283790707588,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 510.14585876464844,
+      "epoch": 0.189,
+      "grad_norm": 2.3438629757333143,
+      "kl": 0.12646484375,
+      "learning_rate": 9.784582267339622e-07,
+      "loss": 0.0447,
+      "reward": 2.517542004585266,
+      "reward_std": 0.44227664172649384,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.07968007400631905,
+      "rewards/tag_count_reward": 1.0,
       "step": 378
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 577.9166717529297,
-      "epoch": 0.379,
-      "grad_norm": 30.827479254698062,
-      "kl": 1.96484375,
-      "learning_rate": 8.029375200334587e-07,
-      "loss": 0.4051,
-      "reward": 2.63044536113739,
-      "reward_std": 0.4179770350456238,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.02059647301211953,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 498.12501525878906,
+      "epoch": 0.1895,
+      "grad_norm": 2.352684614662993,
+      "kl": 0.1328125,
+      "learning_rate": 9.78217481745747e-07,
+      "loss": -0.0119,
+      "reward": 2.3344022035598755,
+      "reward_std": 0.4069585055112839,
+      "rewards/accuracy_reward": 0.4375000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.049278499558568,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 379
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 571.625,
-      "epoch": 0.38,
-      "grad_norm": 11.564342505897628,
-      "kl": 0.927734375,
-      "learning_rate": 8.01636806561836e-07,
-      "loss": 0.286,
-      "reward": 2.6674641370773315,
-      "reward_std": 0.36413097381591797,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.030452590435743332,
-      "rewards/tag_count_reward": 0.96875,
+      "completion_length": 446.18751525878906,
+      "epoch": 0.19,
+      "grad_norm": 2.4157444132247923,
+      "kl": 0.12841796875,
+      "learning_rate": 9.779754323328192e-07,
+      "loss": 0.057,
+      "reward": 2.842611312866211,
+      "reward_std": 0.24086056649684906,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.060166530311107635,
+      "rewards/tag_count_reward": 1.0,
       "step": 380
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 627.7083587646484,
-      "epoch": 0.381,
-      "grad_norm": 20.79576339395751,
-      "kl": 2.859375,
-      "learning_rate": 8.003330269751372e-07,
-      "loss": 0.473,
-      "reward": 2.5600701570510864,
-      "reward_std": 0.5540246367454529,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.02152709010988474,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 444.62501525878906,
+      "epoch": 0.1905,
+      "grad_norm": 2.419479532121062,
+      "kl": 0.14306640625,
+      "learning_rate": 9.777320792325025e-07,
+      "loss": 0.1048,
+      "reward": 2.7526127099990845,
+      "reward_std": 0.14365122094750404,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.05467919073998928,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 381
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 549.8125305175781,
-      "epoch": 0.382,
-      "grad_norm": 5.584732514662178,
-      "kl": 0.400390625,
-      "learning_rate": 7.990261971595048e-07,
-      "loss": 0.037,
-      "reward": 2.806509017944336,
-      "reward_std": 0.29304996132850647,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.030296694487333298,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 459.0833435058594,
+      "epoch": 0.191,
+      "grad_norm": 2.372783987416792,
+      "kl": 0.12890625,
+      "learning_rate": 9.774874231860935e-07,
+      "loss": 0.0066,
+      "reward": 2.371148109436035,
+      "reward_std": 0.40637652575969696,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.07329631596803665,
+      "rewards/tag_count_reward": 1.0,
       "step": 382
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 804.3541870117188,
-      "epoch": 0.383,
-      "grad_norm": 11.399359693772158,
-      "kl": 2.4296875,
-      "learning_rate": 7.977163330382479e-07,
-      "loss": 0.4973,
-      "reward": 2.5117881298065186,
-      "reward_std": 0.5285173058509827,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.028142395429313183,
-      "rewards/tag_count_reward": 0.8177083730697632,
+      "completion_length": 471.2291717529297,
+      "epoch": 0.1915,
+      "grad_norm": 2.226194607367302,
+      "kl": 0.13232421875,
+      "learning_rate": 9.772414649388568e-07,
+      "loss": 0.044,
+      "reward": 2.6466528177261353,
+      "reward_std": 0.2924405038356781,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.08251398801803589,
+      "rewards/tag_count_reward": 1.0,
       "step": 383
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 529.2291870117188,
-      "epoch": 0.384,
-      "grad_norm": 15.184188235037553,
-      "kl": 1.04296875,
-      "learning_rate": 7.964034505716476e-07,
-      "loss": 0.3124,
-      "reward": 2.5725693702697754,
-      "reward_std": 0.3802667409181595,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666567325592,
-      "rewards/repetition_penalty_reward": -0.015972374938428402,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 500.6041717529297,
+      "epoch": 0.192,
+      "grad_norm": 2.132124535999422,
+      "kl": 0.14892578125,
+      "learning_rate": 9.769942052400235e-07,
+      "loss": -0.031,
+      "reward": 2.7737842798233032,
+      "reward_std": 0.23282310366630554,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.07343816570937634,
+      "rewards/tag_count_reward": 1.0,
       "step": 384
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 677.3333435058594,
-      "epoch": 0.385,
-      "grad_norm": 6.513600915558159,
-      "kl": 2.9375,
-      "learning_rate": 7.950875657567621e-07,
-      "loss": 0.5198,
-      "reward": 2.4022648334503174,
-      "reward_std": 0.6920067071914673,
-      "rewards/accuracy_reward": 0.5833333730697632,
+      "completion_length": 534.7708435058594,
+      "epoch": 0.1925,
+      "grad_norm": 2.1795003146204626,
+      "kl": 0.1416015625,
+      "learning_rate": 9.767456448427896e-07,
+      "loss": 0.0287,
+      "reward": 2.5758708715438843,
+      "reward_std": 0.38380637764930725,
+      "rewards/accuracy_reward": 0.6666666865348816,
       "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.0178740993142128,
-      "rewards/tag_count_reward": 0.84375,
+      "rewards/repetition_penalty_reward": -0.07864313945174217,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 385
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 572.7708435058594,
-      "epoch": 0.386,
-      "grad_norm": 10.893931836031141,
-      "kl": 1.95703125,
-      "learning_rate": 7.93768694627233e-07,
-      "loss": 0.4043,
-      "reward": 2.5226422548294067,
-      "reward_std": 0.502282902598381,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.022496800869703293,
-      "rewards/tag_count_reward": 0.90625,
+      "completion_length": 500.5625305175781,
+      "epoch": 0.193,
+      "grad_norm": 2.3063809528831736,
+      "kl": 0.14013671875,
+      "learning_rate": 9.764957845043135e-07,
+      "loss": -0.021,
+      "reward": 2.755507707595825,
+      "reward_std": 0.29851196706295013,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.0500478558242321,
+      "rewards/tag_count_reward": 1.0,
       "step": 386
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 650.7083435058594,
-      "epoch": 0.387,
-      "grad_norm": 11.694285585261715,
-      "kl": 2.919921875,
-      "learning_rate": 7.924468532530883e-07,
-      "loss": 0.3373,
-      "reward": 2.491190791130066,
-      "reward_std": 0.4670180529356003,
-      "rewards/accuracy_reward": 0.645833358168602,
+      "completion_length": 503.33335876464844,
+      "epoch": 0.1935,
+      "grad_norm": 2.1645139159061717,
+      "kl": 0.131591796875,
+      "learning_rate": 9.76244624985713e-07,
+      "loss": 0.0266,
+      "reward": 2.502621650695801,
+      "reward_std": 0.42333585023880005,
+      "rewards/accuracy_reward": 0.583333358168602,
       "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.03311488404870033,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "rewards/repetition_penalty_reward": -0.06335068121552467,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 387
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 834.1041870117188,
-      "epoch": 0.388,
-      "grad_norm": 11.93709525868574,
-      "kl": 5.203125,
-      "learning_rate": 7.911220577405484e-07,
-      "loss": 0.6908,
-      "reward": 2.1863056421279907,
-      "reward_std": 0.6824186444282532,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.020291661843657494,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 477.0833435058594,
+      "epoch": 0.194,
+      "grad_norm": 2.2028964179218256,
+      "kl": 0.14697265625,
+      "learning_rate": 9.759921670520634e-07,
+      "loss": -0.0024,
+      "reward": 2.746849298477173,
+      "reward_std": 0.3060501739382744,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.058706341311335564,
+      "rewards/tag_count_reward": 1.0,
       "step": 388
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 692.3333435058594,
-      "epoch": 0.389,
-      "grad_norm": 9.021930665026163,
-      "kl": 3.2265625,
-      "learning_rate": 7.897943242318285e-07,
-      "loss": 0.5433,
-      "reward": 2.5233638286590576,
-      "reward_std": 0.5364340543746948,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.026983547024428844,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 488.56251525878906,
+      "epoch": 0.1945,
+      "grad_norm": 2.6481468308941714,
+      "kl": 0.1396484375,
+      "learning_rate": 9.757384114723953e-07,
+      "loss": -0.0251,
+      "reward": 2.859673857688904,
+      "reward_std": 0.1833506003022194,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.06220110505819321,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 389
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 705.8750305175781,
-      "epoch": 0.39,
-      "grad_norm": 13.205610036016818,
-      "kl": 3.453125,
-      "learning_rate": 7.884636689049422e-07,
-      "loss": 0.76,
-      "reward": 2.53139591217041,
-      "reward_std": 0.6152072250843048,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.018951344303786755,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 490.0625,
+      "epoch": 0.195,
+      "grad_norm": 2.245242392422714,
+      "kl": 0.14990234375,
+      "learning_rate": 9.754833590196926e-07,
+      "loss": -0.0097,
+      "reward": 2.5275352001190186,
+      "reward_std": 0.3732472062110901,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.055798135697841644,
+      "rewards/tag_count_reward": 1.0,
       "step": 390
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 733.6250305175781,
-      "epoch": 0.391,
-      "grad_norm": 7.373593323078646,
-      "kl": 3.265625,
-      "learning_rate": 7.871301079735049e-07,
-      "loss": 0.5816,
-      "reward": 2.51271653175354,
-      "reward_std": 0.5442458987236023,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.028950226493179798,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 527.5833587646484,
+      "epoch": 0.1955,
+      "grad_norm": 2.988740973444905,
+      "kl": 0.15478515625,
+      "learning_rate": 9.752270104708888e-07,
+      "loss": 0.1089,
+      "reward": 2.5857293605804443,
+      "reward_std": 0.4840521216392517,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.06531241536140442,
+      "rewards/tag_count_reward": 0.984375,
       "step": 391
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 897.1666870117188,
-      "epoch": 0.392,
-      "grad_norm": 24.00799478501999,
-      "kl": 6.640625,
-      "learning_rate": 7.857936576865356e-07,
-      "loss": 0.9816,
-      "reward": 2.3230772018432617,
-      "reward_std": 0.7363496124744415,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.013728468678891659,
-      "rewards/tag_count_reward": 0.6979166865348816,
+      "completion_length": 483.75001525878906,
+      "epoch": 0.196,
+      "grad_norm": 2.355763241309473,
+      "kl": 0.146484375,
+      "learning_rate": 9.749693666068663e-07,
+      "loss": 0.024,
+      "reward": 2.677127242088318,
+      "reward_std": 0.3999278247356415,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.06072022393345833,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 392
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 753.1666870117188,
-      "epoch": 0.393,
-      "grad_norm": 12.878430832163207,
-      "kl": 3.63671875,
-      "learning_rate": 7.844543343282595e-07,
-      "loss": 0.7256,
-      "reward": 2.5792795419692993,
-      "reward_std": 0.5005324482917786,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.026623360812664032,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 443.3541717529297,
+      "epoch": 0.1965,
+      "grad_norm": 2.370253055532675,
+      "kl": 0.15625,
+      "learning_rate": 9.747104282124531e-07,
+      "loss": 0.0625,
+      "reward": 2.6037063598632812,
+      "reward_std": 0.2899606078863144,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.0490715391933918,
+      "rewards/tag_count_reward": 1.0,
       "step": 393
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 723.9791870117188,
-      "epoch": 0.394,
-      "grad_norm": 22.790620068840962,
-      "kl": 3.4140625,
-      "learning_rate": 7.831121542179086e-07,
-      "loss": 0.794,
-      "reward": 2.396026372909546,
-      "reward_std": 0.5379204005002975,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02237648330628872,
-      "rewards/tag_count_reward": 0.8489583730697632,
+      "completion_length": 456.3333435058594,
+      "epoch": 0.197,
+      "grad_norm": 2.4610089541965694,
+      "kl": 0.15625,
+      "learning_rate": 9.744501960764203e-07,
+      "loss": 0.0754,
+      "reward": 2.8426754474639893,
+      "reward_std": 0.21852993965148926,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.07051906362175941,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 394
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 863.2291870117188,
-      "epoch": 0.395,
-      "grad_norm": 19.51867731167796,
-      "kl": 5.0625,
-      "learning_rate": 7.817671337095244e-07,
-      "loss": 0.5991,
-      "reward": 2.26111102104187,
-      "reward_std": 0.6565420031547546,
-      "rewards/accuracy_reward": 0.5000000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02013904694467783,
-      "rewards/tag_count_reward": 0.7812500298023224,
+      "completion_length": 597.6875,
+      "epoch": 0.1975,
+      "grad_norm": 3.405083047196158,
+      "kl": 0.166015625,
+      "learning_rate": 9.741886709914803e-07,
+      "loss": 0.1408,
+      "reward": 2.701059341430664,
+      "reward_std": 0.3547600954771042,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.06803778372704983,
+      "rewards/tag_count_reward": 0.984375,
       "step": 395
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 900.2708740234375,
-      "epoch": 0.396,
-      "grad_norm": 17.88490632655815,
-      "kl": 6.6171875,
-      "learning_rate": 7.804192891917571e-07,
-      "loss": 0.7442,
-      "reward": 2.096524953842163,
-      "reward_std": 0.6609176695346832,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.01458633691072464,
-      "rewards/tag_count_reward": 0.6875,
+      "completion_length": 473.7916717529297,
+      "epoch": 0.198,
+      "grad_norm": 2.7966864501073188,
+      "kl": 0.17822265625,
+      "learning_rate": 9.739258537542835e-07,
+      "loss": -0.039,
+      "reward": 2.5628772974014282,
+      "reward_std": 0.3083910197019577,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.0725395604968071,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 396
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 727.7083435058594,
-      "epoch": 0.397,
-      "grad_norm": 7.902766105189536,
-      "kl": 3.4375,
-      "learning_rate": 7.79068637087667e-07,
-      "loss": 0.619,
-      "reward": 2.3182718753814697,
-      "reward_std": 0.6975450813770294,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.027214372530579567,
-      "rewards/tag_count_reward": 0.796875,
+      "completion_length": 553.5833435058594,
+      "epoch": 0.1985,
+      "grad_norm": 5.916243349638,
+      "kl": 0.1669921875,
+      "learning_rate": 9.73661745165417e-07,
+      "loss": 0.1781,
+      "reward": 2.6928629875183105,
+      "reward_std": 0.3948093354701996,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.08144245855510235,
+      "rewards/tag_count_reward": 0.96875,
       "step": 397
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 923.0833435058594,
-      "epoch": 0.398,
-      "grad_norm": 8.813050784830676,
-      "kl": 4.828125,
-      "learning_rate": 7.777151938545235e-07,
-      "loss": 0.744,
-      "reward": 2.218664765357971,
-      "reward_std": 0.687986433506012,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026126965880393982,
-      "rewards/tag_count_reward": 0.7239583730697632,
+      "completion_length": 466.97918701171875,
+      "epoch": 0.199,
+      "grad_norm": 2.568535579007926,
+      "kl": 0.17041015625,
+      "learning_rate": 9.733963460294015e-07,
+      "loss": -0.0109,
+      "reward": 2.8347532749176025,
+      "reward_std": 0.23304447531700134,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.061080172657966614,
+      "rewards/tag_count_reward": 1.0,
       "step": 398
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 731.9583435058594,
-      "epoch": 0.399,
-      "grad_norm": 17.369257892310884,
-      "kl": 3.0703125,
-      "learning_rate": 7.763589759836058e-07,
-      "loss": 0.5943,
-      "reward": 2.4915345907211304,
-      "reward_std": 0.6682489514350891,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.020618132315576077,
-      "rewards/tag_count_reward": 0.8177083432674408,
+      "completion_length": 444.31251525878906,
+      "epoch": 0.1995,
+      "grad_norm": 2.6070633311402096,
+      "kl": 0.1962890625,
+      "learning_rate": 9.731296571546885e-07,
+      "loss": 0.0098,
+      "reward": 2.824558138847351,
+      "reward_std": 0.24229364097118378,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.07821962051093578,
+      "rewards/tag_count_reward": 1.0,
       "step": 399
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 669.4583435058594,
-      "epoch": 0.4,
-      "grad_norm": 28.5217854549646,
-      "kl": 2.2109375,
-      "learning_rate": 7.75e-07,
-      "loss": 0.495,
-      "reward": 2.6985886096954346,
-      "reward_std": 0.4976983517408371,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.027106057852506638,
-      "rewards/tag_count_reward": 0.8645833730697632,
+      "completion_length": 473.3541717529297,
+      "epoch": 0.2,
+      "grad_norm": 3.67637754483244,
+      "kl": 0.19482421875,
+      "learning_rate": 9.728616793536587e-07,
+      "loss": 0.0453,
+      "reward": 2.5268030166625977,
+      "reward_std": 0.3516158163547516,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.06521105766296387,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 400
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 747.625,
-      "epoch": 0.401,
-      "grad_norm": 8.8439464948951,
-      "kl": 2.84375,
-      "learning_rate": 7.736382824623999e-07,
-      "loss": 0.4971,
-      "reward": 2.371925711631775,
-      "reward_std": 0.5836665034294128,
+      "completion_length": 438.5416717529297,
+      "epoch": 0.2005,
+      "grad_norm": 6.51901226376256,
+      "kl": 0.208984375,
+      "learning_rate": 9.72592413442619e-07,
+      "loss": -0.0584,
+      "reward": 2.4788198471069336,
+      "reward_std": 0.30706703662872314,
       "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02390767727047205,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.0489580724388361,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 401
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 632.9791870117188,
-      "epoch": 0.402,
-      "grad_norm": 7.326080260643342,
-      "kl": 2.3515625,
-      "learning_rate": 7.72273839962904e-07,
-      "loss": 0.4332,
-      "reward": 2.304866313934326,
-      "reward_std": 0.5863041877746582,
-      "rewards/accuracy_reward": 0.5,
+      "completion_length": 488.50001525878906,
+      "epoch": 0.201,
+      "grad_norm": 2.34033624330632,
+      "kl": 0.18994140625,
+      "learning_rate": 9.723218602418e-07,
+      "loss": -0.016,
+      "reward": 2.7911014556884766,
+      "reward_std": 0.32635878026485443,
+      "rewards/accuracy_reward": 0.8750000298023224,
       "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.026731058955192566,
-      "rewards/tag_count_reward": 0.859375,
+      "rewards/repetition_penalty_reward": -0.05091256648302078,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 402
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 852.3958740234375,
-      "epoch": 0.403,
-      "grad_norm": 12.84430423544328,
-      "kl": 4.61328125,
-      "learning_rate": 7.709066891268133e-07,
-      "loss": 0.5593,
-      "reward": 2.23075133562088,
-      "reward_std": 0.5370394438505173,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.029665381647646427,
-      "rewards/tag_count_reward": 0.7395833432674408,
+      "completion_length": 476.56251525878906,
+      "epoch": 0.2015,
+      "grad_norm": 2.5623947724894274,
+      "kl": 0.189453125,
+      "learning_rate": 9.720500205753538e-07,
+      "loss": -0.0322,
+      "reward": 2.698399066925049,
+      "reward_std": 0.403295561671257,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.06548996269702911,
+      "rewards/tag_count_reward": 1.0,
       "step": 403
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 815.9375,
-      "epoch": 0.404,
-      "grad_norm": 19.101834754723175,
-      "kl": 5.0,
-      "learning_rate": 7.695368466124296e-07,
-      "loss": 0.5902,
-      "reward": 2.4460870027542114,
-      "reward_std": 0.5804774314165115,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.0174547852948308,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 471.64585876464844,
+      "epoch": 0.202,
+      "grad_norm": 2.6351050859979885,
+      "kl": 0.20703125,
+      "learning_rate": 9.717768952713511e-07,
+      "loss": 0.0172,
+      "reward": 2.64251971244812,
+      "reward_std": 0.2700469493865967,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.04845273122191429,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 404
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 951.1458740234375,
-      "epoch": 0.405,
-      "grad_norm": 11.760380958455592,
-      "kl": 6.1875,
-      "learning_rate": 7.681643291108517e-07,
-      "loss": 0.8655,
-      "reward": 2.131704032421112,
-      "reward_std": 0.7297748029232025,
-      "rewards/accuracy_reward": 0.4375000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.017601476050913334,
-      "rewards/tag_count_reward": 0.7187500298023224,
+      "completion_length": 490.85418701171875,
+      "epoch": 0.2025,
+      "grad_norm": 2.5295824552139585,
+      "kl": 0.18994140625,
+      "learning_rate": 9.71502485161779e-07,
+      "loss": -0.0095,
+      "reward": 2.682486653327942,
+      "reward_std": 0.47308868169784546,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.0432078093290329,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 405
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 657.4375305175781,
-      "epoch": 0.406,
-      "grad_norm": 8.483281318881998,
-      "kl": 2.7734375,
-      "learning_rate": 7.667891533457718e-07,
-      "loss": 0.3648,
-      "reward": 2.377819776535034,
-      "reward_std": 0.6631312072277069,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.019749819301068783,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 449.97918701171875,
+      "epoch": 0.203,
+      "grad_norm": 4.590231605072331,
+      "kl": 0.1875,
+      "learning_rate": 9.71226791082538e-07,
+      "loss": 0.0868,
+      "reward": 2.853319525718689,
+      "reward_std": 0.2601998746395111,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03730554319918156,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 406
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 532.6458587646484,
-      "epoch": 0.407,
-      "grad_norm": 14.061905199941569,
-      "kl": 0.779296875,
-      "learning_rate": 7.654113360732732e-07,
-      "loss": 0.1726,
-      "reward": 2.7471961975097656,
-      "reward_std": 0.508865624666214,
-      "rewards/accuracy_reward": 0.8333333730697632,
+      "completion_length": 459.7083435058594,
+      "epoch": 0.2035,
+      "grad_norm": 3.798867450655633,
+      "kl": 0.1826171875,
+      "learning_rate": 9.709498138734403e-07,
+      "loss": 0.0078,
+      "reward": 2.8343251943588257,
+      "reward_std": 0.3080275356769562,
+      "rewards/accuracy_reward": 0.8958333432674408,
       "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.027109289541840553,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "rewards/repetition_penalty_reward": -0.03893873654305935,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 407
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 706.9583435058594,
-      "epoch": 0.408,
-      "grad_norm": 11.779388411702078,
-      "kl": 2.3046875,
-      "learning_rate": 7.640308940816239e-07,
-      "loss": 0.4889,
-      "reward": 2.349950671195984,
-      "reward_std": 0.6774967312812805,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.014632808044552803,
-      "rewards/tag_count_reward": 0.8020833730697632,
+      "completion_length": 480.1041717529297,
+      "epoch": 0.204,
+      "grad_norm": 3.339773656845583,
+      "kl": 0.2431640625,
+      "learning_rate": 9.706715543782064e-07,
+      "loss": 0.0546,
+      "reward": 2.397212266921997,
+      "reward_std": 0.38606369495391846,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04376017488539219,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 408
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 765.8125305175781,
-      "epoch": 0.409,
-      "grad_norm": 15.586280796607687,
-      "kl": 3.109375,
-      "learning_rate": 7.626478441910744e-07,
-      "loss": 0.578,
-      "reward": 2.306677460670471,
-      "reward_std": 0.7071038484573364,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.016239337623119354,
-      "rewards/tag_count_reward": 0.7812500298023224,
+      "completion_length": 438.5833435058594,
+      "epoch": 0.2045,
+      "grad_norm": 4.374047431584956,
+      "kl": 0.2890625,
+      "learning_rate": 9.703920134444632e-07,
+      "loss": -0.0037,
+      "reward": 2.664395570755005,
+      "reward_std": 0.20635664463043213,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.08039623126387596,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 409
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 665.75,
-      "epoch": 0.41,
-      "grad_norm": 7.201254956040354,
-      "kl": 2.546875,
-      "learning_rate": 7.612622032536507e-07,
-      "loss": 0.3043,
-      "reward": 2.2321949005126953,
-      "reward_std": 0.5692197978496552,
-      "rewards/accuracy_reward": 0.4375,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.019541208632290363,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 437.2708435058594,
+      "epoch": 0.205,
+      "grad_norm": 4.577231643859023,
+      "kl": 0.33203125,
+      "learning_rate": 9.701111919237408e-07,
+      "loss": 0.0983,
+      "reward": 2.6673845052719116,
+      "reward_std": 0.28793033957481384,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03574047051370144,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 410
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 557.1250305175781,
-      "epoch": 0.411,
-      "grad_norm": 12.200566811103256,
-      "kl": 1.060546875,
-      "learning_rate": 7.59873988152951e-07,
-      "loss": 0.2802,
-      "reward": 2.6842163801193237,
-      "reward_std": 0.292233943939209,
-      "rewards/accuracy_reward": 0.7708333730697632,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.031061380170285702,
-      "rewards/tag_count_reward": 0.9583333730697632,
+      "completion_length": 411.68751525878906,
+      "epoch": 0.2055,
+      "grad_norm": 3.9132489725820614,
+      "kl": 0.314453125,
+      "learning_rate": 9.698290906714702e-07,
+      "loss": 0.058,
+      "reward": 2.741675853729248,
+      "reward_std": 0.25958670675754547,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.07256042212247849,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 411
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.4375305175781,
-      "epoch": 0.412,
-      "grad_norm": 20.90214024265737,
-      "kl": 1.5390625,
-      "learning_rate": 7.584832158039378e-07,
-      "loss": 0.2278,
-      "reward": 2.5838751792907715,
-      "reward_std": 0.5441871881484985,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03070823848247528,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "completion_length": 489.3125,
+      "epoch": 0.206,
+      "grad_norm": 11.654921545662503,
+      "kl": 0.4736328125,
+      "learning_rate": 9.695457105469804e-07,
+      "loss": 0.1564,
+      "reward": 2.6907659769058228,
+      "reward_std": 0.36029884219169617,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.07659525983035564,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 412
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 641.4791870117188,
-      "epoch": 0.413,
-      "grad_norm": 15.755392833494586,
-      "kl": 3.9609375,
-      "learning_rate": 7.570899031527332e-07,
-      "loss": 0.4663,
-      "reward": 2.525357723236084,
-      "reward_std": 0.607357531785965,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.0145728699862957,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 472.91668701171875,
+      "epoch": 0.2065,
+      "grad_norm": 8.960365959818047,
+      "kl": 0.87109375,
+      "learning_rate": 9.69261052413497e-07,
+      "loss": 0.1558,
+      "reward": 2.5001312494277954,
+      "reward_std": 0.5244300961494446,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.05542440339922905,
+      "rewards/tag_count_reward": 0.8958333730697632,
       "step": 413
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 899.6250305175781,
-      "epoch": 0.414,
-      "grad_norm": 21.34036856462466,
-      "kl": 6.265625,
-      "learning_rate": 7.556940671764124e-07,
-      "loss": 0.7498,
-      "reward": 2.206377148628235,
-      "reward_std": 0.7690772414207458,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.021053357981145382,
+      "completion_length": 789.2500305175781,
+      "epoch": 0.207,
+      "grad_norm": 13.034420101645079,
+      "kl": 2.1484375,
+      "learning_rate": 9.689751171381377e-07,
+      "loss": 0.4389,
+      "reward": 2.2836403846740723,
+      "reward_std": 0.55134117603302,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.06184585019946098,
       "rewards/tag_count_reward": 0.7135416865348816,
       "step": 414
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 902.6250305175781,
-      "epoch": 0.415,
-      "grad_norm": 14.645548610311325,
-      "kl": 5.703125,
-      "learning_rate": 7.54295724882796e-07,
-      "loss": 0.7326,
-      "reward": 2.3041797876358032,
-      "reward_std": 0.5700157135725021,
-      "rewards/accuracy_reward": 0.6041666865348816,
+      "completion_length": 678.8750305175781,
+      "epoch": 0.2075,
+      "grad_norm": 24.87345626627445,
+      "kl": 2.734375,
+      "learning_rate": 9.68687905591911e-07,
+      "loss": 0.2528,
+      "reward": 2.3144073486328125,
+      "reward_std": 0.4746939539909363,
+      "rewards/accuracy_reward": 0.6250000298023224,
       "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.020473050884902477,
+      "rewards/repetition_penalty_reward": -0.031078746542334557,
       "rewards/tag_count_reward": 0.7343750298023224,
       "step": 415
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 601.4375,
-      "epoch": 0.416,
-      "grad_norm": 9.196475410607844,
-      "kl": 1.935546875,
-      "learning_rate": 7.528948933102438e-07,
-      "loss": 0.3524,
-      "reward": 2.62975537776947,
-      "reward_std": 0.5219498127698898,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.024758631363511086,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 756.8958435058594,
+      "epoch": 0.208,
+      "grad_norm": 10.29769838863137,
+      "kl": 1.921875,
+      "learning_rate": 9.683994186497132e-07,
+      "loss": 0.1948,
+      "reward": 2.4876515865325928,
+      "reward_std": 0.3638898953795433,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.05227917805314064,
+      "rewards/tag_count_reward": 0.8177083432674408,
       "step": 416
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 541.1666870117188,
-      "epoch": 0.417,
-      "grad_norm": 6.279974669254126,
-      "kl": 1.14453125,
-      "learning_rate": 7.514915895274463e-07,
-      "loss": 0.2295,
-      "reward": 2.4813188314437866,
-      "reward_std": 0.307245125528425,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.022153427824378014,
-      "rewards/tag_count_reward": 0.96875,
+      "completion_length": 458.4583435058594,
+      "epoch": 0.2085,
+      "grad_norm": 4.792568485298934,
+      "kl": 0.650390625,
+      "learning_rate": 9.681096571903252e-07,
+      "loss": 0.0885,
+      "reward": 2.4742881059646606,
+      "reward_std": 0.4346305727958679,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.050017524510622025,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 417
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 608.6666870117188,
-      "epoch": 0.418,
-      "grad_norm": 13.670357428378258,
-      "kl": 2.1640625,
-      "learning_rate": 7.500858306332172e-07,
-      "loss": 0.4449,
-      "reward": 2.4430500268936157,
-      "reward_std": 0.5130402147769928,
-      "rewards/accuracy_reward": 0.5625,
+      "completion_length": 424.3125,
+      "epoch": 0.209,
+      "grad_norm": 5.038602331803735,
+      "kl": 0.322265625,
+      "learning_rate": 9.67818622096411e-07,
+      "loss": 0.0884,
+      "reward": 2.7949297428131104,
+      "reward_std": 0.28741036355495453,
+      "rewards/accuracy_reward": 0.8333333730697632,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.015283319167792797,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.038403624668717384,
+      "rewards/tag_count_reward": 1.0,
       "step": 418
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 513.375,
-      "epoch": 0.419,
-      "grad_norm": 8.83072473507965,
-      "kl": 1.2265625,
-      "learning_rate": 7.486776337562853e-07,
-      "loss": 0.2575,
-      "reward": 2.5676904916763306,
-      "reward_std": 0.4607061445713043,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.026059521362185478,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "completion_length": 450.9375,
+      "epoch": 0.2095,
+      "grad_norm": 5.384359454777853,
+      "kl": 0.287109375,
+      "learning_rate": 9.67526314254514e-07,
+      "loss": 0.0816,
+      "reward": 2.835049033164978,
+      "reward_std": 0.24460270255804062,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.048631489276885986,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 419
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.9166717529297,
-      "epoch": 0.42,
-      "grad_norm": 9.224829985446288,
-      "kl": 1.548828125,
-      "learning_rate": 7.472670160550848e-07,
-      "loss": 0.1764,
-      "reward": 2.311278820037842,
-      "reward_std": 0.4501464366912842,
-      "rewards/accuracy_reward": 0.3958333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.023790807463228703,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 541.2083435058594,
+      "epoch": 0.21,
+      "grad_norm": 20.48316004254315,
+      "kl": 0.9375,
+      "learning_rate": 9.672327345550543e-07,
+      "loss": 0.2865,
+      "reward": 2.627635359764099,
+      "reward_std": 0.5200231969356537,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.023406220600008965,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 420
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 707.0208435058594,
-      "epoch": 0.421,
-      "grad_norm": 9.815342189617178,
-      "kl": 3.53125,
-      "learning_rate": 7.458539947175473e-07,
-      "loss": 0.7108,
-      "reward": 2.3707098960876465,
-      "reward_std": 0.5124354809522629,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.019915098324418068,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 564.5833435058594,
+      "epoch": 0.2105,
+      "grad_norm": 12.49478929221166,
+      "kl": 1.31640625,
+      "learning_rate": 9.669378838923267e-07,
+      "loss": 0.4444,
+      "reward": 2.715390205383301,
+      "reward_std": 0.4231880307197571,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.03460977412760258,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 421
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 500.8750305175781,
-      "epoch": 0.422,
-      "grad_norm": 5.188831712175436,
-      "kl": 1.2841796875,
-      "learning_rate": 7.444385869608921e-07,
-      "loss": 0.2224,
-      "reward": 2.582339286804199,
-      "reward_std": 0.25641510635614395,
-      "rewards/accuracy_reward": 0.6875000149011612,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.02529969811439514,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 962.4583435058594,
+      "epoch": 0.211,
+      "grad_norm": 23.747666802030537,
+      "kl": 3.1328125,
+      "learning_rate": 9.666417631644976e-07,
+      "loss": 0.4393,
+      "reward": 2.1522072553634644,
+      "reward_std": 0.5904508531093597,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.02487611025571823,
+      "rewards/tag_count_reward": 0.7812500298023224,
       "step": 422
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 470.8333435058594,
-      "epoch": 0.423,
-      "grad_norm": 8.589041831141525,
-      "kl": 0.720703125,
-      "learning_rate": 7.430208100314156e-07,
-      "loss": 0.1102,
-      "reward": 2.710915446281433,
-      "reward_std": 0.390904039144516,
-      "rewards/accuracy_reward": 0.7708333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.026931931264698505,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 1850.354248046875,
+      "epoch": 0.2115,
+      "grad_norm": 22.007156665570427,
+      "kl": 4.5625,
+      "learning_rate": 9.66344373273602e-07,
+      "loss": 0.3166,
+      "reward": 1.7171184420585632,
+      "reward_std": 0.5808897018432617,
+      "rewards/accuracy_reward": 0.1875,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.0033677266910672188,
+      "rewards/tag_count_reward": 0.5885416716337204,
       "step": 423
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 489.8333435058594,
-      "epoch": 0.424,
-      "grad_norm": 10.226503973140625,
-      "kl": 1.58984375,
-      "learning_rate": 7.416006812042827e-07,
-      "loss": 0.3074,
-      "reward": 2.462380290031433,
-      "reward_std": 0.39044664800167084,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.016786448657512665,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 1172.25,
+      "epoch": 0.212,
+      "grad_norm": 13.046087807963628,
+      "kl": 3.3359375,
+      "learning_rate": 9.66045715125541e-07,
+      "loss": 0.5982,
+      "reward": 2.00005042552948,
+      "reward_std": 0.6750081181526184,
+      "rewards/accuracy_reward": 0.375,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.0121024283580482,
+      "rewards/tag_count_reward": 0.6927083432674408,
       "step": 424
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 642.6666870117188,
-      "epoch": 0.425,
-      "grad_norm": 14.863563634108994,
-      "kl": 4.3671875,
-      "learning_rate": 7.401782177833147e-07,
-      "loss": 0.7175,
-      "reward": 2.3808945417404175,
-      "reward_std": 0.6382828205823898,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.03750831447541714,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 777.1666870117188,
+      "epoch": 0.2125,
+      "grad_norm": 8.697075914848647,
+      "kl": 1.41015625,
+      "learning_rate": 9.657457896300791e-07,
+      "loss": 0.4061,
+      "reward": 2.4428231716156006,
+      "reward_std": 0.44499611109495163,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03634357824921608,
+      "rewards/tag_count_reward": 0.8541666865348816,
       "step": 425
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 619.1666870117188,
-      "epoch": 0.426,
-      "grad_norm": 10.866155494400388,
-      "kl": 2.935546875,
-      "learning_rate": 7.387534371007797e-07,
-      "loss": 0.3152,
-      "reward": 2.6118627786636353,
-      "reward_std": 0.29477719962596893,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.028762279078364372,
-      "rewards/tag_count_reward": 0.9114583730697632,
+      "completion_length": 629.6875152587891,
+      "epoch": 0.213,
+      "grad_norm": 12.127551785913408,
+      "kl": 1.099609375,
+      "learning_rate": 9.654445977008414e-07,
+      "loss": 0.4644,
+      "reward": 2.2689449787139893,
+      "reward_std": 0.3217324912548065,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.026193935424089432,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 426
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 638.8333435058594,
-      "epoch": 0.427,
-      "grad_norm": 11.888028322568847,
-      "kl": 3.5625,
-      "learning_rate": 7.373263565171805e-07,
-      "loss": 0.3712,
-      "reward": 2.460385322570801,
-      "reward_std": 0.48997509479522705,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.025725997984409332,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 435.3958435058594,
+      "epoch": 0.2135,
+      "grad_norm": 14.647819824321752,
+      "kl": 0.37890625,
+      "learning_rate": 9.651421402553108e-07,
+      "loss": 0.2392,
+      "reward": 2.611443519592285,
+      "reward_std": 0.21422222256660461,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.04654255323112011,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 427
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 579.4375,
-      "epoch": 0.428,
-      "grad_norm": 12.847512777948014,
-      "kl": 2.53515625,
-      "learning_rate": 7.358969934210438e-07,
-      "loss": 0.6338,
-      "reward": 2.6187379360198975,
-      "reward_std": 0.53090900182724,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.02362314984202385,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 741.8125,
+      "epoch": 0.214,
+      "grad_norm": 6.865030392191665,
+      "kl": 1.41015625,
+      "learning_rate": 9.648384182148252e-07,
+      "loss": 0.3917,
+      "reward": 2.2415451407432556,
+      "reward_std": 0.5755177438259125,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.03970504552125931,
+      "rewards/tag_count_reward": 0.8229166865348816,
       "step": 428
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 483.70835876464844,
-      "epoch": 0.429,
-      "grad_norm": 11.418979135114895,
-      "kl": 1.544921875,
-      "learning_rate": 7.344653652287077e-07,
-      "loss": 0.2723,
-      "reward": 2.708911895751953,
-      "reward_std": 0.45899534225463867,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.02199092786759138,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 539.2083435058594,
+      "epoch": 0.2145,
+      "grad_norm": 4.7239397954996205,
+      "kl": 0.71875,
+      "learning_rate": 9.645334325045745e-07,
+      "loss": 0.1712,
+      "reward": 2.2595887184143066,
+      "reward_std": 0.39921020716428757,
+      "rewards/accuracy_reward": 0.4375000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.04423084668815136,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 429
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 705.2083435058594,
-      "epoch": 0.43,
-      "grad_norm": 25.61496750803982,
-      "kl": 5.7578125,
-      "learning_rate": 7.330314893841101e-07,
-      "loss": 0.6231,
-      "reward": 2.4273844957351685,
-      "reward_std": 0.5663270652294159,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.02400454506278038,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "completion_length": 687.2083435058594,
+      "epoch": 0.215,
+      "grad_norm": 5.41675055528973,
+      "kl": 1.40234375,
+      "learning_rate": 9.64227184053598e-07,
+      "loss": 0.5379,
+      "reward": 2.37734055519104,
+      "reward_std": 0.4606352895498276,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.05668727494776249,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 430
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 462.1666717529297,
-      "epoch": 0.431,
-      "grad_norm": 16.91868590026825,
-      "kl": 1.2421875,
-      "learning_rate": 7.315953833585755e-07,
-      "loss": 0.1736,
-      "reward": 2.821496605873108,
-      "reward_std": 0.40228843688964844,
-      "rewards/accuracy_reward": 0.8958333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.020517424680292606,
-      "rewards/tag_count_reward": 0.9531250298023224,
-      "step": 431
-    },
-    {
+      "completion_length": 852.3958435058594,
+      "epoch": 0.2155,
+      "grad_norm": 12.180552388615105,
+      "kl": 2.3359375,
+      "learning_rate": 9.63919673794782e-07,
+      "loss": 0.55,
+      "reward": 2.2885630130767822,
+      "reward_std": 0.5578717291355133,
+      "rewards/accuracy_reward": 0.5416666716337204,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.03088139370083809,
+      "rewards/tag_count_reward": 0.8333333730697632,
+      "step": 431
+    },
+    {
       "clip_ratio": 0.0,
-      "completion_length": 564.7708435058594,
-      "epoch": 0.432,
-      "grad_norm": 10.05316608237381,
-      "kl": 3.078125,
-      "learning_rate": 7.301570646506027e-07,
-      "loss": 0.7108,
-      "reward": 2.748389720916748,
-      "reward_std": 0.52855084836483,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02070755325257778,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 1157.7291870117188,
+      "epoch": 0.216,
+      "grad_norm": 21.263145759520075,
+      "kl": 3.4140625,
+      "learning_rate": 9.636109026648554e-07,
+      "loss": 0.514,
+      "reward": 1.711805820465088,
+      "reward_std": 0.5387972891330719,
+      "rewards/accuracy_reward": 0.1458333358168602,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.02777751348912716,
+      "rewards/tag_count_reward": 0.6562500298023224,
       "step": 432
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 779.3333435058594,
-      "epoch": 0.433,
-      "grad_norm": 27.40142017382295,
-      "kl": 7.359375,
-      "learning_rate": 7.287165507856512e-07,
-      "loss": 0.8687,
-      "reward": 2.269049644470215,
-      "reward_std": 0.7071286737918854,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.015672642271965742,
-      "rewards/tag_count_reward": 0.75,
+      "completion_length": 740.2500305175781,
+      "epoch": 0.2165,
+      "grad_norm": 8.545418331851147,
+      "kl": 1.56640625,
+      "learning_rate": 9.633008716043892e-07,
+      "loss": 0.3121,
+      "reward": 2.3099186420440674,
+      "reward_std": 0.5152548253536224,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.037303626537323,
+      "rewards/tag_count_reward": 0.8333333730697632,
       "step": 433
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 671.7916870117188,
-      "epoch": 0.434,
-      "grad_norm": 21.75332490966816,
-      "kl": 4.8125,
-      "learning_rate": 7.27273859315928e-07,
-      "loss": 0.6014,
-      "reward": 2.483176350593567,
-      "reward_std": 0.5736861526966095,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222983837128,
-      "rewards/repetition_penalty_reward": -0.02550439164042473,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 547.2708435058594,
+      "epoch": 0.217,
+      "grad_norm": 4.67948611938176,
+      "kl": 1.0,
+      "learning_rate": 9.629895815577915e-07,
+      "loss": 0.358,
+      "reward": 2.493169069290161,
+      "reward_std": 0.6183354258537292,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.022456128150224686,
+      "rewards/tag_count_reward": 0.8906250298023224,
       "step": 434
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 847.4375305175781,
-      "epoch": 0.435,
-      "grad_norm": 27.079037368491598,
-      "kl": 6.625,
-      "learning_rate": 7.258290078201731e-07,
-      "loss": 0.8844,
-      "reward": 1.987673282623291,
-      "reward_std": 0.6426667273044586,
-      "rewards/accuracy_reward": 0.2708333358168602,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.027951689437031746,
-      "rewards/tag_count_reward": 0.7447916865348816,
+      "completion_length": 468.25,
+      "epoch": 0.2175,
+      "grad_norm": 5.989166621610532,
+      "kl": 0.51123046875,
+      "learning_rate": 9.626770334733058e-07,
+      "loss": 0.108,
+      "reward": 2.392013192176819,
+      "reward_std": 0.47137293219566345,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.036806222051382065,
+      "rewards/tag_count_reward": 0.921875,
       "step": 435
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 611.125,
-      "epoch": 0.436,
-      "grad_norm": 10.091665515824934,
-      "kl": 4.046875,
-      "learning_rate": 7.243820139034464e-07,
-      "loss": 0.6799,
-      "reward": 2.2479928731918335,
-      "reward_std": 0.6026458740234375,
-      "rewards/accuracy_reward": 0.458333358168602,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.015896069817245007,
-      "rewards/tag_count_reward": 0.8125000298023224,
+      "completion_length": 488.8958435058594,
+      "epoch": 0.218,
+      "grad_norm": 9.738518456979653,
+      "kl": 0.54296875,
+      "learning_rate": 9.623632283030077e-07,
+      "loss": 0.1873,
+      "reward": 2.3263269662857056,
+      "reward_std": 0.3005019724369049,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.05561743676662445,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 436
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 750.5833740234375,
-      "epoch": 0.437,
-      "grad_norm": 9.40390565894805,
-      "kl": 4.390625,
-      "learning_rate": 7.229328951969115e-07,
-      "loss": 0.7124,
-      "reward": 2.4180572032928467,
-      "reward_std": 0.6964816451072693,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.01944286935031414,
-      "rewards/tag_count_reward": 0.7916666865348816,
+      "completion_length": 400.2083435058594,
+      "epoch": 0.2185,
+      "grad_norm": 8.158385637923423,
+      "kl": 0.357421875,
+      "learning_rate": 9.620481670028026e-07,
+      "loss": 0.1567,
+      "reward": 2.596799850463867,
+      "reward_std": 0.24532540142536163,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.033408566378057,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 437
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 597.7500305175781,
-      "epoch": 0.438,
-      "grad_norm": 14.732573596337566,
-      "kl": 1.73828125,
-      "learning_rate": 7.214816693576234e-07,
-      "loss": 0.4652,
-      "reward": 2.5944327116012573,
-      "reward_std": 0.4662973880767822,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.021886682137846947,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 425.375,
+      "epoch": 0.219,
+      "grad_norm": 5.659616408154006,
+      "kl": 0.5888671875,
+      "learning_rate": 9.617318505324212e-07,
+      "loss": 0.0578,
+      "reward": 2.552333116531372,
+      "reward_std": 0.32321399450302124,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.04488925263285637,
+      "rewards/tag_count_reward": 0.9375,
       "step": 438
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 666.2083587646484,
-      "epoch": 0.439,
-      "grad_norm": 15.776670792271592,
-      "kl": 3.07421875,
-      "learning_rate": 7.200283540683102e-07,
-      "loss": 0.5074,
-      "reward": 2.3149194717407227,
-      "reward_std": 0.568027138710022,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.02014992106705904,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 420.0833435058594,
+      "epoch": 0.2195,
+      "grad_norm": 8.173779954105225,
+      "kl": 0.58984375,
+      "learning_rate": 9.614142798554186e-07,
+      "loss": 0.2351,
+      "reward": 2.7584837675094604,
+      "reward_std": 0.42886343598365784,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.040127404034137726,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 439
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 579.7500305175781,
-      "epoch": 0.44,
-      "grad_norm": 13.110846882696643,
-      "kl": 1.84765625,
-      "learning_rate": 7.185729670371604e-07,
-      "loss": 0.3424,
-      "reward": 2.5865591764450073,
-      "reward_std": 0.5619917958974838,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.024551907554268837,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 584.5208740234375,
+      "epoch": 0.22,
+      "grad_norm": 12.102598421917286,
+      "kl": 1.78515625,
+      "learning_rate": 9.610954559391704e-07,
+      "loss": 0.4177,
+      "reward": 2.3829479217529297,
+      "reward_std": 0.5593430995941162,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.06496894918382168,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 440
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 691.2916870117188,
-      "epoch": 0.441,
-      "grad_norm": 13.083680161708491,
-      "kl": 2.3515625,
-      "learning_rate": 7.171155259976057e-07,
-      "loss": 0.5228,
-      "reward": 2.60994291305542,
-      "reward_std": 0.5069815963506699,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.034154389053583145,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 666.3541870117188,
+      "epoch": 0.2205,
+      "grad_norm": 12.188104046012809,
+      "kl": 2.4296875,
+      "learning_rate": 9.607753797548691e-07,
+      "loss": 0.5983,
+      "reward": 2.4352097511291504,
+      "reward_std": 0.7042646259069443,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.8888888955116272,
+      "rewards/repetition_penalty_reward": -0.052637550979852676,
+      "rewards/tag_count_reward": 0.8489583432674408,
       "step": 441
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 646.6041870117188,
-      "epoch": 0.442,
-      "grad_norm": 50.887150157363536,
-      "kl": 4.890625,
-      "learning_rate": 7.156560487081051e-07,
-      "loss": 0.9085,
-      "reward": 2.2685351371765137,
-      "reward_std": 0.7841026484966278,
+      "completion_length": 619.4166870117188,
+      "epoch": 0.221,
+      "grad_norm": 17.355391790641967,
+      "kl": 2.421875,
+      "learning_rate": 9.604540522775227e-07,
+      "loss": 0.3384,
+      "reward": 2.2807594537734985,
+      "reward_std": 0.47133713960647583,
       "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03354842588305473,
-      "rewards/tag_count_reward": 0.7812500298023224,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.03521277289837599,
+      "rewards/tag_count_reward": 0.8645833730697632,
       "step": 442
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 755.8750305175781,
-      "epoch": 0.443,
-      "grad_norm": 15.956258650623388,
-      "kl": 5.609375,
-      "learning_rate": 7.141945529519288e-07,
-      "loss": 0.7977,
-      "reward": 2.418252944946289,
-      "reward_std": 0.683345377445221,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.027927756309509277,
-      "rewards/tag_count_reward": 0.828125,
+      "completion_length": 727.7291870117188,
+      "epoch": 0.2215,
+      "grad_norm": 19.28638169980206,
+      "kl": 2.24609375,
+      "learning_rate": 9.601314744859504e-07,
+      "loss": 0.692,
+      "reward": 2.0420679450035095,
+      "reward_std": 0.7488097846508026,
+      "rewards/accuracy_reward": 0.4166666716337204,
+      "rewards/reasoning_steps_reward": 0.8819444179534912,
+      "rewards/repetition_penalty_reward": -0.02737656608223915,
+      "rewards/tag_count_reward": 0.7708333730697632,
       "step": 443
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 743.5416870117188,
-      "epoch": 0.444,
-      "grad_norm": 24.6724782512694,
-      "kl": 7.890625,
-      "learning_rate": 7.127310565369415e-07,
-      "loss": 1.0148,
-      "reward": 2.227971076965332,
-      "reward_std": 0.8724583983421326,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.04980688448995352,
-      "rewards/tag_count_reward": 0.7500000298023224,
+      "completion_length": 447.93751525878906,
+      "epoch": 0.222,
+      "grad_norm": 3862503.859265399,
+      "kl": 2170.0,
+      "learning_rate": 9.598076473627796e-07,
+      "loss": 259.2312,
+      "reward": 2.4890825748443604,
+      "reward_std": 0.6550872325897217,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.03695919178426266,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 444
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 617.1458435058594,
-      "epoch": 0.445,
-      "grad_norm": 18.521282427965453,
-      "kl": 4.4140625,
-      "learning_rate": 7.11265577295385e-07,
-      "loss": 0.5287,
-      "reward": 2.4945348501205444,
-      "reward_std": 0.5981617867946625,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.01761806895956397,
-      "rewards/tag_count_reward": 0.8385416865348816,
+      "completion_length": 389.00001525878906,
+      "epoch": 0.2225,
+      "grad_norm": 20458711.67176828,
+      "kl": 37120.201171875,
+      "learning_rate": 9.594825718944444e-07,
+      "loss": 1892.2664,
+      "reward": 2.576767325401306,
+      "reward_std": 0.32869401574134827,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.8611111044883728,
+      "rewards/repetition_penalty_reward": -0.055177152156829834,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 445
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 732.5625305175781,
-      "epoch": 0.446,
-      "grad_norm": 34.24653385750582,
-      "kl": 6.484375,
-      "learning_rate": 7.097981330836616e-07,
-      "loss": 0.7675,
-      "reward": 1.965806484222412,
-      "reward_std": 0.6163901686668396,
-      "rewards/accuracy_reward": 0.2708333358168602,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.04113807622343302,
-      "rewards/tag_count_reward": 0.75,
+      "completion_length": 375.6666717529297,
+      "epoch": 0.223,
+      "grad_norm": 6.226734549137509,
+      "kl": 0.50830078125,
+      "learning_rate": 9.59156249071181e-07,
+      "loss": 0.01,
+      "reward": 2.7551599740982056,
+      "reward_std": 0.20863118767738342,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.04345105215907097,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 446
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 672.2708435058594,
-      "epoch": 0.447,
-      "grad_norm": 15.152108234478831,
-      "kl": 5.34375,
-      "learning_rate": 7.083287417821157e-07,
-      "loss": 0.7173,
-      "reward": 2.316531181335449,
-      "reward_std": 0.5732046067714691,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.018538246862590313,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 473.45835876464844,
+      "epoch": 0.2235,
+      "grad_norm": 5.567081842645192,
+      "kl": 0.68359375,
+      "learning_rate": 9.588286798870248e-07,
+      "loss": 0.1501,
+      "reward": 2.574015259742737,
+      "reward_std": 0.45759186148643494,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.049248674884438515,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 447
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 730.1875305175781,
-      "epoch": 0.448,
-      "grad_norm": 9.314353876415918,
-      "kl": 4.265625,
-      "learning_rate": 7.068574212948169e-07,
-      "loss": 0.7717,
-      "reward": 2.5467538833618164,
-      "reward_std": 0.5365928113460541,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.03137107007205486,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 493.04168701171875,
+      "epoch": 0.224,
+      "grad_norm": 4.077704475094219,
+      "kl": 0.634765625,
+      "learning_rate": 9.58499865339809e-07,
+      "loss": 0.0452,
+      "reward": 2.5271689891815186,
+      "reward_std": 0.3818424344062805,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.05963664874434471,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 448
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 650.6875305175781,
-      "epoch": 0.449,
-      "grad_norm": 46.24553233658842,
-      "kl": 2.59375,
-      "learning_rate": 7.053841895493406e-07,
-      "loss": 0.778,
-      "reward": 2.618269443511963,
-      "reward_std": 0.6291141211986542,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.029300065711140633,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 453.5833435058594,
+      "epoch": 0.2245,
+      "grad_norm": 4.624099127542048,
+      "kl": 0.5810546875,
+      "learning_rate": 9.581698064311592e-07,
+      "loss": 0.0303,
+      "reward": 2.3410807847976685,
+      "reward_std": 0.4072131812572479,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.054752614349126816,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 449
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 628.1458435058594,
-      "epoch": 0.45,
-      "grad_norm": 22.01768222586756,
-      "kl": 1.95703125,
-      "learning_rate": 7.039090644965509e-07,
-      "loss": 0.4547,
-      "reward": 2.60259747505188,
-      "reward_std": 0.5832863450050354,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.02934711705893278,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 336.81251525878906,
+      "epoch": 0.225,
+      "grad_norm": 4.480015408899521,
+      "kl": 0.248046875,
+      "learning_rate": 9.578385041664925e-07,
+      "loss": 0.03,
+      "reward": 2.6452146768569946,
+      "reward_std": 0.1535702757537365,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.8958334028720856,
+      "rewards/repetition_penalty_reward": -0.04228539019823074,
+      "rewards/tag_count_reward": 1.0,
       "step": 450
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 517.1458435058594,
-      "epoch": 0.451,
-      "grad_norm": 33.402974244613176,
-      "kl": 1.6484375,
-      "learning_rate": 7.024320641103811e-07,
-      "loss": 0.4482,
-      "reward": 2.561974287033081,
-      "reward_std": 0.5507599115371704,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.017886925488710403,
-      "rewards/tag_count_reward": 0.90625,
+      "completion_length": 464.5833435058594,
+      "epoch": 0.2255,
+      "grad_norm": 6.047533958166858,
+      "kl": 0.5625,
+      "learning_rate": 9.575059595550127e-07,
+      "loss": 0.1545,
+      "reward": 2.5978249311447144,
+      "reward_std": 0.3733315169811249,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.041063896380364895,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 451
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 501.41668701171875,
-      "epoch": 0.452,
-      "grad_norm": 12.388889897731671,
-      "kl": 1.1015625,
-      "learning_rate": 7.009532063876148e-07,
-      "loss": -0.028,
-      "reward": 2.4829607009887695,
-      "reward_std": 0.4351721853017807,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.022247745655477047,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 398.43751525878906,
+      "epoch": 0.226,
+      "grad_norm": 11.490949893657614,
+      "kl": 0.505859375,
+      "learning_rate": 9.571721736097088e-07,
+      "loss": 0.1815,
+      "reward": 2.7292104959487915,
+      "reward_std": 0.36046914756298065,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.05898405984044075,
+      "rewards/tag_count_reward": 0.96875,
       "step": 452
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 663.8125305175781,
-      "epoch": 0.453,
-      "grad_norm": 8.122371694455087,
-      "kl": 3.46875,
-      "learning_rate": 6.994725093476664e-07,
-      "loss": 0.586,
-      "reward": 2.5677295923233032,
-      "reward_std": 0.5909285247325897,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.0173398619517684,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 461.81251525878906,
+      "epoch": 0.2265,
+      "grad_norm": 6.831195619894885,
+      "kl": 1.2734375,
+      "learning_rate": 9.568371473473503e-07,
+      "loss": 0.2574,
+      "reward": 2.4620362520217896,
+      "reward_std": 0.4237401932477951,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.03796376567333937,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 453
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 724.4375457763672,
-      "epoch": 0.454,
-      "grad_norm": 24.750328325907926,
-      "kl": 3.8046875,
-      "learning_rate": 6.979899910323624e-07,
-      "loss": 0.4282,
-      "reward": 2.418397545814514,
-      "reward_std": 0.5793856680393219,
-      "rewards/accuracy_reward": 0.6458333730697632,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.022574756294488907,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 518.1458435058594,
+      "epoch": 0.227,
+      "grad_norm": 17.14178008843787,
+      "kl": 2.23828125,
+      "learning_rate": 9.565008817884854e-07,
+      "loss": 0.3425,
+      "reward": 2.2992377281188965,
+      "reward_std": 0.7252504229545593,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.86111119389534,
+      "rewards/repetition_penalty_reward": -0.025415126234292984,
+      "rewards/tag_count_reward": 0.8593750298023224,
       "step": 454
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 670.1875,
-      "epoch": 0.455,
-      "grad_norm": 32.10532320367166,
-      "kl": 2.8515625,
-      "learning_rate": 6.965056695057204e-07,
-      "loss": 0.5623,
-      "reward": 2.5078253746032715,
-      "reward_std": 0.7500589489936829,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.04252194054424763,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 386.7083435058594,
+      "epoch": 0.2275,
+      "grad_norm": 11.113475621778726,
+      "kl": 1.48828125,
+      "learning_rate": 9.561633779574372e-07,
+      "loss": 0.1915,
+      "reward": 2.620706558227539,
+      "reward_std": 0.4370867908000946,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.8958334028720856,
+      "rewards/repetition_penalty_reward": -0.03554350510239601,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 455
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 769.7083435058594,
-      "epoch": 0.456,
-      "grad_norm": 35.5370005747395,
-      "kl": 5.8203125,
-      "learning_rate": 6.950195628537299e-07,
-      "loss": 0.7239,
-      "reward": 2.236100912094116,
-      "reward_std": 0.6354574412107468,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.024315819144248962,
-      "rewards/tag_count_reward": 0.8020833432674408,
+      "completion_length": 441.91668701171875,
+      "epoch": 0.228,
+      "grad_norm": 4.925397785816655,
+      "kl": 0.98046875,
+      "learning_rate": 9.55824636882301e-07,
+      "loss": 0.3033,
+      "reward": 2.390235185623169,
+      "reward_std": 0.44205035269260406,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.875,
+      "rewards/repetition_penalty_reward": -0.042056621983647346,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 456
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 598.4166870117188,
-      "epoch": 0.457,
-      "grad_norm": 15.54748803668102,
-      "kl": 2.38671875,
-      "learning_rate": 6.935316891841315e-07,
-      "loss": 0.5143,
-      "reward": 2.461942434310913,
-      "reward_std": 0.5678964406251907,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02764105051755905,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 543.9583435058594,
+      "epoch": 0.2285,
+      "grad_norm": 6.065645620729275,
+      "kl": 0.68359375,
+      "learning_rate": 9.554846595949413e-07,
+      "loss": 0.2455,
+      "reward": 2.332500398159027,
+      "reward_std": 0.3640214204788208,
+      "rewards/accuracy_reward": 0.5000000111758709,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.06159696541726589,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 457
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 687.0208740234375,
-      "epoch": 0.458,
-      "grad_norm": 9.674950437917838,
-      "kl": 3.5078125,
-      "learning_rate": 6.920420666261961e-07,
-      "loss": 0.525,
-      "reward": 2.373473048210144,
-      "reward_std": 0.6431159377098083,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.02409642282873392,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 538.5416717529297,
+      "epoch": 0.229,
+      "grad_norm": 13.405805070703696,
+      "kl": 0.806640625,
+      "learning_rate": 9.55143447130987e-07,
+      "loss": 0.2765,
+      "reward": 2.613925814628601,
+      "reward_std": 0.4678662419319153,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.04926872253417969,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 458
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 597.1666870117188,
-      "epoch": 0.459,
-      "grad_norm": 11.267949423552796,
-      "kl": 2.6328125,
-      "learning_rate": 6.905507133305047e-07,
-      "loss": 0.5289,
-      "reward": 2.4548171758651733,
-      "reward_std": 0.5164096057415009,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.017405156511813402,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 684.75,
+      "epoch": 0.2295,
+      "grad_norm": 5.353993038328121,
+      "kl": 1.541015625,
+      "learning_rate": 9.54801000529831e-07,
+      "loss": 0.3747,
+      "reward": 2.072951376438141,
+      "reward_std": 0.6868791580200195,
+      "rewards/accuracy_reward": 0.458333358168602,
+      "rewards/reasoning_steps_reward": 0.8402778506278992,
+      "rewards/repetition_penalty_reward": -0.03295143134891987,
+      "rewards/tag_count_reward": 0.8072916865348816,
       "step": 459
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 624.125,
-      "epoch": 0.46,
-      "grad_norm": 11.88684382845764,
-      "kl": 3.703125,
-      "learning_rate": 6.890576474687263e-07,
-      "loss": 0.6898,
-      "reward": 2.5472575426101685,
-      "reward_std": 0.6651158332824707,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.016978577245026827,
-      "rewards/tag_count_reward": 0.8489583730697632,
+      "completion_length": 697.5416870117188,
+      "epoch": 0.23,
+      "grad_norm": 6.2141087927742555,
+      "kl": 1.39453125,
+      "learning_rate": 9.54457320834625e-07,
+      "loss": 0.5803,
+      "reward": 2.205751061439514,
+      "reward_std": 0.5500113666057587,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.018207357730716467,
+      "rewards/tag_count_reward": 0.8489583432674408,
       "step": 460
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 752.25,
-      "epoch": 0.461,
-      "grad_norm": 14.448078905297072,
-      "kl": 5.765625,
-      "learning_rate": 6.875628872333975e-07,
-      "loss": 0.7562,
-      "reward": 2.2633214592933655,
-      "reward_std": 0.6216214001178741,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02660918142646551,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 374.7708435058594,
+      "epoch": 0.2305,
+      "grad_norm": 4.32996026619283,
+      "kl": 0.51171875,
+      "learning_rate": 9.54112409092277e-07,
+      "loss": 0.1318,
+      "reward": 2.5193281173706055,
+      "reward_std": 0.2269902043044567,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.03969983570277691,
+      "rewards/tag_count_reward": 0.96875,
       "step": 461
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 708.3958740234375,
-      "epoch": 0.462,
-      "grad_norm": 20.449164875872643,
-      "kl": 7.078125,
-      "learning_rate": 6.860664508377001e-07,
-      "loss": 0.7331,
-      "reward": 2.1556172370910645,
-      "reward_std": 0.7438893914222717,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.01799399685114622,
-      "rewards/tag_count_reward": 0.7291666865348816,
+      "completion_length": 522.4791870117188,
+      "epoch": 0.231,
+      "grad_norm": 6.138638623677164,
+      "kl": 0.767578125,
+      "learning_rate": 9.537662663534477e-07,
+      "loss": 0.2944,
+      "reward": 2.0977752208709717,
+      "reward_std": 0.47963058948516846,
+      "rewards/accuracy_reward": 0.2708333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.06368312053382397,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 462
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 650.5000305175781,
-      "epoch": 0.463,
-      "grad_norm": 14.487854173063429,
-      "kl": 4.7265625,
-      "learning_rate": 6.84568356515239e-07,
-      "loss": 0.818,
-      "reward": 2.314304828643799,
-      "reward_std": 0.6960805356502533,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.025973046198487282,
-      "rewards/tag_count_reward": 0.8125000298023224,
+      "completion_length": 549.1666870117188,
+      "epoch": 0.2315,
+      "grad_norm": 4.983489188907526,
+      "kl": 0.802734375,
+      "learning_rate": 9.534188936725483e-07,
+      "loss": 0.209,
+      "reward": 2.581421971321106,
+      "reward_std": 0.44028259813785553,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.03316138405352831,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 463
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 727.3125,
-      "epoch": 0.464,
-      "grad_norm": 26.696071790504657,
-      "kl": 7.0625,
-      "learning_rate": 6.83068622519821e-07,
-      "loss": 0.8028,
-      "reward": 2.3517472743988037,
-      "reward_std": 0.6738306879997253,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.02498907968401909,
-      "rewards/tag_count_reward": 0.7864583432674408,
+      "completion_length": 517.0000152587891,
+      "epoch": 0.232,
+      "grad_norm": 519611.1496287819,
+      "kl": 284.578125,
+      "learning_rate": 9.530702921077358e-07,
+      "loss": 46.5488,
+      "reward": 2.31371533870697,
+      "reward_std": 0.47075986862182617,
+      "rewards/accuracy_reward": 0.5000000111758709,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.028298573568463326,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 464
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 644.5000305175781,
-      "epoch": 0.465,
-      "grad_norm": 13.62258323479659,
-      "kl": 3.65625,
-      "learning_rate": 6.815672671252315e-07,
-      "loss": 0.8709,
-      "reward": 2.5122939348220825,
-      "reward_std": 0.6799641847610474,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.0224284203723073,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 600.5208587646484,
+      "epoch": 0.2325,
+      "grad_norm": 17.962574871609352,
+      "kl": 1.216796875,
+      "learning_rate": 9.527204627209112e-07,
+      "loss": 0.357,
+      "reward": 2.464834451675415,
+      "reward_std": 0.5079529285430908,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.07162389159202576,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 465
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 696.3125305175781,
-      "epoch": 0.466,
-      "grad_norm": 16.003897480590176,
-      "kl": 4.6328125,
-      "learning_rate": 6.800643086250121e-07,
-      "loss": 0.8121,
-      "reward": 2.034106969833374,
-      "reward_std": 0.537964329123497,
-      "rewards/accuracy_reward": 0.2500000074505806,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.021448652260005474,
-      "rewards/tag_count_reward": 0.8125000298023224,
+      "completion_length": 338.62501525878906,
+      "epoch": 0.233,
+      "grad_norm": 14.124197845323842,
+      "kl": 0.435546875,
+      "learning_rate": 9.523694065777156e-07,
+      "loss": 0.3118,
+      "reward": 2.8386720418930054,
+      "reward_std": 0.370292603969574,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.020702947862446308,
+      "rewards/tag_count_reward": 0.984375,
       "step": 466
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 834.3333740234375,
-      "epoch": 0.467,
-      "grad_norm": 78.01009343549657,
-      "kl": 9.53125,
-      "learning_rate": 6.78559765332238e-07,
-      "loss": 0.9044,
-      "reward": 1.9304699301719666,
-      "reward_std": 0.6387233734130859,
-      "rewards/accuracy_reward": 0.3125,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02786353789269924,
-      "rewards/tag_count_reward": 0.6666666865348816,
+      "completion_length": 639.2708740234375,
+      "epoch": 0.2335,
+      "grad_norm": 22.217309088723397,
+      "kl": 2.3359375,
+      "learning_rate": 9.520171247475268e-07,
+      "loss": 0.5155,
+      "reward": 2.592587947845459,
+      "reward_std": 0.3498579263687134,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.0410926416516304,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 467
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 653.5000305175781,
-      "epoch": 0.468,
-      "grad_norm": 26.98679371952424,
-      "kl": 4.2109375,
-      "learning_rate": 6.770536555792944e-07,
-      "loss": 0.8219,
-      "reward": 2.4220420122146606,
-      "reward_std": 0.6534742116928101,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.018930173479020596,
-      "rewards/tag_count_reward": 0.84375,
+      "completion_length": 587.0208435058594,
+      "epoch": 0.234,
+      "grad_norm": 26.970069655903107,
+      "kl": 1.96484375,
+      "learning_rate": 9.516636183034564e-07,
+      "loss": 0.4617,
+      "reward": 2.5388882160186768,
+      "reward_std": 0.5101897567510605,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.02882016822695732,
+      "rewards/tag_count_reward": 0.921875,
       "step": 468
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 627.3750305175781,
-      "epoch": 0.469,
-      "grad_norm": 149.59812362533228,
-      "kl": 6.53125,
-      "learning_rate": 6.755459977176532e-07,
-      "loss": 0.8048,
-      "reward": 2.1985827684402466,
-      "reward_std": 0.603963702917099,
-      "rewards/accuracy_reward": 0.4375,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.03405630309134722,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 551.1666870117188,
+      "epoch": 0.2345,
+      "grad_norm": 10.814287456638,
+      "kl": 1.375,
+      "learning_rate": 9.513088883223463e-07,
+      "loss": 0.3882,
+      "reward": 2.5434218645095825,
+      "reward_std": 0.49620410799980164,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889849185944,
+      "rewards/repetition_penalty_reward": -0.03817532956600189,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 469
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 470.20835876464844,
-      "epoch": 0.47,
-      "grad_norm": 17.189852933380237,
-      "kl": 1.8984375,
-      "learning_rate": 6.740368101176495e-07,
-      "loss": 0.2061,
-      "reward": 2.4457980394363403,
-      "reward_std": 0.44828473031520844,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.01774365920573473,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 476.70835876464844,
+      "epoch": 0.235,
+      "grad_norm": 15.36470721622306,
+      "kl": 1.408203125,
+      "learning_rate": 9.509529358847654e-07,
+      "loss": 0.2803,
+      "reward": 2.6564310789108276,
+      "reward_std": 0.321887843310833,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.024124516174197197,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 470
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 547.5625,
-      "epoch": 0.471,
-      "grad_norm": 15.50803120857332,
-      "kl": 2.32421875,
-      "learning_rate": 6.725261111682584e-07,
-      "loss": 0.4077,
-      "reward": 2.327059745788574,
-      "reward_std": 0.5171791017055511,
-      "rewards/accuracy_reward": 0.4791666716337204,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.023634711280465126,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 579.4791717529297,
+      "epoch": 0.2355,
+      "grad_norm": 26.432404929557944,
+      "kl": 1.341796875,
+      "learning_rate": 9.505957620750069e-07,
+      "loss": 0.3973,
+      "reward": 2.486717700958252,
+      "reward_std": 0.4216475263237953,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.05494917556643486,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 471
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 642.6666870117188,
-      "epoch": 0.472,
-      "grad_norm": 21.870962603035192,
-      "kl": 3.46875,
-      "learning_rate": 6.710139192768694e-07,
-      "loss": 0.7413,
-      "reward": 2.383070707321167,
-      "reward_std": 0.5808857977390289,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02838774584233761,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 638.1666870117188,
+      "epoch": 0.236,
+      "grad_norm": 34.94887870173885,
+      "kl": 1.388671875,
+      "learning_rate": 9.502373679810839e-07,
+      "loss": 0.5501,
+      "reward": 2.225524663925171,
+      "reward_std": 0.6959312558174133,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.05225317180156708,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 472
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 642.8958435058594,
-      "epoch": 0.473,
-      "grad_norm": 10.842744450293143,
-      "kl": 3.8203125,
-      "learning_rate": 6.695002528690639e-07,
-      "loss": 0.6264,
-      "reward": 2.4516608715057373,
-      "reward_std": 0.5648539513349533,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.024033674970269203,
-      "rewards/tag_count_reward": 0.8020833432674408,
+      "completion_length": 379.0416717529297,
+      "epoch": 0.2365,
+      "grad_norm": 5.72045110026632,
+      "kl": 0.39453125,
+      "learning_rate": 9.49877754694727e-07,
+      "loss": -0.0209,
+      "reward": 2.2306947112083435,
+      "reward_std": 0.19987352192401886,
+      "rewards/accuracy_reward": 0.29166667722165585,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.027985903434455395,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 473
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 724.2708435058594,
-      "epoch": 0.474,
-      "grad_norm": 29.595685417395913,
-      "kl": 5.6640625,
-      "learning_rate": 6.679851303883891e-07,
-      "loss": 0.4416,
-      "reward": 2.2453513741493225,
-      "reward_std": 0.5234881341457367,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.020273788832128048,
-      "rewards/tag_count_reward": 0.7656250298023224,
+      "completion_length": 407.0833435058594,
+      "epoch": 0.237,
+      "grad_norm": 32.792554292092795,
+      "kl": 0.365234375,
+      "learning_rate": 9.495169233113806e-07,
+      "loss": 0.1724,
+      "reward": 2.5822925567626953,
+      "reward_std": 0.33261267840862274,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.8680555820465088,
+      "rewards/repetition_penalty_reward": -0.061804771423339844,
+      "rewards/tag_count_reward": 0.984375,
       "step": 474
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 648.6458435058594,
-      "epoch": 0.475,
-      "grad_norm": 12.791295320250482,
-      "kl": 4.8125,
-      "learning_rate": 6.664685702961344e-07,
-      "loss": 0.6559,
-      "reward": 2.5282609462738037,
-      "reward_std": 0.6692875623703003,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.030766917392611504,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 486.6458435058594,
+      "epoch": 0.2375,
+      "grad_norm": 74.12575189964451,
+      "kl": 1.287109375,
+      "learning_rate": 9.491548749301997e-07,
+      "loss": 0.4262,
+      "reward": 2.6630338430404663,
+      "reward_std": 0.4112485349178314,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.07134135626256466,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 475
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 548.0833435058594,
-      "epoch": 0.476,
-      "grad_norm": 14.816045187511754,
-      "kl": 2.8046875,
-      "learning_rate": 6.649505910711058e-07,
-      "loss": 0.4964,
-      "reward": 2.352592945098877,
-      "reward_std": 0.5124609172344208,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.03282391466200352,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 404.4375,
+      "epoch": 0.238,
+      "grad_norm": 18.64237541974552,
+      "kl": 0.95751953125,
+      "learning_rate": 9.487916106540465e-07,
+      "loss": 0.2883,
+      "reward": 2.847753882408142,
+      "reward_std": 0.25422170013189316,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.05676013603806496,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 476
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 653.0208435058594,
-      "epoch": 0.477,
-      "grad_norm": 17.711933807676367,
-      "kl": 4.5625,
-      "learning_rate": 6.634312112094013e-07,
-      "loss": 0.5825,
-      "reward": 2.517646074295044,
-      "reward_std": 0.46930187940597534,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.020548363216221333,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 495.93751525878906,
+      "epoch": 0.2385,
+      "grad_norm": 528.7163050198875,
+      "kl": 15.3125,
+      "learning_rate": 9.484271315894871e-07,
+      "loss": 0.9612,
+      "reward": 2.6119762659072876,
+      "reward_std": 0.4331580549478531,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.04427382908761501,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 477
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 569.8541870117188,
-      "epoch": 0.478,
-      "grad_norm": 15.060185969581552,
-      "kl": 2.9296875,
-      "learning_rate": 6.619104492241847e-07,
-      "loss": 0.6103,
-      "reward": 2.419161558151245,
-      "reward_std": 0.5570693910121918,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.01833842322230339,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 383.9375,
+      "epoch": 0.239,
+      "grad_norm": 132.3126145140169,
+      "kl": 3.9453125,
+      "learning_rate": 9.480614388467877e-07,
+      "loss": 0.2224,
+      "reward": 2.751030921936035,
+      "reward_std": 0.23816969990730286,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.047580257058143616,
+      "rewards/tag_count_reward": 1.0,
       "step": 478
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 820.8541870117188,
-      "epoch": 0.479,
-      "grad_norm": 67.25826820013505,
-      "kl": 5.484375,
-      "learning_rate": 6.603883236454612e-07,
-      "loss": 0.759,
-      "reward": 2.107418715953827,
-      "reward_std": 0.7182769775390625,
-      "rewards/accuracy_reward": 0.4375000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.050567377358675,
-      "rewards/tag_count_reward": 0.7343750298023224,
+      "completion_length": 376.375,
+      "epoch": 0.2395,
+      "grad_norm": 44.2595079452565,
+      "kl": 0.7900390625,
+      "learning_rate": 9.47694533539912e-07,
+      "loss": 0.2293,
+      "reward": 2.831836700439453,
+      "reward_std": 0.23659071326255798,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.04142730962485075,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 479
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 534.2291870117188,
-      "epoch": 0.48,
-      "grad_norm": 19.548359681394896,
-      "kl": 2.03515625,
-      "learning_rate": 6.588648530198504e-07,
-      "loss": 0.5902,
-      "reward": 2.236539840698242,
-      "reward_std": 0.47467684745788574,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.03082139603793621,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 381.4583435058594,
+      "epoch": 0.24,
+      "grad_norm": 5.3725546070452115,
+      "kl": 0.3857421875,
+      "learning_rate": 9.473264167865171e-07,
+      "loss": 0.1157,
+      "reward": 2.626276731491089,
+      "reward_std": 0.19565383344888687,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9097222089767456,
+      "rewards/repetition_penalty_reward": -0.05427892133593559,
+      "rewards/tag_count_reward": 1.0,
       "step": 480
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 557.5833587646484,
-      "epoch": 0.481,
-      "grad_norm": 27.83836565318464,
-      "kl": 2.5,
-      "learning_rate": 6.573400559103613e-07,
-      "loss": 0.6866,
-      "reward": 2.6256160736083984,
-      "reward_std": 0.6559298038482666,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.018481258302927017,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 394.4166717529297,
+      "epoch": 0.2405,
+      "grad_norm": 17.822823313036796,
+      "kl": 0.291015625,
+      "learning_rate": 9.469570897079504e-07,
+      "loss": 0.1667,
+      "reward": 2.8159937858581543,
+      "reward_std": 0.23275446519255638,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.036437010392546654,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 481
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 544.4791870117188,
-      "epoch": 0.482,
-      "grad_norm": 16.75197360847501,
-      "kl": 2.359375,
-      "learning_rate": 6.558139508961654e-07,
-      "loss": 0.4691,
-      "reward": 2.4990181922912598,
-      "reward_std": 0.5695154368877411,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02528748568147421,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 494.6250305175781,
+      "epoch": 0.241,
+      "grad_norm": 37.12601936582567,
+      "kl": 0.4287109375,
+      "learning_rate": 9.465865534292464e-07,
+      "loss": 0.3572,
+      "reward": 2.4508873224258423,
+      "reward_std": 0.49121496081352234,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.06300175003707409,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 482
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 513.3125152587891,
-      "epoch": 0.483,
-      "grad_norm": 11.150322642215151,
-      "kl": 3.21875,
-      "learning_rate": 6.542865565723707e-07,
-      "loss": 0.6304,
-      "reward": 2.340551257133484,
-      "reward_std": 0.5843985080718994,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.9513889849185944,
-      "rewards/repetition_penalty_reward": -0.02229604311287403,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 472.6875,
+      "epoch": 0.2415,
+      "grad_norm": 17.990973842263713,
+      "kl": 0.3642578125,
+      "learning_rate": 9.462148090791228e-07,
+      "loss": 0.0626,
+      "reward": 2.607269763946533,
+      "reward_std": 0.3627200424671173,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666269302368,
+      "rewards/repetition_penalty_reward": -0.08543861098587513,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 483
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 555.0,
-      "epoch": 0.484,
-      "grad_norm": 25.161999743535905,
-      "kl": 4.7734375,
-      "learning_rate": 6.527578915497951e-07,
-      "loss": 0.7346,
-      "reward": 2.534891724586487,
-      "reward_std": 0.6275463998317719,
-      "rewards/accuracy_reward": 0.7708333730697632,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.029344591312110424,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 348.5625,
+      "epoch": 0.242,
+      "grad_norm": 3.9182155917466623,
+      "kl": 0.2919921875,
+      "learning_rate": 9.458418577899774e-07,
+      "loss": 0.0877,
+      "reward": 2.6882437467575073,
+      "reward_std": 0.4514298141002655,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.05828405171632767,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 484
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 780.125,
-      "epoch": 0.485,
-      "grad_norm": 51.74758202769363,
-      "kl": 10.09375,
-      "learning_rate": 6.512279744547392e-07,
-      "loss": 0.9894,
-      "reward": 1.9237075448036194,
-      "reward_std": 0.6222628057003021,
-      "rewards/accuracy_reward": 0.2708333432674408,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.01726479548960924,
-      "rewards/tag_count_reward": 0.6979166865348816,
+      "completion_length": 449.2916717529297,
+      "epoch": 0.2425,
+      "grad_norm": 25.86980665658515,
+      "kl": 0.513671875,
+      "learning_rate": 9.454677006978842e-07,
+      "loss": 0.2189,
+      "reward": 2.6299872398376465,
+      "reward_std": 0.3183176666498184,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.07834616675972939,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 485
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 694.1875305175781,
-      "epoch": 0.486,
-      "grad_norm": 26.567315708037814,
-      "kl": 5.890625,
-      "learning_rate": 6.496968239287603e-07,
-      "loss": 0.7402,
-      "reward": 2.210293173789978,
-      "reward_std": 0.7222527861595154,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02060973085463047,
-      "rewards/tag_count_reward": 0.7656250298023224,
+      "completion_length": 400.9583435058594,
+      "epoch": 0.243,
+      "grad_norm": 5.7429495684505,
+      "kl": 1.07421875,
+      "learning_rate": 9.450923389425911e-07,
+      "loss": 0.2514,
+      "reward": 2.870607614517212,
+      "reward_std": 0.19876050017774105,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.05647587403655052,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 486
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 701.7291870117188,
-      "epoch": 0.487,
-      "grad_norm": 23.60086720039633,
-      "kl": 5.703125,
-      "learning_rate": 6.481644586284442e-07,
-      "loss": 0.8466,
-      "reward": 2.353806734085083,
-      "reward_std": 0.6038880944252014,
-      "rewards/accuracy_reward": 0.583333358168602,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.03334605507552624,
-      "rewards/tag_count_reward": 0.8177083432674408,
+      "completion_length": 434.25001525878906,
+      "epoch": 0.2435,
+      "grad_norm": 13.342575140209078,
+      "kl": 0.84912109375,
+      "learning_rate": 9.44715773667515e-07,
+      "loss": 0.2413,
+      "reward": 2.700625419616699,
+      "reward_std": 0.31798748672008514,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.06673579290509224,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 487
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 748.4583435058594,
-      "epoch": 0.488,
-      "grad_norm": 19.03851418441834,
-      "kl": 6.2734375,
-      "learning_rate": 6.466308972251785e-07,
-      "loss": 1.0066,
-      "reward": 2.5121508836746216,
-      "reward_std": 0.7225759625434875,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.038196416571736336,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 322.5,
+      "epoch": 0.244,
+      "grad_norm": 5.343676568554896,
+      "kl": 0.31591796875,
+      "learning_rate": 9.443380060197385e-07,
+      "loss": 0.0725,
+      "reward": 2.9111135005950928,
+      "reward_std": 0.15393263846635818,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9722221791744232,
+      "rewards/repetition_penalty_reward": -0.03506710007786751,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 488
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.0416870117188,
-      "epoch": 0.489,
-      "grad_norm": 12.839961288051784,
-      "kl": 3.6796875,
-      "learning_rate": 6.45096158404925e-07,
-      "loss": 0.7412,
-      "reward": 2.5207144021987915,
-      "reward_std": 0.6570230424404144,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.015743907541036606,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 386.1458435058594,
+      "epoch": 0.2445,
+      "grad_norm": 5.09132953467604,
+      "kl": 0.3408203125,
+      "learning_rate": 9.43959037150008e-07,
+      "loss": 0.0758,
+      "reward": 2.8277556896209717,
+      "reward_std": 0.2439712956547737,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.06286950036883354,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 489
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 594.1666870117188,
-      "epoch": 0.49,
-      "grad_norm": 11.918218758688853,
-      "kl": 4.0390625,
-      "learning_rate": 6.435602608679916e-07,
-      "loss": 0.5896,
-      "reward": 2.576896905899048,
-      "reward_std": 0.6758164465427399,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.018589303828775883,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 448.2916717529297,
+      "epoch": 0.245,
+      "grad_norm": 23.327613224213003,
+      "kl": 1.8515625,
+      "learning_rate": 9.43578868212728e-07,
+      "loss": 0.2664,
+      "reward": 2.1128222346305847,
+      "reward_std": 0.39175570011138916,
+      "rewards/accuracy_reward": 0.2916666716337204,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.03821949101984501,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 490
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 531.25,
-      "epoch": 0.491,
-      "grad_norm": 12.572669494859374,
-      "kl": 4.7421875,
-      "learning_rate": 6.420232233288055e-07,
-      "loss": 0.4827,
-      "reward": 2.256177067756653,
-      "reward_std": 0.543413519859314,
-      "rewards/accuracy_reward": 0.4375000223517418,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.01639253133907914,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 404.3958435058594,
+      "epoch": 0.2455,
+      "grad_norm": 6.480996990735909,
+      "kl": 0.669921875,
+      "learning_rate": 9.431975003659594e-07,
+      "loss": 0.1923,
+      "reward": 2.416967749595642,
+      "reward_std": 0.4523526728153229,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.044837912544608116,
+      "rewards/tag_count_reward": 0.96875,
       "step": 491
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 400.9375,
-      "epoch": 0.492,
-      "grad_norm": 25.196826952891467,
-      "kl": 1.765625,
-      "learning_rate": 6.404850645156841e-07,
-      "loss": 0.3313,
-      "reward": 2.747916102409363,
-      "reward_std": 0.490363210439682,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.022917456924915314,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 474.22918701171875,
+      "epoch": 0.246,
+      "grad_norm": 17.872104813321265,
+      "kl": 2.296875,
+      "learning_rate": 9.428149347714143e-07,
+      "loss": 0.3238,
+      "reward": 2.4794113636016846,
+      "reward_std": 0.3524549901485443,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9513889849185944,
+      "rewards/repetition_penalty_reward": -0.05531090311706066,
+      "rewards/tag_count_reward": 0.9375,
       "step": 492
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 536.6458435058594,
-      "epoch": 0.493,
-      "grad_norm": 12.14349711767654,
-      "kl": 2.984375,
-      "learning_rate": 6.389458031706068e-07,
-      "loss": 0.5739,
-      "reward": 2.6953213214874268,
-      "reward_std": 0.5101250112056732,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.026901046745479107,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 429.5625305175781,
+      "epoch": 0.2465,
+      "grad_norm": 25.50926751724998,
+      "kl": 2.3203125,
+      "learning_rate": 9.424311725944543e-07,
+      "loss": 0.3623,
+      "reward": 2.5525494813919067,
+      "reward_std": 0.40840910375118256,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.037728333845734596,
+      "rewards/tag_count_reward": 0.9583333730697632,
       "step": 493
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 399.7083435058594,
-      "epoch": 0.494,
-      "grad_norm": 19.752623374452337,
-      "kl": 2.17578125,
-      "learning_rate": 6.374054580489873e-07,
-      "loss": 0.1944,
-      "reward": 2.5966198444366455,
-      "reward_std": 0.4636584371328354,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02317184768617153,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 353.12501525878906,
+      "epoch": 0.247,
+      "grad_norm": 5.9043836446965186,
+      "kl": 0.50390625,
+      "learning_rate": 9.420462150040852e-07,
+      "loss": 0.0674,
+      "reward": 2.6686885356903076,
+      "reward_std": 0.2424444481730461,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.0465893279761076,
+      "rewards/tag_count_reward": 1.0,
       "step": 494
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 465.91668701171875,
-      "epoch": 0.495,
-      "grad_norm": 14.16827974955118,
-      "kl": 1.6328125,
-      "learning_rate": 6.358640479194451e-07,
-      "loss": 0.3459,
-      "reward": 2.6907471418380737,
-      "reward_std": 0.4236067831516266,
+      "completion_length": 375.7708435058594,
+      "epoch": 0.2475,
+      "grad_norm": 8.837736782220068,
+      "kl": 0.705078125,
+      "learning_rate": 9.416600631729548e-07,
+      "loss": 0.1302,
+      "reward": 2.721538782119751,
+      "reward_std": 0.13726279512047768,
       "rewards/accuracy_reward": 0.7916666865348816,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.0332112442702055,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "rewards/repetition_penalty_reward": -0.0597112700343132,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 495
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 587.9791870117188,
-      "epoch": 0.496,
-      "grad_norm": 28.995423975163437,
-      "kl": 3.40625,
-      "learning_rate": 6.343215915635761e-07,
-      "loss": 0.5617,
-      "reward": 2.512778878211975,
-      "reward_std": 0.5804081857204437,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666567325592,
-      "rewards/repetition_penalty_reward": -0.028887868858873844,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 385.7708435058594,
+      "epoch": 0.248,
+      "grad_norm": 11.405430982188452,
+      "kl": 1.044921875,
+      "learning_rate": 9.412727182773486e-07,
+      "loss": 0.201,
+      "reward": 2.7786643505096436,
+      "reward_std": 0.3927687704563141,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.03730800375342369,
+      "rewards/tag_count_reward": 0.96875,
       "step": 496
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 411.7083435058594,
-      "epoch": 0.497,
-      "grad_norm": 14.922976729616561,
-      "kl": 2.7109375,
-      "learning_rate": 6.327781077757241e-07,
-      "loss": 0.3736,
-      "reward": 2.570963978767395,
-      "reward_std": 0.578457772731781,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.019313913770020008,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 473.85418701171875,
+      "epoch": 0.2485,
+      "grad_norm": 11.676094147579299,
+      "kl": 1.74609375,
+      "learning_rate": 9.408841814971861e-07,
+      "loss": 0.3357,
+      "reward": 2.485311985015869,
+      "reward_std": 0.4508149325847626,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.8958334028720856,
+      "rewards/repetition_penalty_reward": -0.045938010327517986,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 497
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 570.1875,
-      "epoch": 0.498,
-      "grad_norm": 12.22816448398453,
-      "kl": 3.7890625,
-      "learning_rate": 6.31233615362752e-07,
-      "loss": 0.7994,
-      "reward": 2.4899239540100098,
-      "reward_std": 0.45974400639533997,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.0222289408557117,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 342.25001525878906,
+      "epoch": 0.249,
+      "grad_norm": 6.336940055320027,
+      "kl": 0.8125,
+      "learning_rate": 9.404944540160177e-07,
+      "loss": 0.0604,
+      "reward": 2.6701611280441284,
+      "reward_std": 0.355712890625,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.055533334612846375,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 498
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 541.2083435058594,
-      "epoch": 0.499,
-      "grad_norm": 24.07992840689996,
-      "kl": 3.2890625,
-      "learning_rate": 6.296881331438126e-07,
-      "loss": 0.6263,
-      "reward": 2.4488918781280518,
-      "reward_std": 0.6255036890506744,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.02680276893079281,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 369.79168701171875,
+      "epoch": 0.2495,
+      "grad_norm": 18.051875973959618,
+      "kl": 1.072265625,
+      "learning_rate": 9.401035370210212e-07,
+      "loss": 0.2629,
+      "reward": 2.47028648853302,
+      "reward_std": 0.44121938943862915,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.05401917174458504,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 499
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 539.4375,
-      "epoch": 0.5,
-      "grad_norm": 13.625300381502933,
-      "kl": 3.3359375,
-      "learning_rate": 6.281416799501187e-07,
-      "loss": 0.3615,
-      "reward": 2.6264405250549316,
-      "reward_std": 0.5198075622320175,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.01939287781715393,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 440.31251525878906,
+      "epoch": 0.25,
+      "grad_norm": 20.96572788542454,
+      "kl": 2.51171875,
+      "learning_rate": 9.397114317029974e-07,
+      "loss": 0.4272,
+      "reward": 2.3335598707199097,
+      "reward_std": 0.625691831111908,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.02581525407731533,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 500
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.1041870117188,
-      "epoch": 0.501,
-      "grad_norm": 14.175570746863878,
-      "kl": 4.4375,
-      "learning_rate": 6.265942746247146e-07,
-      "loss": 0.5509,
-      "reward": 2.34303081035614,
-      "reward_std": 0.5519561469554901,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.021552613005042076,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 325.9583435058594,
+      "epoch": 0.2505,
+      "grad_norm": 4.461987602762651,
+      "kl": 0.57421875,
+      "learning_rate": 9.393181392563669e-07,
+      "loss": 0.0403,
+      "reward": 2.7603014707565308,
+      "reward_std": 0.20193683356046677,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.03136515896767378,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 501
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 698.0625305175781,
-      "epoch": 0.502,
-      "grad_norm": 11.766159336622879,
-      "kl": 5.6875,
-      "learning_rate": 6.25045936022246e-07,
-      "loss": 0.9051,
-      "reward": 2.2460073232650757,
-      "reward_std": 0.7623308002948761,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.019617782905697823,
-      "rewards/tag_count_reward": 0.765625,
+      "completion_length": 325.9583435058594,
+      "epoch": 0.251,
+      "grad_norm": 3.5280127016605,
+      "kl": 0.2861328125,
+      "learning_rate": 9.38923660879167e-07,
+      "loss": 0.0561,
+      "reward": 2.8909952640533447,
+      "reward_std": 0.14273904263973236,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03435206413269043,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 502
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 680.4375305175781,
-      "epoch": 0.503,
-      "grad_norm": 29.510968471399966,
-      "kl": 7.453125,
-      "learning_rate": 6.2349668300873e-07,
-      "loss": 1.0744,
-      "reward": 2.1472885608673096,
-      "reward_std": 0.7113551497459412,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.03847557958215475,
-      "rewards/tag_count_reward": 0.7552083432674408,
+      "completion_length": 347.4583435058594,
+      "epoch": 0.2515,
+      "grad_norm": 12.53255748366457,
+      "kl": 1.65576171875,
+      "learning_rate": 9.385279977730472e-07,
+      "loss": 0.2711,
+      "reward": 2.432243227958679,
+      "reward_std": 0.06772075779736042,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.033034625463187695,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 503
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 477.1041717529297,
-      "epoch": 0.504,
-      "grad_norm": 14.23791373020675,
-      "kl": 3.9609375,
-      "learning_rate": 6.219465344613258e-07,
-      "loss": 0.6783,
-      "reward": 2.5481714010238647,
-      "reward_std": 0.5729511976242065,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.012592657934874296,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 365.0,
+      "epoch": 0.252,
+      "grad_norm": 12.118654747376015,
+      "kl": 1.572265625,
+      "learning_rate": 9.381311511432658e-07,
+      "loss": 0.1095,
+      "reward": 2.758858323097229,
+      "reward_std": 0.3722696304321289,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.025863975286483765,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 504
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 497.35418701171875,
-      "epoch": 0.505,
-      "grad_norm": 11.708756841965316,
-      "kl": 3.3359375,
-      "learning_rate": 6.203955092681039e-07,
-      "loss": 0.4639,
-      "reward": 2.479733467102051,
-      "reward_std": 0.38708843290805817,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.022002640645951033,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 332.56251525878906,
+      "epoch": 0.2525,
+      "grad_norm": 12.675358816220538,
+      "kl": 0.609375,
+      "learning_rate": 9.377331221986866e-07,
+      "loss": 0.1461,
+      "reward": 2.7681565284729004,
+      "reward_std": 0.3903198540210724,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.04087108187377453,
+      "rewards/tag_count_reward": 0.96875,
       "step": 505
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 681.0625305175781,
-      "epoch": 0.506,
-      "grad_norm": 18.2185049016805,
-      "kl": 5.5390625,
-      "learning_rate": 6.188436263278172e-07,
-      "loss": 0.7,
-      "reward": 2.064489424228668,
-      "reward_std": 0.6310321092605591,
-      "rewards/accuracy_reward": 0.3333333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.022316260263323784,
-      "rewards/tag_count_reward": 0.7812500298023224,
+      "completion_length": 366.9166717529297,
+      "epoch": 0.253,
+      "grad_norm": 8.279601199477703,
+      "kl": 0.5146484375,
+      "learning_rate": 9.373339121517746e-07,
+      "loss": 0.0372,
+      "reward": 2.770949602127075,
+      "reward_std": 0.34754781424999237,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.046758661046624184,
+      "rewards/tag_count_reward": 0.984375,
       "step": 506
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 550.75,
-      "epoch": 0.507,
-      "grad_norm": 17.7803570241784,
-      "kl": 3.546875,
-      "learning_rate": 6.172909045496694e-07,
-      "loss": 0.8123,
-      "reward": 2.2492624521255493,
-      "reward_std": 0.5181691646575928,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02330716885626316,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 474.625,
+      "epoch": 0.2535,
+      "grad_norm": 12.183520807470723,
+      "kl": 1.2109375,
+      "learning_rate": 9.36933522218593e-07,
+      "loss": 0.2878,
+      "reward": 2.3224886655807495,
+      "reward_std": 0.6696373820304871,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.026469644159078598,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 507
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 807.4375305175781,
-      "epoch": 0.508,
-      "grad_norm": 42.53879608641054,
-      "kl": 6.71875,
-      "learning_rate": 6.157373628530852e-07,
-      "loss": 0.9312,
-      "reward": 2.1224422454833984,
-      "reward_std": 0.8008854687213898,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.06505792587995529,
-      "rewards/tag_count_reward": 0.75,
+      "completion_length": 355.1458435058594,
+      "epoch": 0.254,
+      "grad_norm": 17.57821101350406,
+      "kl": 0.7890625,
+      "learning_rate": 9.36531953618799e-07,
+      "loss": 0.2123,
+      "reward": 2.774104595184326,
+      "reward_std": 0.31567811965942383,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.026242737658321857,
+      "rewards/tag_count_reward": 0.9739583730697632,
       "step": 508
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 703.625,
-      "epoch": 0.509,
-      "grad_norm": 38.46406571699496,
-      "kl": 5.875,
-      "learning_rate": 6.141830201674802e-07,
-      "loss": 1.0649,
-      "reward": 2.162044405937195,
-      "reward_std": 0.7470089793205261,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 0.9097223281860352,
-      "rewards/repetition_penalty_reward": -0.04976117052137852,
-      "rewards/tag_count_reward": 0.78125,
+      "completion_length": 477.3333435058594,
+      "epoch": 0.2545,
+      "grad_norm": 20.44793363137805,
+      "kl": 2.091796875,
+      "learning_rate": 9.361292075756401e-07,
+      "loss": 0.4205,
+      "reward": 2.6159251928329468,
+      "reward_std": 0.4438074082136154,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03338045813143253,
+      "rewards/tag_count_reward": 0.9270833730697632,
       "step": 509
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 537.3541870117188,
-      "epoch": 0.51,
-      "grad_norm": 23.200007837390707,
-      "kl": 2.9609375,
-      "learning_rate": 6.126278954320294e-07,
-      "loss": 0.8415,
-      "reward": 2.5392757654190063,
-      "reward_std": 0.7050909399986267,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9513889849185944,
-      "rewards/repetition_penalty_reward": -0.02148817665874958,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 320.56251525878906,
+      "epoch": 0.255,
+      "grad_norm": 7.717284653235476,
+      "kl": 0.5712890625,
+      "learning_rate": 9.357252853159505e-07,
+      "loss": 0.0217,
+      "reward": 2.717986583709717,
+      "reward_std": 0.2879996597766876,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03895781189203262,
+      "rewards/tag_count_reward": 1.0,
       "step": 510
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 605.9375152587891,
-      "epoch": 0.511,
-      "grad_norm": 12.217834862993339,
-      "kl": 4.390625,
-      "learning_rate": 6.11072007595437e-07,
-      "loss": 0.5492,
-      "reward": 2.281681537628174,
-      "reward_std": 0.5939317345619202,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.018665821757167578,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 503.25001525878906,
+      "epoch": 0.2555,
+      "grad_norm": 23.04840840134793,
+      "kl": 2.5546875,
+      "learning_rate": 9.353201880701477e-07,
+      "loss": 0.7477,
+      "reward": 2.436654567718506,
+      "reward_std": 0.581574410200119,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.039039863273501396,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 511
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 659.0000305175781,
-      "epoch": 0.512,
-      "grad_norm": 11.142187913544113,
-      "kl": 5.203125,
-      "learning_rate": 6.095153756157051e-07,
-      "loss": 0.8536,
-      "reward": 2.404225468635559,
-      "reward_std": 0.7626966536045074,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.01591357495635748,
-      "rewards/tag_count_reward": 0.8020833730697632,
+      "completion_length": 369.75001525878906,
+      "epoch": 0.256,
+      "grad_norm": 6.140144067454673,
+      "kl": 1.4453125,
+      "learning_rate": 9.34913917072228e-07,
+      "loss": 0.1252,
+      "reward": 2.3770639896392822,
+      "reward_std": 0.4367944300174713,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.04654715396463871,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 512
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 482.1875,
-      "epoch": 0.513,
-      "grad_norm": 27.609945341526714,
-      "kl": 2.7734375,
-      "learning_rate": 6.079580184599032e-07,
-      "loss": 0.5479,
-      "reward": 2.167789936065674,
-      "reward_std": 0.3920954018831253,
-      "rewards/accuracy_reward": 0.27083333395421505,
-      "rewards/reasoning_steps_reward": 0.9791666567325592,
-      "rewards/repetition_penalty_reward": -0.019710074178874493,
+      "completion_length": 412.6458435058594,
+      "epoch": 0.2565,
+      "grad_norm": 14.724218719110599,
+      "kl": 1.90625,
+      "learning_rate": 9.345064735597633e-07,
+      "loss": 0.2505,
+      "reward": 2.666745901107788,
+      "reward_std": 0.5124035775661469,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.04158730432391167,
       "rewards/tag_count_reward": 0.9375000298023224,
       "step": 513
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 602.0208435058594,
-      "epoch": 0.514,
-      "grad_norm": 21.276603242655675,
-      "kl": 4.65625,
-      "learning_rate": 6.06399955103937e-07,
-      "loss": 0.9838,
-      "reward": 2.233444333076477,
-      "reward_std": 0.626254141330719,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03218065481632948,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 344.0833435058594,
+      "epoch": 0.257,
+      "grad_norm": 12.147136815173782,
+      "kl": 2.13671875,
+      "learning_rate": 9.340978587738972e-07,
+      "loss": 0.2299,
+      "reward": 2.7790675163269043,
+      "reward_std": 0.28115659207105637,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.042113007977604866,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 514
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 548.7708587646484,
-      "epoch": 0.515,
-      "grad_norm": 8.99915341715262,
-      "kl": 2.984375,
-      "learning_rate": 6.048412045323164e-07,
-      "loss": 0.5247,
-      "reward": 2.4330430030822754,
-      "reward_std": 0.4966660887002945,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.02702661231160164,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 465.04168701171875,
+      "epoch": 0.2575,
+      "grad_norm": 42.65752880873406,
+      "kl": 4.375,
+      "learning_rate": 9.336880739593415e-07,
+      "loss": 0.3814,
+      "reward": 2.218036413192749,
+      "reward_std": 0.5327357053756714,
+      "rewards/accuracy_reward": 0.3333333358168602,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.021547013893723488,
+      "rewards/tag_count_reward": 0.9270833730697632,
       "step": 515
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 607.3125305175781,
-      "epoch": 0.516,
-      "grad_norm": 12.744716558501036,
-      "kl": 4.21875,
-      "learning_rate": 6.032817857379256e-07,
-      "loss": 0.6478,
-      "reward": 2.4581196308135986,
-      "reward_std": 0.674341470003128,
-      "rewards/accuracy_reward": 0.6458333730697632,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.017574850469827652,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 355.1458435058594,
+      "epoch": 0.258,
+      "grad_norm": 12.695454347179494,
+      "kl": 2.56640625,
+      "learning_rate": 9.332771203643714e-07,
+      "loss": 0.3547,
+      "reward": 2.7579511404037476,
+      "reward_std": 0.3366604894399643,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.023298936896026134,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 516
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 550.0833587646484,
-      "epoch": 0.517,
-      "grad_norm": 25.040001378142637,
-      "kl": 3.984375,
-      "learning_rate": 6.017217177217899e-07,
-      "loss": 0.8487,
-      "reward": 2.251497983932495,
-      "reward_std": 0.6527323424816132,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.015863300301134586,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 314.2708435058594,
+      "epoch": 0.2585,
+      "grad_norm": 8.577846989702975,
+      "kl": 0.6416015625,
+      "learning_rate": 9.328649992408231e-07,
+      "loss": 0.0241,
+      "reward": 2.9229161739349365,
+      "reward_std": 0.11198093183338642,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.042361509054899216,
+      "rewards/tag_count_reward": 1.0,
       "step": 517
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 413.8333435058594,
-      "epoch": 0.518,
-      "grad_norm": 25.626709792783128,
-      "kl": 2.48046875,
-      "learning_rate": 6.001610194928464e-07,
-      "loss": 0.4396,
-      "reward": 2.7058017253875732,
-      "reward_std": 0.5824707746505737,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.019892807584255934,
-      "rewards/tag_count_reward": 0.9270833730697632,
+      "completion_length": 386.1458435058594,
+      "epoch": 0.259,
+      "grad_norm": 17.4553779455926,
+      "kl": 2.1640625,
+      "learning_rate": 9.324517118440888e-07,
+      "loss": 0.4768,
+      "reward": 2.671400547027588,
+      "reward_std": 0.4989718794822693,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.02825235854834318,
+      "rewards/tag_count_reward": 0.9427083730697632,
       "step": 518
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 741.0833435058594,
-      "epoch": 0.519,
-      "grad_norm": 59.3566513287147,
-      "kl": 8.140625,
-      "learning_rate": 5.985997100677103e-07,
-      "loss": 1.4268,
-      "reward": 2.335380792617798,
-      "reward_std": 0.8169020414352417,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.03441094420850277,
-      "rewards/tag_count_reward": 0.7864583432674408,
+      "completion_length": 299.3333435058594,
+      "epoch": 0.2595,
+      "grad_norm": 11.830503067220192,
+      "kl": 1.0546875,
+      "learning_rate": 9.320372594331137e-07,
+      "loss": 0.0964,
+      "reward": 2.7533164024353027,
+      "reward_std": 0.3934124857187271,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03140602447092533,
+      "rewards/tag_count_reward": 1.0,
       "step": 519
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 692.6875,
-      "epoch": 0.52,
-      "grad_norm": 30.718619323707678,
-      "kl": 7.515625,
-      "learning_rate": 5.97037808470444e-07,
-      "loss": 0.9893,
-      "reward": 1.9578059911727905,
-      "reward_std": 0.5474497377872467,
-      "rewards/accuracy_reward": 0.2083333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.014416326768696308,
-      "rewards/tag_count_reward": 0.7916666865348816,
+      "completion_length": 347.8333435058594,
+      "epoch": 0.26,
+      "grad_norm": 6.457880743365504,
+      "kl": 0.3310546875,
+      "learning_rate": 9.316216432703916e-07,
+      "loss": -0.0572,
+      "reward": 2.5639514923095703,
+      "reward_std": 0.2705356106162071,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.04021530598402023,
+      "rewards/tag_count_reward": 1.0,
       "step": 520
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 486.4166717529297,
-      "epoch": 0.521,
-      "grad_norm": 18.357835870055244,
-      "kl": 4.421875,
-      "learning_rate": 5.954753337323259e-07,
-      "loss": 0.4358,
-      "reward": 2.387876033782959,
-      "reward_std": 0.5994940996170044,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.02184621151536703,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 315.2916717529297,
+      "epoch": 0.2605,
+      "grad_norm": 6.392279281489095,
+      "kl": 0.775390625,
+      "learning_rate": 9.312048646219617e-07,
+      "loss": 0.1502,
+      "reward": 2.8109084367752075,
+      "reward_std": 0.21814071387052536,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03978624939918518,
+      "rewards/tag_count_reward": 0.96875,
       "step": 521
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.25,
-      "epoch": 0.522,
-      "grad_norm": 23.01587053197016,
-      "kl": 4.734375,
-      "learning_rate": 5.939123048916173e-07,
-      "loss": 0.7259,
-      "reward": 2.393532395362854,
-      "reward_std": 0.7403541803359985,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.028342580422759056,
-      "rewards/tag_count_reward": 0.8177083730697632,
+      "completion_length": 343.22918701171875,
+      "epoch": 0.261,
+      "grad_norm": 6.132812824609326,
+      "kl": 0.546875,
+      "learning_rate": 9.307869247574038e-07,
+      "loss": 0.0556,
+      "reward": 2.6175711154937744,
+      "reward_std": 0.26942718029022217,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02826231624931097,
+      "rewards/tag_count_reward": 1.0,
       "step": 522
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 502.4791717529297,
-      "epoch": 0.523,
-      "grad_norm": 81.64058877460842,
-      "kl": 4.53125,
-      "learning_rate": 5.923487409933315e-07,
-      "loss": 0.9372,
-      "reward": 2.3321489691734314,
-      "reward_std": 0.47031281888484955,
-      "rewards/accuracy_reward": 0.4791666679084301,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.018545696511864662,
-      "rewards/tag_count_reward": 0.90625,
+      "completion_length": 304.93751525878906,
+      "epoch": 0.2615,
+      "grad_norm": 4.409570648820208,
+      "kl": 0.3115234375,
+      "learning_rate": 9.303678249498352e-07,
+      "loss": 0.0109,
+      "reward": 2.9457967281341553,
+      "reward_std": 0.0824959184974432,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.033369969576597214,
+      "rewards/tag_count_reward": 1.0,
       "step": 523
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 536.5416870117188,
-      "epoch": 0.524,
-      "grad_norm": 15.605773200090558,
-      "kl": 4.109375,
-      "learning_rate": 5.907846610890011e-07,
-      "loss": 0.5794,
-      "reward": 2.2709691524505615,
-      "reward_std": 0.644355833530426,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.01375302067026496,
-      "rewards/tag_count_reward": 0.7916666865348816,
+      "completion_length": 335.1458435058594,
+      "epoch": 0.262,
+      "grad_norm": 10.6806703810238,
+      "kl": 0.740234375,
+      "learning_rate": 9.299475664759068e-07,
+      "loss": 0.1467,
+      "reward": 2.5661017894744873,
+      "reward_std": 0.3743235468864441,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03632892295718193,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 524
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 655.5833740234375,
-      "epoch": 0.525,
-      "grad_norm": 13.665522547941404,
-      "kl": 5.3125,
-      "learning_rate": 5.892200842364462e-07,
-      "loss": 0.839,
-      "reward": 2.3515175580978394,
-      "reward_std": 0.7731278240680695,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.028690868988633156,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 346.7916717529297,
+      "epoch": 0.2625,
+      "grad_norm": 7.022946103666297,
+      "kl": 0.4912109375,
+      "learning_rate": 9.295261506157985e-07,
+      "loss": -0.0496,
+      "reward": 2.487242579460144,
+      "reward_std": 0.4531702846288681,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.028382405638694763,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 525
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 641.8750305175781,
-      "epoch": 0.526,
-      "grad_norm": 9.492149650000716,
-      "kl": 3.7890625,
-      "learning_rate": 5.87655029499542e-07,
-      "loss": 0.7068,
-      "reward": 2.473028063774109,
-      "reward_std": 0.6603610515594482,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.018291576765477657,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 344.6666717529297,
+      "epoch": 0.263,
+      "grad_norm": 5.783429974481017,
+      "kl": 0.4072265625,
+      "learning_rate": 9.291035786532163e-07,
+      "loss": 0.0516,
+      "reward": 2.7101194858551025,
+      "reward_std": 0.01704893447458744,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03988067805767059,
+      "rewards/tag_count_reward": 1.0,
       "step": 526
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 498.0625,
-      "epoch": 0.527,
-      "grad_norm": 9.861486616032105,
-      "kl": 3.2421875,
-      "learning_rate": 5.860895159479864e-07,
-      "loss": 0.5474,
-      "reward": 2.4325687885284424,
-      "reward_std": 0.5291797816753387,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.020556333474814892,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 356.72918701171875,
+      "epoch": 0.2635,
+      "grad_norm": 5.172841688416704,
+      "kl": 0.6357421875,
+      "learning_rate": 9.286798518753878e-07,
+      "loss": 0.0223,
+      "reward": 2.5386565923690796,
+      "reward_std": 0.35626935213804245,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.022107405588030815,
+      "rewards/tag_count_reward": 0.984375,
       "step": 527
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 562.8958587646484,
-      "epoch": 0.528,
-      "grad_norm": 11.54211464068912,
-      "kl": 4.3984375,
-      "learning_rate": 5.845235626570683e-07,
-      "loss": 0.6108,
-      "reward": 2.3765957355499268,
-      "reward_std": 0.6778049767017365,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.01229327404871583,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 335.0625,
+      "epoch": 0.264,
+      "grad_norm": 4.496157225131026,
+      "kl": 0.2734375,
+      "learning_rate": 9.282549715730579e-07,
+      "loss": 0.0545,
+      "reward": 2.803345203399658,
+      "reward_std": 0.2169586569070816,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.043877096846699715,
+      "rewards/tag_count_reward": 1.0,
       "step": 528
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 580.6041870117188,
-      "epoch": 0.529,
-      "grad_norm": 12.470353874580706,
-      "kl": 4.125,
-      "learning_rate": 5.829571887074343e-07,
-      "loss": 0.5239,
-      "reward": 2.4933345317840576,
-      "reward_std": 0.509730190038681,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666567325592,
-      "rewards/repetition_penalty_reward": -0.02229050174355507,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 417.2708435058594,
+      "epoch": 0.2645,
+      "grad_norm": 6.658335679549266,
+      "kl": 0.66796875,
+      "learning_rate": 9.278289390404859e-07,
+      "loss": 0.0364,
+      "reward": 2.3210874795913696,
+      "reward_std": 0.5226413607597351,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.05217655561864376,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 529
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 833.7917175292969,
-      "epoch": 0.53,
-      "grad_norm": 18.321092863583786,
-      "kl": 6.75,
-      "learning_rate": 5.813904131848564e-07,
-      "loss": 0.6788,
-      "reward": 1.967248022556305,
-      "reward_std": 0.6098497807979584,
-      "rewards/accuracy_reward": 0.3333333358168602,
-      "rewards/reasoning_steps_reward": 0.9375000298023224,
-      "rewards/repetition_penalty_reward": -0.0327520202845335,
-      "rewards/tag_count_reward": 0.7291666865348816,
+      "completion_length": 357.1458435058594,
+      "epoch": 0.265,
+      "grad_norm": 3.9330369236217146,
+      "kl": 0.5556640625,
+      "learning_rate": 9.274017555754407e-07,
+      "loss": 0.0323,
+      "reward": 2.767940640449524,
+      "reward_std": 0.2660471647977829,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9722221791744232,
+      "rewards/repetition_penalty_reward": -0.03240684233605862,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 530
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 697.2916870117188,
-      "epoch": 0.531,
-      "grad_norm": 19.33500544989958,
-      "kl": 5.546875,
-      "learning_rate": 5.798232551800002e-07,
-      "loss": 0.6624,
-      "reward": 2.311483144760132,
-      "reward_std": 0.6170333027839661,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.018378185108304024,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 328.41668701171875,
+      "epoch": 0.2655,
+      "grad_norm": 9.064114909150554,
+      "kl": 0.44140625,
+      "learning_rate": 9.269734224791974e-07,
+      "loss": 0.1061,
+      "reward": 2.6510632038116455,
+      "reward_std": 0.4317702651023865,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.03643686696887016,
+      "rewards/tag_count_reward": 0.9583333730697632,
       "step": 531
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 734.2916870117188,
-      "epoch": 0.532,
-      "grad_norm": 16.207591684417004,
-      "kl": 4.796875,
-      "learning_rate": 5.78255733788191e-07,
-      "loss": 0.8369,
-      "reward": 2.042443633079529,
-      "reward_std": 0.6627461314201355,
-      "rewards/accuracy_reward": 0.354166679084301,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.02873703371733427,
-      "rewards/tag_count_reward": 0.7447916865348816,
+      "completion_length": 356.25001525878906,
+      "epoch": 0.266,
+      "grad_norm": 8.929164182079782,
+      "kl": 0.3369140625,
+      "learning_rate": 9.265439410565328e-07,
+      "loss": 0.0927,
+      "reward": 2.6343066692352295,
+      "reward_std": 0.19984594732522964,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.04624889977276325,
+      "rewards/tag_count_reward": 1.0,
       "step": 532
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 670.3333435058594,
-      "epoch": 0.533,
-      "grad_norm": 17.149503798992292,
-      "kl": 5.671875,
-      "learning_rate": 5.766878681091828e-07,
-      "loss": 0.8688,
-      "reward": 1.991923749446869,
-      "reward_std": 0.7662231922149658,
-      "rewards/accuracy_reward": 0.3541666679084301,
-      "rewards/reasoning_steps_reward": 0.9305556416511536,
-      "rewards/repetition_penalty_reward": -0.04279853031039238,
-      "rewards/tag_count_reward": 0.7500000298023224,
+      "completion_length": 351.93751525878906,
+      "epoch": 0.2665,
+      "grad_norm": 12.113702687047534,
+      "kl": 0.48828125,
+      "learning_rate": 9.261133126157217e-07,
+      "loss": 0.1091,
+      "reward": 2.534152388572693,
+      "reward_std": 0.40387988090515137,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.05438925884664059,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 533
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 579.8125,
-      "epoch": 0.534,
-      "grad_norm": 18.63845179074038,
-      "kl": 4.59375,
-      "learning_rate": 5.751196772469237e-07,
-      "loss": 0.7095,
-      "reward": 2.073606848716736,
-      "reward_std": 0.7093684077262878,
-      "rewards/accuracy_reward": 0.375,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.023615523241460323,
-      "rewards/tag_count_reward": 0.7500000298023224,
+      "completion_length": 401.18751525878906,
+      "epoch": 0.267,
+      "grad_norm": 11.16084804643565,
+      "kl": 1.84375,
+      "learning_rate": 9.256815384685328e-07,
+      "loss": 0.3193,
+      "reward": 2.4775713682174683,
+      "reward_std": 0.46769386529922485,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03631755895912647,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 534
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 618.0,
-      "epoch": 0.535,
-      "grad_norm": 26.379982268973194,
-      "kl": 4.984375,
-      "learning_rate": 5.735511803093248e-07,
-      "loss": 0.8534,
-      "reward": 2.1693702936172485,
-      "reward_std": 0.578357994556427,
-      "rewards/accuracy_reward": 0.4375000298023224,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.0389630775898695,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "completion_length": 492.7708435058594,
+      "epoch": 0.2675,
+      "grad_norm": 31.0072808202828,
+      "kl": 3.375,
+      "learning_rate": 9.252486199302256e-07,
+      "loss": 0.4648,
+      "reward": 2.288800835609436,
+      "reward_std": 0.569076657295227,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.02890750952064991,
+      "rewards/tag_count_reward": 0.8802083432674408,
       "step": 535
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 630.9791870117188,
-      "epoch": 0.536,
-      "grad_norm": 16.107255305572483,
-      "kl": 3.328125,
-      "learning_rate": 5.71982396408026e-07,
-      "loss": 0.6713,
-      "reward": 2.3328845500946045,
-      "reward_std": 0.762814462184906,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.024754411540925503,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 418.3541717529297,
+      "epoch": 0.268,
+      "grad_norm": 21.076285472286582,
+      "kl": 2.7109375,
+      "learning_rate": 9.248145583195447e-07,
+      "loss": 0.3675,
+      "reward": 2.3963736295700073,
+      "reward_std": 0.5206663012504578,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.05501541867852211,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 536
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 644.5416870117188,
-      "epoch": 0.537,
-      "grad_norm": 9.431209455223057,
-      "kl": 4.125,
-      "learning_rate": 5.704133446581642e-07,
-      "loss": 0.5649,
-      "reward": 2.1833406686782837,
-      "reward_std": 0.6518435776233673,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.023256704211235046,
-      "rewards/tag_count_reward": 0.796875,
-      "step": 537
-    },
+      "completion_length": 446.3958435058594,
+      "epoch": 0.2685,
+      "grad_norm": 21.190596871719485,
+      "kl": 3.33203125,
+      "learning_rate": 9.243793549587171e-07,
+      "loss": 0.4056,
+      "reward": 2.491150140762329,
+      "reward_std": 0.5588173568248749,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.027947167865931988,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 537
+    },
     {
       "clip_ratio": 0.0,
-      "completion_length": 521.5625,
-      "epoch": 0.538,
-      "grad_norm": 13.691614334173481,
-      "kl": 2.828125,
-      "learning_rate": 5.688440441781398e-07,
-      "loss": 0.532,
-      "reward": 2.5351574420928955,
-      "reward_std": 0.5137946009635925,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02213435433804989,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 453.68751525878906,
+      "epoch": 0.269,
+      "grad_norm": 15.39662311115678,
+      "kl": 3.171875,
+      "learning_rate": 9.239430111734476e-07,
+      "loss": 0.3442,
+      "reward": 2.3229269981384277,
+      "reward_std": 0.5908633470535278,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.0347119364887476,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 538
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 504.20835876464844,
-      "epoch": 0.539,
-      "grad_norm": 15.514611157640976,
-      "kl": 3.296875,
-      "learning_rate": 5.672745140893839e-07,
-      "loss": 0.4842,
-      "reward": 2.3555127382278442,
-      "reward_std": 0.6762471795082092,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.017751268576830626,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 402.1458435058594,
+      "epoch": 0.2695,
+      "grad_norm": 16.818966771875186,
+      "kl": 1.31640625,
+      "learning_rate": 9.235055282929153e-07,
+      "loss": 0.3005,
+      "reward": 2.69978666305542,
+      "reward_std": 0.5072591304779053,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.025908033829182386,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 539
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 500.2083435058594,
-      "epoch": 0.54,
-      "grad_norm": 33.71631385555809,
-      "kl": 4.0390625,
-      "learning_rate": 5.657047735161255e-07,
-      "loss": 0.5623,
-      "reward": 2.4260315895080566,
-      "reward_std": 0.7543806433677673,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.02014908567070961,
-      "rewards/tag_count_reward": 0.8489583730697632,
+      "completion_length": 387.8125,
+      "epoch": 0.27,
+      "grad_norm": 12.06656265047348,
+      "kl": 0.79296875,
+      "learning_rate": 9.230669076497687e-07,
+      "loss": 0.2646,
+      "reward": 2.6891995668411255,
+      "reward_std": 0.4540497958660126,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9722223281860352,
+      "rewards/repetition_penalty_reward": -0.038231078535318375,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 540
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 645.2083435058594,
-      "epoch": 0.541,
-      "grad_norm": 18.05553096101935,
-      "kl": 5.75,
-      "learning_rate": 5.641348415851577e-07,
-      "loss": 0.9497,
-      "reward": 2.2081546783447266,
-      "reward_std": 0.5377081632614136,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.017539918422698975,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 411.1458435058594,
+      "epoch": 0.2705,
+      "grad_norm": 10.66189870366724,
+      "kl": 0.7666015625,
+      "learning_rate": 9.226271505801224e-07,
+      "loss": 0.2545,
+      "reward": 2.7011055946350098,
+      "reward_std": 0.16866168193519115,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.05410290136933327,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 541
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 739.2291870117188,
-      "epoch": 0.542,
-      "grad_norm": 16.273564979773667,
-      "kl": 7.109375,
-      "learning_rate": 5.625647374256061e-07,
-      "loss": 0.902,
-      "reward": 2.1837064027786255,
-      "reward_std": 0.6955267190933228,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.014210469089448452,
-      "rewards/tag_count_reward": 0.71875,
+      "completion_length": 433.5625,
+      "epoch": 0.271,
+      "grad_norm": 8.633799282198257,
+      "kl": 1.1171875,
+      "learning_rate": 9.221862584235526e-07,
+      "loss": 0.1229,
+      "reward": 2.806559443473816,
+      "reward_std": 0.2913419157266617,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.035454532131552696,
+      "rewards/tag_count_reward": 0.9739583730697632,
       "step": 542
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 586.3125,
-      "epoch": 0.543,
-      "grad_norm": 13.748732846527533,
-      "kl": 3.7109375,
-      "learning_rate": 5.60994480168694e-07,
-      "loss": 0.8548,
-      "reward": 2.5483158826828003,
-      "reward_std": 0.593627005815506,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.024600773118436337,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 450.375,
+      "epoch": 0.2715,
+      "grad_norm": 12.521383665249783,
+      "kl": 2.5703125,
+      "learning_rate": 9.217442325230936e-07,
+      "loss": 0.3894,
+      "reward": 2.494984745979309,
+      "reward_std": 0.5038753598928452,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03626537322998047,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 543
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 609.5416870117188,
-      "epoch": 0.544,
-      "grad_norm": 17.36578790885452,
-      "kl": 6.140625,
-      "learning_rate": 5.594240889475106e-07,
-      "loss": 1.0892,
-      "reward": 2.431919813156128,
-      "reward_std": 0.8149698972702026,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.8958333730697632,
-      "rewards/repetition_penalty_reward": -0.021205355413258076,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 462.29168701171875,
+      "epoch": 0.272,
+      "grad_norm": 13.38780899364238,
+      "kl": 2.8515625,
+      "learning_rate": 9.213010742252327e-07,
+      "loss": 0.4016,
+      "reward": 2.420200288295746,
+      "reward_std": 0.4867652505636215,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.930555522441864,
+      "rewards/repetition_penalty_reward": -0.025980187579989433,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 544
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.9791870117188,
-      "epoch": 0.545,
-      "grad_norm": 30.417556701818675,
-      "kl": 2.8984375,
-      "learning_rate": 5.578535828967777e-07,
-      "loss": 0.8559,
-      "reward": 2.590558648109436,
-      "reward_std": 0.6466452777385712,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.030969200655817986,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 420.22918701171875,
+      "epoch": 0.2725,
+      "grad_norm": 29.596911334695022,
+      "kl": 2.2265625,
+      "learning_rate": 9.208567848799069e-07,
+      "loss": 0.4394,
+      "reward": 2.488977313041687,
+      "reward_std": 0.3840087577700615,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.040536582469940186,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 545
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 633.8958435058594,
-      "epoch": 0.546,
-      "grad_norm": 15.846035378270587,
-      "kl": 6.0234375,
-      "learning_rate": 5.562829811526154e-07,
-      "loss": 1.0087,
-      "reward": 2.4130187034606934,
-      "reward_std": 0.680559515953064,
-      "rewards/accuracy_reward": 0.6666666716337204,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.01580093428492546,
-      "rewards/tag_count_reward": 0.8177083432674408,
+      "completion_length": 510.3541717529297,
+      "epoch": 0.273,
+      "grad_norm": 26.31735095796572,
+      "kl": 4.09375,
+      "learning_rate": 9.204113658404989e-07,
+      "loss": 0.706,
+      "reward": 2.371821641921997,
+      "reward_std": 0.6036520600318909,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.02574794925749302,
+      "rewards/tag_count_reward": 0.8906250298023224,
       "step": 546
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 432.8958435058594,
-      "epoch": 0.547,
-      "grad_norm": 11.403437149970372,
-      "kl": 1.345703125,
-      "learning_rate": 5.547123028523106e-07,
-      "loss": 0.0869,
-      "reward": 2.532930016517639,
-      "reward_std": 0.4439847320318222,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.03130619879812002,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 454.18751525878906,
+      "epoch": 0.2735,
+      "grad_norm": 30.066527015445274,
+      "kl": 4.25,
+      "learning_rate": 9.199648184638318e-07,
+      "loss": 0.3264,
+      "reward": 2.3577940464019775,
+      "reward_std": 0.4556938707828522,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.05019213631749153,
+      "rewards/tag_count_reward": 0.8802083432674408,
       "step": 547
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 646.3750305175781,
-      "epoch": 0.548,
-      "grad_norm": 67.38569099061115,
-      "kl": 7.3125,
-      "learning_rate": 5.531415671340826e-07,
-      "loss": 1.0126,
-      "reward": 2.4480226039886475,
-      "reward_std": 0.7663466334342957,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.02767193131148815,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 548.5,
+      "epoch": 0.274,
+      "grad_norm": 12.579666635260093,
+      "kl": 2.2412109375,
+      "learning_rate": 9.195171441101668e-07,
+      "loss": 0.2961,
+      "reward": 2.1001862287521362,
+      "reward_std": 0.2963993363082409,
+      "rewards/accuracy_reward": 0.2291666716337204,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.030022156424820423,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 548
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 534.4166717529297,
-      "epoch": 0.549,
-      "grad_norm": 15.126469736221692,
-      "kl": 3.61328125,
-      "learning_rate": 5.515707931368507e-07,
-      "loss": 0.4266,
-      "reward": 2.385190963745117,
-      "reward_std": 0.47373223304748535,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.01932289730757475,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 341.3333435058594,
+      "epoch": 0.2745,
+      "grad_norm": 6.4673055298014965,
+      "kl": 0.5400390625,
+      "learning_rate": 9.190683441431974e-07,
+      "loss": 0.0892,
+      "reward": 2.7759108543395996,
+      "reward_std": 0.3490236699581146,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.047005822882056236,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 549
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 622.2916870117188,
-      "epoch": 0.55,
-      "grad_norm": 16.89430814413906,
-      "kl": 5.453125,
-      "learning_rate": 5.5e-07,
-      "loss": 0.6763,
-      "reward": 2.388463854789734,
-      "reward_std": 0.5397979617118835,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.017786113545298576,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 361.62501525878906,
+      "epoch": 0.275,
+      "grad_norm": 12.47875427754187,
+      "kl": 0.83984375,
+      "learning_rate": 9.186184199300463e-07,
+      "loss": 0.082,
+      "reward": 2.453283429145813,
+      "reward_std": 0.3730267733335495,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.04498073272407055,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 550
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 533.1666717529297,
-      "epoch": 0.551,
-      "grad_norm": 13.91558589732855,
-      "kl": 4.0078125,
-      "learning_rate": 5.484292068631494e-07,
-      "loss": 0.6629,
-      "reward": 2.265714168548584,
-      "reward_std": 0.6671868860721588,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.020744211971759796,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 372.97918701171875,
+      "epoch": 0.2755,
+      "grad_norm": 6.003413536757677,
+      "kl": 0.533203125,
+      "learning_rate": 9.181673728412605e-07,
+      "loss": 0.0929,
+      "reward": 2.6774885654449463,
+      "reward_std": 0.32011160254478455,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.06556697562336922,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 551
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 573.5625305175781,
-      "epoch": 0.552,
-      "grad_norm": 16.883404148942905,
-      "kl": 5.0,
-      "learning_rate": 5.468584328659172e-07,
-      "loss": 0.8272,
-      "reward": 2.377629041671753,
-      "reward_std": 0.6626934707164764,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.021676572039723396,
-      "rewards/tag_count_reward": 0.84375,
+      "completion_length": 371.2083435058594,
+      "epoch": 0.276,
+      "grad_norm": 11.901672638559532,
+      "kl": 0.51806640625,
+      "learning_rate": 9.177152042508077e-07,
+      "loss": 0.0808,
+      "reward": 2.7159184217453003,
+      "reward_std": 0.22979847341775894,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.05317908897995949,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 552
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 641.9583435058594,
-      "epoch": 0.553,
-      "grad_norm": 12.510249676603848,
-      "kl": 6.515625,
-      "learning_rate": 5.452876971476896e-07,
-      "loss": 1.1594,
-      "reward": 2.224699914455414,
-      "reward_std": 0.7095433175563812,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.020091742277145386,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 487.4791717529297,
+      "epoch": 0.2765,
+      "grad_norm": 15.136748639101315,
+      "kl": 1.328125,
+      "learning_rate": 9.17261915536072e-07,
+      "loss": 0.334,
+      "reward": 2.423860192298889,
+      "reward_std": 0.5218981206417084,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.04836214520037174,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 553
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 621.0000305175781,
-      "epoch": 0.554,
-      "grad_norm": 20.15924923331073,
-      "kl": 4.90625,
-      "learning_rate": 5.437170188473847e-07,
-      "loss": 1.009,
-      "reward": 2.4554017782211304,
-      "reward_std": 0.6323766112327576,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.025501138530671597,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 461.7291717529297,
+      "epoch": 0.277,
+      "grad_norm": 26.981007155408008,
+      "kl": 1.1640625,
+      "learning_rate": 9.168075080778494e-07,
+      "loss": 0.5053,
+      "reward": 2.441664218902588,
+      "reward_std": 0.5236180424690247,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.04791930224746466,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 554
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 516.3125305175781,
-      "epoch": 0.555,
-      "grad_norm": 15.439703900925965,
-      "kl": 3.703125,
-      "learning_rate": 5.421464171032224e-07,
-      "loss": 0.6462,
-      "reward": 2.5977327823638916,
-      "reward_std": 0.5676628202199936,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.016850699670612812,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 317.31251525878906,
+      "epoch": 0.2775,
+      "grad_norm": 11.002217558072193,
+      "kl": 0.91796875,
+      "learning_rate": 9.163519832603436e-07,
+      "loss": 0.147,
+      "reward": 2.499881386756897,
+      "reward_std": 0.4323354959487915,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.03831310383975506,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 555
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 483.5625,
-      "epoch": 0.556,
-      "grad_norm": 11.490664107960336,
-      "kl": 3.9921875,
-      "learning_rate": 5.405759110524894e-07,
-      "loss": 0.5193,
-      "reward": 2.5783601999282837,
-      "reward_std": 0.6397126764059067,
-      "rewards/accuracy_reward": 0.7708333730697632,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.0292786480858922,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 379.5208435058594,
+      "epoch": 0.278,
+      "grad_norm": 12.09872643997718,
+      "kl": 1.6318359375,
+      "learning_rate": 9.158953424711624e-07,
+      "loss": 0.1784,
+      "reward": 2.6038191318511963,
+      "reward_std": 0.46002739667892456,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.04027818236500025,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 556
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 516.6041870117188,
-      "epoch": 0.557,
-      "grad_norm": 27.558157304060057,
-      "kl": 4.078125,
-      "learning_rate": 5.390055198313061e-07,
-      "loss": 0.7408,
-      "reward": 2.3512425422668457,
-      "reward_std": 0.5882950127124786,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.025493742898106575,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 318.625,
+      "epoch": 0.2785,
+      "grad_norm": 10.050531964018983,
+      "kl": 1.021484375,
+      "learning_rate": 9.154375871013128e-07,
+      "loss": 0.1225,
+      "reward": 2.6270726919174194,
+      "reward_std": 0.3049708902835846,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04306626692414284,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 557
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 540.9166870117188,
-      "epoch": 0.558,
-      "grad_norm": 22.15871149302614,
-      "kl": 5.25,
-      "learning_rate": 5.37435262574394e-07,
-      "loss": 0.7217,
-      "reward": 2.273875594139099,
-      "reward_std": 0.604903370141983,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.01605505309998989,
-      "rewards/tag_count_reward": 0.8177083432674408,
+      "completion_length": 459.52085876464844,
+      "epoch": 0.279,
+      "grad_norm": 39.99093411330565,
+      "kl": 5.265625,
+      "learning_rate": 9.149787185451969e-07,
+      "loss": 0.4026,
+      "reward": 2.456482768058777,
+      "reward_std": 0.5135739296674728,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03830893710255623,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 558
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 564.1875305175781,
-      "epoch": 0.559,
-      "grad_norm": 20.132907345221618,
-      "kl": 5.90625,
-      "learning_rate": 5.358651584148423e-07,
-      "loss": 0.975,
-      "reward": 2.391770362854004,
-      "reward_std": 0.7469497919082642,
+      "completion_length": 568.2291870117188,
+      "epoch": 0.2795,
+      "grad_norm": 28.34326913260782,
+      "kl": 4.4140625,
+      "learning_rate": 9.145187382006081e-07,
+      "loss": 0.9447,
+      "reward": 2.393509268760681,
+      "reward_std": 0.7060167789459229,
       "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.958333432674408,
-      "rewards/repetition_penalty_reward": -0.0248964074999094,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.01794920302927494,
+      "rewards/tag_count_reward": 0.8489583730697632,
       "step": 559
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 602.7500305175781,
-      "epoch": 0.56,
-      "grad_norm": 20.768809175821836,
-      "kl": 5.4296875,
-      "learning_rate": 5.342952264838747e-07,
-      "loss": 0.7266,
-      "reward": 1.9830502271652222,
-      "reward_std": 0.5142233371734619,
-      "rewards/accuracy_reward": 0.2500000111758709,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.02563036046922207,
-      "rewards/tag_count_reward": 0.7864583432674408,
+      "completion_length": 346.6666717529297,
+      "epoch": 0.28,
+      "grad_norm": 16.615012130478203,
+      "kl": 1.669921875,
+      "learning_rate": 9.140576474687263e-07,
+      "loss": 0.3664,
+      "reward": 2.7198301553726196,
+      "reward_std": 0.38381946086883545,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.030169978737831116,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 560
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 522.9583435058594,
-      "epoch": 0.561,
-      "grad_norm": 17.71687219992085,
-      "kl": 6.890625,
-      "learning_rate": 5.32725485910616e-07,
-      "loss": 0.4285,
-      "reward": 2.2837836742401123,
-      "reward_std": 0.7183063626289368,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333134651184,
-      "rewards/repetition_penalty_reward": -0.02350800298154354,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 556.7916870117188,
+      "epoch": 0.2805,
+      "grad_norm": 17.593736355103562,
+      "kl": 2.7265625,
+      "learning_rate": 9.135954477541137e-07,
+      "loss": 0.8128,
+      "reward": 2.2776578664779663,
+      "reward_std": 0.6524731516838074,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.024425473995506763,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 561
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 533.6666870117188,
-      "epoch": 0.562,
-      "grad_norm": 17.98874232563595,
-      "kl": 2.96875,
-      "learning_rate": 5.311559558218603e-07,
-      "loss": 0.7934,
-      "reward": 2.4511380195617676,
-      "reward_std": 0.5958435237407684,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.021084393840283155,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 403.9583435058594,
+      "epoch": 0.281,
+      "grad_norm": 9.443969242253909,
+      "kl": 0.826171875,
+      "learning_rate": 9.131321404647109e-07,
+      "loss": 0.3193,
+      "reward": 2.3959869742393494,
+      "reward_std": 0.171494722366333,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.034568555653095245,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 562
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 678.0625305175781,
-      "epoch": 0.563,
-      "grad_norm": 20.342890063735087,
-      "kl": 6.2265625,
-      "learning_rate": 5.295866553418358e-07,
-      "loss": 0.8234,
-      "reward": 2.3168532848358154,
-      "reward_std": 0.6717122793197632,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.025160694494843483,
-      "rewards/tag_count_reward": 0.7864583730697632,
+      "completion_length": 376.12501525878906,
+      "epoch": 0.2815,
+      "grad_norm": 235.68207779933957,
+      "kl": 1.48046875,
+      "learning_rate": 9.126677270118322e-07,
+      "loss": 0.3893,
+      "reward": 2.7154510021209717,
+      "reward_std": 0.47927525639533997,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.031076885759830475,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 563
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 617.2500305175781,
-      "epoch": 0.564,
-      "grad_norm": 16.061198883154272,
-      "kl": 3.953125,
-      "learning_rate": 5.28017603591974e-07,
-      "loss": 0.9804,
-      "reward": 2.5708781480789185,
-      "reward_std": 0.6792239546775818,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.024608048610389233,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 512.9375305175781,
+      "epoch": 0.282,
+      "grad_norm": 14.828955399249537,
+      "kl": 1.9921875,
+      "learning_rate": 9.122022088101613e-07,
+      "loss": 0.6226,
+      "reward": 2.4416096210479736,
+      "reward_std": 0.424076110124588,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.03408493287861347,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 564
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 500.29168701171875,
-      "epoch": 0.565,
-      "grad_norm": 17.72408984099671,
-      "kl": 2.89453125,
-      "learning_rate": 5.264488196906752e-07,
-      "loss": 0.7968,
-      "reward": 2.518498420715332,
-      "reward_std": 0.6503982543945312,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.024904441088438034,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 453.0208435058594,
+      "epoch": 0.2825,
+      "grad_norm": 71.8537331742799,
+      "kl": 3.9765625,
+      "learning_rate": 9.117355872777477e-07,
+      "loss": 0.6687,
+      "reward": 2.5574655532836914,
+      "reward_std": 0.44046418368816376,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.046701231971383095,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 565
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 576.8958435058594,
-      "epoch": 0.566,
-      "grad_norm": 22.26454438337593,
-      "kl": 4.609375,
-      "learning_rate": 5.248803227530763e-07,
-      "loss": 1.0678,
-      "reward": 2.483291506767273,
-      "reward_std": 0.4923284649848938,
-      "rewards/accuracy_reward": 0.6666666865348816,
+      "completion_length": 585.7708435058594,
+      "epoch": 0.283,
+      "grad_norm": 133.27147379184413,
+      "kl": 6.51171875,
+      "learning_rate": 9.112678638360015e-07,
+      "loss": 0.8637,
+      "reward": 2.124269187450409,
+      "reward_std": 0.5485763549804688,
+      "rewards/accuracy_reward": 0.3333333358168602,
       "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.020180770196020603,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "rewards/repetition_penalty_reward": -0.025036394596099854,
+      "rewards/tag_count_reward": 0.8229166865348816,
       "step": 566
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 555.2708587646484,
-      "epoch": 0.567,
-      "grad_norm": 15.919462093451902,
-      "kl": 5.5703125,
-      "learning_rate": 5.233121318908173e-07,
-      "loss": 0.7808,
-      "reward": 2.2207025289535522,
-      "reward_std": 0.6656961143016815,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9375001192092896,
-      "rewards/repetition_penalty_reward": -0.018881036899983883,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 390.8333435058594,
+      "epoch": 0.2835,
+      "grad_norm": 68.07666811218944,
+      "kl": 2.1826171875,
+      "learning_rate": 9.107990399096893e-07,
+      "loss": 0.3028,
+      "reward": 2.555752396583557,
+      "reward_std": 0.39118392765522003,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.03452543169260025,
+      "rewards/tag_count_reward": 0.9375,
       "step": 567
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 642.1041870117188,
-      "epoch": 0.568,
-      "grad_norm": 21.902991875998683,
-      "kl": 6.234375,
-      "learning_rate": 5.21744266211809e-07,
-      "loss": 0.7828,
-      "reward": 2.2312402725219727,
-      "reward_std": 0.6396611332893372,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.023968255147337914,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 411.5416717529297,
+      "epoch": 0.284,
+      "grad_norm": 15.00691010233856,
+      "kl": 1.05078125,
+      "learning_rate": 9.103291169269299e-07,
+      "loss": 0.3439,
+      "reward": 2.596803307533264,
+      "reward_std": 0.5693235397338867,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.031668941490352154,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 568
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 493.2708435058594,
-      "epoch": 0.569,
-      "grad_norm": 11.796721143909934,
-      "kl": 2.9140625,
-      "learning_rate": 5.2017674482e-07,
-      "loss": 0.4568,
-      "reward": 2.5521020889282227,
-      "reward_std": 0.6523496210575104,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.02081457804888487,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 466.6666717529297,
+      "epoch": 0.2845,
+      "grad_norm": 444.847404957193,
+      "kl": 3.9453125,
+      "learning_rate": 9.098580963191907e-07,
+      "loss": 0.8048,
+      "reward": 2.4301793575286865,
+      "reward_std": 0.7720136642456055,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.021209627389907837,
+      "rewards/tag_count_reward": 0.8541666865348816,
       "step": 569
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 563.8750305175781,
-      "epoch": 0.57,
-      "grad_norm": 16.743725013071696,
-      "kl": 4.6015625,
-      "learning_rate": 5.186095868151436e-07,
-      "loss": 0.542,
-      "reward": 2.138919949531555,
-      "reward_std": 0.6218420267105103,
-      "rewards/accuracy_reward": 0.3958333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.017330123111605644,
-      "rewards/tag_count_reward": 0.8020833432674408,
+      "completion_length": 373.87501525878906,
+      "epoch": 0.285,
+      "grad_norm": 32.33608353360448,
+      "kl": 1.00390625,
+      "learning_rate": 9.093859795212817e-07,
+      "loss": 0.2574,
+      "reward": 2.688089966773987,
+      "reward_std": 0.47320832312107086,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.039340706542134285,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 570
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 593.0625305175781,
-      "epoch": 0.571,
-      "grad_norm": 31.64391481670669,
-      "kl": 5.453125,
-      "learning_rate": 5.170428112925659e-07,
-      "loss": 0.6923,
-      "reward": 2.3365062475204468,
-      "reward_std": 0.6305623352527618,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.02460480574518442,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "completion_length": 374.75001525878906,
+      "epoch": 0.2855,
+      "grad_norm": 17.560098791576593,
+      "kl": 0.44921875,
+      "learning_rate": 9.089127679713529e-07,
+      "loss": 0.1948,
+      "reward": 2.7453384399414062,
+      "reward_std": 0.3349706828594208,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.048064423725008965,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 571
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 431.81251525878906,
-      "epoch": 0.572,
-      "grad_norm": 18.846581957559952,
-      "kl": 1.83203125,
-      "learning_rate": 5.154764373429315e-07,
-      "loss": 0.3648,
-      "reward": 2.5100269317626953,
-      "reward_std": 0.38545116782188416,
-      "rewards/accuracy_reward": 0.625,
+      "completion_length": 322.18751525878906,
+      "epoch": 0.286,
+      "grad_norm": 18.129855941015013,
+      "kl": 0.556640625,
+      "learning_rate": 9.084384631108882e-07,
+      "loss": 0.1961,
+      "reward": 2.65604305267334,
+      "reward_std": 0.520334392786026,
+      "rewards/accuracy_reward": 0.7291666865348816,
       "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.022959282621741295,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "rewards/repetition_penalty_reward": -0.027984846383333206,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 572
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 553.2291717529297,
-      "epoch": 0.573,
-      "grad_norm": 13.121698712408023,
-      "kl": 3.3359375,
-      "learning_rate": 5.139104840520135e-07,
-      "loss": 0.7169,
-      "reward": 2.331916570663452,
-      "reward_std": 0.6915982961654663,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.030930647626519203,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 459.5208435058594,
+      "epoch": 0.2865,
+      "grad_norm": 18.09227325152774,
+      "kl": 2.9375,
+      "learning_rate": 9.079630663847031e-07,
+      "loss": 0.5215,
+      "reward": 2.2212284803390503,
+      "reward_std": 0.5788050144910812,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.027035493403673172,
+      "rewards/tag_count_reward": 0.8385416865348816,
       "step": 573
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 730.5625,
-      "epoch": 0.574,
-      "grad_norm": 35.573664674103505,
-      "kl": 7.890625,
-      "learning_rate": 5.123449705004581e-07,
-      "loss": 0.8329,
-      "reward": 1.8582618832588196,
-      "reward_std": 0.7164104580879211,
-      "rewards/accuracy_reward": 0.2708333432674408,
-      "rewards/reasoning_steps_reward": 0.9027777910232544,
-      "rewards/repetition_penalty_reward": -0.018474191427230835,
-      "rewards/tag_count_reward": 0.703125,
+      "completion_length": 358.81251525878906,
+      "epoch": 0.287,
+      "grad_norm": 29.6582060019414,
+      "kl": 2.4140625,
+      "learning_rate": 9.074865792409381e-07,
+      "loss": 0.4198,
+      "reward": 2.4573017358779907,
+      "reward_std": 0.4562261551618576,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.027073362842202187,
+      "rewards/tag_count_reward": 0.921875,
       "step": 574
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 706.9166870117188,
-      "epoch": 0.575,
-      "grad_norm": 12.890226305523278,
-      "kl": 5.65625,
-      "learning_rate": 5.107799157635538e-07,
-      "loss": 0.7742,
-      "reward": 2.287819743156433,
-      "reward_std": 0.8128792345523834,
-      "rewards/accuracy_reward": 0.6458333730697632,
-      "rewards/reasoning_steps_reward": 0.9027777910232544,
-      "rewards/repetition_penalty_reward": -0.02641639020293951,
-      "rewards/tag_count_reward": 0.7656250298023224,
+      "completion_length": 335.18751525878906,
+      "epoch": 0.2875,
+      "grad_norm": 59.93954067426397,
+      "kl": 2.78857421875,
+      "learning_rate": 9.070090031310558e-07,
+      "loss": 0.3862,
+      "reward": 2.782563328742981,
+      "reward_std": 0.31483452301472425,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.04382561706006527,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 575
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 678.9166870117188,
-      "epoch": 0.576,
-      "grad_norm": 22.0467723806393,
-      "kl": 6.984375,
-      "learning_rate": 5.09215338910999e-07,
-      "loss": 0.8976,
-      "reward": 2.1274478435516357,
-      "reward_std": 0.8857446014881134,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.016649481374770403,
-      "rewards/tag_count_reward": 0.7552083432674408,
+      "completion_length": 397.4166717529297,
+      "epoch": 0.288,
+      "grad_norm": 23.817675077634586,
+      "kl": 3.94921875,
+      "learning_rate": 9.065303395098358e-07,
+      "loss": 0.48,
+      "reward": 2.5549936294555664,
+      "reward_std": 0.3741450160741806,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.017923136241734028,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 576
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 843.8333435058594,
-      "epoch": 0.577,
-      "grad_norm": 24.290171025234606,
-      "kl": 7.203125,
-      "learning_rate": 5.076512590066685e-07,
-      "loss": 0.9397,
-      "reward": 1.9734002351760864,
-      "reward_std": 0.8574240803718567,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.012710887007415295,
-      "rewards/tag_count_reward": 0.6458333432674408,
+      "completion_length": 289.75,
+      "epoch": 0.2885,
+      "grad_norm": 9.736727012422476,
+      "kl": 0.5712890625,
+      "learning_rate": 9.060505898353705e-07,
+      "loss": 0.1109,
+      "reward": 2.759944438934326,
+      "reward_std": 0.2943428307771683,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9791667461395264,
+      "rewards/repetition_penalty_reward": -0.02130552940070629,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 577
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 601.8958435058594,
-      "epoch": 0.578,
-      "grad_norm": 13.84698671695797,
-      "kl": 4.265625,
-      "learning_rate": 5.060876951083828e-07,
-      "loss": 0.6815,
-      "reward": 2.2862409353256226,
-      "reward_std": 0.7043590843677521,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.01757870987057686,
-      "rewards/tag_count_reward": 0.796875,
+      "completion_length": 458.87501525878906,
+      "epoch": 0.289,
+      "grad_norm": 32.414744280583555,
+      "kl": 5.390625,
+      "learning_rate": 9.055697555690607e-07,
+      "loss": 0.549,
+      "reward": 2.2597694396972656,
+      "reward_std": 0.8498954474925995,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.888888955116272,
+      "rewards/repetition_penalty_reward": -0.01974454615265131,
+      "rewards/tag_count_reward": 0.8281250298023224,
       "step": 578
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 473.2083435058594,
-      "epoch": 0.579,
-      "grad_norm": 21.25841736364304,
-      "kl": 2.7265625,
-      "learning_rate": 5.045246662676741e-07,
-      "loss": 0.4115,
-      "reward": 2.3248329162597656,
-      "reward_std": 0.6839044392108917,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.022389397025108337,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 442.91668701171875,
+      "epoch": 0.2895,
+      "grad_norm": 18.23142677889983,
+      "kl": 3.25390625,
+      "learning_rate": 9.050878381756107e-07,
+      "loss": 0.4236,
+      "reward": 2.289909839630127,
+      "reward_std": 0.6958013772964478,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.02432657964527607,
+      "rewards/tag_count_reward": 0.8072916865348816,
       "step": 579
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 506.2708435058594,
-      "epoch": 0.58,
-      "grad_norm": 13.768352502526005,
-      "kl": 2.55859375,
-      "learning_rate": 5.02962191529556e-07,
-      "loss": 0.4207,
-      "reward": 2.34650194644928,
-      "reward_std": 0.4788215011358261,
-      "rewards/accuracy_reward": 0.5208333358168602,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.03023417294025421,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 504.58335876464844,
+      "epoch": 0.29,
+      "grad_norm": 29.70121123782455,
+      "kl": 2.6796875,
+      "learning_rate": 9.046048391230247e-07,
+      "loss": 0.4642,
+      "reward": 2.1951472759246826,
+      "reward_std": 0.8341259658336639,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.02013048715889454,
+      "rewards/tag_count_reward": 0.7916666865348816,
       "step": 580
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 509.14585876464844,
-      "epoch": 0.581,
-      "grad_norm": 31.447352351340843,
-      "kl": 2.5546875,
-      "learning_rate": 5.014002899322896e-07,
-      "loss": 0.7482,
-      "reward": 2.202039122581482,
-      "reward_std": 0.689550518989563,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.02712748385965824,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 460.6458435058594,
+      "epoch": 0.2905,
+      "grad_norm": 19.307347308311925,
+      "kl": 2.3359375,
+      "learning_rate": 9.041207598826017e-07,
+      "loss": 0.3662,
+      "reward": 2.1872280836105347,
+      "reward_std": 0.5830141305923462,
+      "rewards/accuracy_reward": 0.4375000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.02804980892688036,
+      "rewards/tag_count_reward": 0.8333333730697632,
       "step": 581
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 636.8750305175781,
-      "epoch": 0.582,
-      "grad_norm": 31.988630155115366,
-      "kl": 5.734375,
-      "learning_rate": 4.998389805071536e-07,
-      "loss": 0.9076,
-      "reward": 2.169505000114441,
-      "reward_std": 0.861674964427948,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.028411878272891045,
-      "rewards/tag_count_reward": 0.7812500298023224,
+      "completion_length": 397.6666717529297,
+      "epoch": 0.291,
+      "grad_norm": 15.271088740961835,
+      "kl": 1.3125,
+      "learning_rate": 9.036356019289309e-07,
+      "loss": 0.3569,
+      "reward": 2.5574233531951904,
+      "reward_std": 0.5160808861255646,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.04847951419651508,
+      "rewards/tag_count_reward": 0.9114583730697632,
       "step": 582
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 605.1666870117188,
-      "epoch": 0.583,
-      "grad_norm": 18.313268799229025,
-      "kl": 3.3671875,
-      "learning_rate": 4.982782822782101e-07,
-      "loss": 0.8556,
-      "reward": 2.253504157066345,
-      "reward_std": 0.5843206644058228,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222983837128,
-      "rewards/repetition_penalty_reward": -0.020801476202905178,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 376.75,
+      "epoch": 0.2915,
+      "grad_norm": 17.403821452003328,
+      "kl": 1.63671875,
+      "learning_rate": 9.031493667398872e-07,
+      "loss": 0.2783,
+      "reward": 2.41110360622406,
+      "reward_std": 0.37185367196798325,
+      "rewards/accuracy_reward": 0.5625000223517418,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.0524381659924984,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 583
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 519.5000305175781,
-      "epoch": 0.584,
-      "grad_norm": 12.893361447903395,
-      "kl": 5.90625,
-      "learning_rate": 4.967182142620745e-07,
-      "loss": 0.5758,
-      "reward": 2.2402533292770386,
-      "reward_std": 0.9014249742031097,
-      "rewards/accuracy_reward": 0.583333358168602,
-      "rewards/reasoning_steps_reward": 0.916666716337204,
-      "rewards/repetition_penalty_reward": -0.020163366571068764,
-      "rewards/tag_count_reward": 0.7604166865348816,
+      "completion_length": 439.8333435058594,
+      "epoch": 0.292,
+      "grad_norm": 63.805319864699065,
+      "kl": 7.046875,
+      "learning_rate": 9.026620557966279e-07,
+      "loss": 0.713,
+      "reward": 2.4353259801864624,
+      "reward_std": 0.5490872263908386,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.01432684762403369,
+      "rewards/tag_count_reward": 0.8385416865348816,
       "step": 584
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 507.9583435058594,
-      "epoch": 0.585,
-      "grad_norm": 2061.671060113234,
-      "kl": 9.4453125,
-      "learning_rate": 4.951587954676837e-07,
-      "loss": 1.0213,
-      "reward": 2.4675891399383545,
-      "reward_std": 0.6850857138633728,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.02546628937125206,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 344.2708435058594,
+      "epoch": 0.2925,
+      "grad_norm": 42.56798762474608,
+      "kl": 3.458984375,
+      "learning_rate": 9.021736705835862e-07,
+      "loss": 0.2748,
+      "reward": 2.646929144859314,
+      "reward_std": 0.3856794238090515,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.02841815911233425,
+      "rewards/tag_count_reward": 0.953125,
       "step": 585
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 554.5208587646484,
-      "epoch": 0.586,
-      "grad_norm": 19.335025088394126,
-      "kl": 3.296875,
-      "learning_rate": 4.93600044896063e-07,
-      "loss": 0.7192,
-      "reward": 2.120868682861328,
-      "reward_std": 0.5526419132947922,
-      "rewards/accuracy_reward": 0.2916666679084301,
-      "rewards/reasoning_steps_reward": 0.9722222983837128,
-      "rewards/repetition_penalty_reward": -0.01802036538720131,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 294.3958435058594,
+      "epoch": 0.293,
+      "grad_norm": 19.403283253518403,
+      "kl": 2.52734375,
+      "learning_rate": 9.016842125884684e-07,
+      "loss": 0.3443,
+      "reward": 2.468275308609009,
+      "reward_std": 0.40393635630607605,
+      "rewards/accuracy_reward": 0.5833333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.03346090763807297,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 586
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 629.6041870117188,
-      "epoch": 0.587,
-      "grad_norm": 18.146859923722392,
-      "kl": 4.8671875,
-      "learning_rate": 4.920419815400968e-07,
-      "loss": 0.886,
-      "reward": 2.283421516418457,
-      "reward_std": 0.7014163732528687,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.02039794623851776,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 469.7291717529297,
+      "epoch": 0.2935,
+      "grad_norm": 73.87564910344872,
+      "kl": 6.8203125,
+      "learning_rate": 9.011936833022484e-07,
+      "loss": 0.8008,
+      "reward": 2.222065567970276,
+      "reward_std": 0.6714163422584534,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.019253874197602272,
+      "rewards/tag_count_reward": 0.8385416865348816,
       "step": 587
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 870.6041870117188,
-      "epoch": 0.588,
-      "grad_norm": 40.04439172717602,
-      "kl": 9.53125,
-      "learning_rate": 4.904846243842949e-07,
-      "loss": 0.8823,
-      "reward": 1.9214681386947632,
-      "reward_std": 0.8858973979949951,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.0368652418255806,
-      "rewards/tag_count_reward": 0.6875000298023224,
+      "completion_length": 461.56251525878906,
+      "epoch": 0.294,
+      "grad_norm": 93.860844445193,
+      "kl": 4.3984375,
+      "learning_rate": 9.007020842191634e-07,
+      "loss": 0.7156,
+      "reward": 2.255267858505249,
+      "reward_std": 0.670578122138977,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.03466268070042133,
+      "rewards/tag_count_reward": 0.859375,
       "step": 588
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 655.1250305175781,
-      "epoch": 0.589,
-      "grad_norm": 38.61615762624294,
-      "kl": 6.953125,
-      "learning_rate": 4.88927992404563e-07,
-      "loss": 0.9102,
-      "reward": 2.024886429309845,
-      "reward_std": 0.7311058044433594,
-      "rewards/accuracy_reward": 0.3333333358168602,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.028933134861290455,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 596.9583740234375,
+      "epoch": 0.2945,
+      "grad_norm": 12.13076498842532,
+      "kl": 3.4375,
+      "learning_rate": 9.002094168367095e-07,
+      "loss": 0.602,
+      "reward": 2.0464794039726257,
+      "reward_std": 0.7266620993614197,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.019492903724312782,
+      "rewards/tag_count_reward": 0.7604166865348816,
       "step": 589
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 536.8333435058594,
-      "epoch": 0.59,
-      "grad_norm": 18.369342242727228,
-      "kl": 4.1875,
-      "learning_rate": 4.873721045679706e-07,
-      "loss": 0.7111,
-      "reward": 2.163462996482849,
-      "reward_std": 0.677480012178421,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.020564796403050423,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 320.5,
+      "epoch": 0.295,
+      "grad_norm": 24.895394841833685,
+      "kl": 1.365234375,
+      "learning_rate": 8.997156826556369e-07,
+      "loss": 0.4022,
+      "reward": 2.6690471172332764,
+      "reward_std": 0.5990716367959976,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.037550284527242184,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 590
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 424.18751525878906,
-      "epoch": 0.591,
-      "grad_norm": 16.75415987323558,
-      "kl": 2.5,
-      "learning_rate": 4.858169798325198e-07,
-      "loss": 0.3791,
-      "reward": 2.6984031200408936,
-      "reward_std": 0.4382772147655487,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.018610700964927673,
-      "rewards/tag_count_reward": 0.9114583730697632,
+      "completion_length": 374.56251525878906,
+      "epoch": 0.2955,
+      "grad_norm": 22.56888130354956,
+      "kl": 1.072265625,
+      "learning_rate": 8.992208831799456e-07,
+      "loss": 0.451,
+      "reward": 2.7043356895446777,
+      "reward_std": 0.4858996868133545,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.03698389232158661,
+      "rewards/tag_count_reward": 0.921875,
       "step": 591
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 541.2083435058594,
-      "epoch": 0.592,
-      "grad_norm": 29.77449375477728,
-      "kl": 5.0546875,
-      "learning_rate": 4.842626371469149e-07,
-      "loss": 0.6774,
-      "reward": 2.1399881839752197,
-      "reward_std": 0.769228607416153,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.0197342149913311,
-      "rewards/tag_count_reward": 0.7708333432674408,
+      "completion_length": 393.0833435058594,
+      "epoch": 0.296,
+      "grad_norm": 15.07305486993295,
+      "kl": 1.38671875,
+      "learning_rate": 8.987250199168808e-07,
+      "loss": 0.3822,
+      "reward": 2.2811846137046814,
+      "reward_std": 0.35361041128635406,
+      "rewards/accuracy_reward": 0.5000000204890966,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.03825983218848705,
+      "rewards/tag_count_reward": 0.875,
       "step": 592
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 536.8125305175781,
-      "epoch": 0.593,
-      "grad_norm": 18.37611006786208,
-      "kl": 2.578125,
-      "learning_rate": 4.827090954503308e-07,
-      "loss": 0.5711,
-      "reward": 2.4687711000442505,
-      "reward_std": 0.6630526483058929,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02428449969738722,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 682.1458435058594,
+      "epoch": 0.2965,
+      "grad_norm": 15.040586253928733,
+      "kl": 3.90625,
+      "learning_rate": 8.982280943769278e-07,
+      "loss": 0.8728,
+      "reward": 2.138270854949951,
+      "reward_std": 0.7431787848472595,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.8611111342906952,
+      "rewards/repetition_penalty_reward": -0.030132046900689602,
+      "rewards/tag_count_reward": 0.8072916865348816,
       "step": 593
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 586.25,
-      "epoch": 0.594,
-      "grad_norm": 14.12630099692924,
-      "kl": 4.84375,
-      "learning_rate": 4.811563736721829e-07,
-      "loss": 0.7276,
-      "reward": 2.257407784461975,
-      "reward_std": 0.7551295161247253,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.030786780640482903,
-      "rewards/tag_count_reward": 0.7395833730697632,
+      "completion_length": 564.4583435058594,
+      "epoch": 0.297,
+      "grad_norm": 22.6147129609363,
+      "kl": 5.03125,
+      "learning_rate": 8.977301080738079e-07,
+      "loss": 0.6899,
+      "reward": 2.386793375015259,
+      "reward_std": 0.6881283521652222,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.03681784123182297,
+      "rewards/tag_count_reward": 0.8333333432674408,
       "step": 594
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 642.9583435058594,
-      "epoch": 0.595,
-      "grad_norm": 21.92582517717831,
-      "kl": 4.5703125,
-      "learning_rate": 4.79604490731896e-07,
-      "loss": 1.0746,
-      "reward": 2.405961036682129,
-      "reward_std": 0.8549763560295105,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.92361119389534,
-      "rewards/repetition_penalty_reward": -0.012441786471754313,
-      "rewards/tag_count_reward": 0.7656250298023224,
+      "completion_length": 401.4583435058594,
+      "epoch": 0.2975,
+      "grad_norm": 3434.828082451659,
+      "kl": 46.5859375,
+      "learning_rate": 8.97231062524474e-07,
+      "loss": 3.8228,
+      "reward": 2.6234763860702515,
+      "reward_std": 0.7114923894405365,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.02582914289087057,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 595
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 626.7500305175781,
-      "epoch": 0.596,
-      "grad_norm": 17.21164988814367,
-      "kl": 4.7109375,
-      "learning_rate": 4.780534655386743e-07,
-      "loss": 0.8805,
-      "reward": 2.197944164276123,
-      "reward_std": 0.7794502377510071,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.020805937238037586,
-      "rewards/tag_count_reward": 0.8020833432674408,
+      "completion_length": 522.4791870117188,
+      "epoch": 0.298,
+      "grad_norm": 81.28452749558095,
+      "kl": 4.79296875,
+      "learning_rate": 8.967309592491052e-07,
+      "loss": 0.9035,
+      "reward": 2.28587806224823,
+      "reward_std": 0.5564829260110855,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.0613443311303854,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 596
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 835.3958435058594,
-      "epoch": 0.597,
-      "grad_norm": 15.129508307858261,
-      "kl": 5.84375,
-      "learning_rate": 4.7650331699127013e-07,
-      "loss": 0.8528,
-      "reward": 2.195970058441162,
-      "reward_std": 0.8639613389968872,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.01930786296725273,
-      "rewards/tag_count_reward": 0.6875000298023224,
+      "completion_length": 384.18751525878906,
+      "epoch": 0.2985,
+      "grad_norm": 27.314706299556043,
+      "kl": 2.41015625,
+      "learning_rate": 8.962297997711027e-07,
+      "loss": 0.6414,
+      "reward": 2.3818459510803223,
+      "reward_std": 0.5823481678962708,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.04697362706065178,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 597
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 786.5208435058594,
-      "epoch": 0.598,
-      "grad_norm": 11.69120320460148,
-      "kl": 5.765625,
-      "learning_rate": 4.749540639777539e-07,
-      "loss": 1.0647,
-      "reward": 2.122739553451538,
-      "reward_std": 0.8401265442371368,
-      "rewards/accuracy_reward": 0.5000000298023224,
-      "rewards/reasoning_steps_reward": 0.9444445073604584,
-      "rewards/repetition_penalty_reward": -0.014413274824619293,
-      "rewards/tag_count_reward": 0.6927083432674408,
+      "completion_length": 362.56251525878906,
+      "epoch": 0.299,
+      "grad_norm": 24.978178060621413,
+      "kl": 3.5078125,
+      "learning_rate": 8.957275856170855e-07,
+      "loss": 0.6036,
+      "reward": 2.4027575254440308,
+      "reward_std": 0.4491366446018219,
+      "rewards/accuracy_reward": 0.5416666716337204,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.017381365410983562,
+      "rewards/tag_count_reward": 0.9270833730697632,
       "step": 598
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 647.9375305175781,
-      "epoch": 0.599,
-      "grad_norm": 56.51129938563486,
-      "kl": 5.953125,
-      "learning_rate": 4.7340572537528547e-07,
-      "loss": 0.9006,
-      "reward": 1.922993242740631,
-      "reward_std": 0.7214177250862122,
-      "rewards/accuracy_reward": 0.2500000074505806,
-      "rewards/reasoning_steps_reward": 0.888888955116272,
-      "rewards/repetition_penalty_reward": -0.023187357001006603,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 610.6041870117188,
+      "epoch": 0.2995,
+      "grad_norm": 56.7569904139385,
+      "kl": 5.6875,
+      "learning_rate": 8.952243183168848e-07,
+      "loss": 0.4682,
+      "reward": 2.0318877696990967,
+      "reward_std": 0.40879975259304047,
+      "rewards/accuracy_reward": 0.3125,
+      "rewards/reasoning_steps_reward": 0.9375000894069672,
+      "rewards/repetition_penalty_reward": -0.0358206108212471,
+      "rewards/tag_count_reward": 0.8177083730697632,
       "step": 599
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 613.0833587646484,
-      "epoch": 0.6,
-      "grad_norm": 19.124513938177042,
-      "kl": 4.421875,
-      "learning_rate": 4.7185832004988133e-07,
-      "loss": 0.8773,
-      "reward": 2.302823781967163,
-      "reward_std": 0.793424665927887,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.03745407238602638,
-      "rewards/tag_count_reward": 0.8125,
+      "completion_length": 329.6041717529297,
+      "epoch": 0.3,
+      "grad_norm": 12.091825230727677,
+      "kl": 1.107421875,
+      "learning_rate": 8.9471999940354e-07,
+      "loss": 0.28,
+      "reward": 2.2527668476104736,
+      "reward_std": 0.4062964767217636,
+      "rewards/accuracy_reward": 0.354166679084301,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.026747104711830616,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 600
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 573.4375305175781,
-      "epoch": 0.601,
-      "grad_norm": 21.633820801259493,
-      "kl": 5.6875,
-      "learning_rate": 4.703118668561875e-07,
-      "loss": 0.8408,
-      "reward": 1.9621968269348145,
-      "reward_std": 0.7473039925098419,
-      "rewards/accuracy_reward": 0.2916666716337204,
-      "rewards/reasoning_steps_reward": 0.9166666567325592,
-      "rewards/repetition_penalty_reward": -0.016969827935099602,
-      "rewards/tag_count_reward": 0.7708333432674408,
+      "completion_length": 358.3541717529297,
+      "epoch": 0.3005,
+      "grad_norm": 18.582827147329553,
+      "kl": 1.2119140625,
+      "learning_rate": 8.942146304132943e-07,
+      "loss": 0.3149,
+      "reward": 2.3439489603042603,
+      "reward_std": 0.46688786149024963,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.03799547627568245,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 601
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 409.9583435058594,
-      "epoch": 0.602,
-      "grad_norm": 21.57642847298448,
-      "kl": 2.21875,
-      "learning_rate": 4.68766384637248e-07,
-      "loss": 0.595,
-      "reward": 2.3019765615463257,
-      "reward_std": 0.5564347207546234,
-      "rewards/accuracy_reward": 0.4166666679084301,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.01573185622692108,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 347.68751525878906,
+      "epoch": 0.301,
+      "grad_norm": 25.38946154820617,
+      "kl": 0.908203125,
+      "learning_rate": 8.937082128855891e-07,
+      "loss": 0.3989,
+      "reward": 2.2832257747650146,
+      "reward_std": 0.44448164105415344,
+      "rewards/accuracy_reward": 0.3958333358168602,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03795480914413929,
+      "rewards/tag_count_reward": 0.953125,
       "step": 602
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 656.0000305175781,
-      "epoch": 0.603,
-      "grad_norm": 18.78803423208506,
-      "kl": 4.390625,
-      "learning_rate": 4.672218922242759e-07,
-      "loss": 0.8124,
-      "reward": 2.2104203701019287,
-      "reward_std": 0.60920250415802,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.0170103432610631,
-      "rewards/tag_count_reward": 0.796875,
+      "completion_length": 391.10418701171875,
+      "epoch": 0.3015,
+      "grad_norm": 37.47070105859472,
+      "kl": 2.890625,
+      "learning_rate": 8.932007483630596e-07,
+      "loss": 0.673,
+      "reward": 2.4643146991729736,
+      "reward_std": 0.42459146678447723,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.025268660858273506,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 603
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 704.3541870117188,
-      "epoch": 0.604,
-      "grad_norm": 17.09263062702093,
-      "kl": 5.625,
-      "learning_rate": 4.656784084364238e-07,
-      "loss": 0.8064,
-      "reward": 1.9496487379074097,
-      "reward_std": 0.6896158158779144,
-      "rewards/accuracy_reward": 0.3333333432674408,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.017365206964313984,
-      "rewards/tag_count_reward": 0.7031250298023224,
+      "completion_length": 499.81251525878906,
+      "epoch": 0.302,
+      "grad_norm": 27.80855047913277,
+      "kl": 3.8515625,
+      "learning_rate": 8.926922383915315e-07,
+      "loss": 0.6036,
+      "reward": 2.298141598701477,
+      "reward_std": 0.6866291165351868,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.03171971254050732,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 604
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 709.5416870117188,
-      "epoch": 0.605,
-      "grad_norm": 14.18186552493172,
-      "kl": 5.40625,
-      "learning_rate": 4.641359520805548e-07,
-      "loss": 0.9465,
-      "reward": 1.9041990041732788,
-      "reward_std": 0.5497189462184906,
-      "rewards/accuracy_reward": 0.25000000558793545,
-      "rewards/reasoning_steps_reward": 0.92361119389534,
-      "rewards/repetition_penalty_reward": -0.014204025734215975,
-      "rewards/tag_count_reward": 0.7447916865348816,
+      "completion_length": 437.6041717529297,
+      "epoch": 0.3025,
+      "grad_norm": 14.735625956308642,
+      "kl": 2.66015625,
+      "learning_rate": 8.921826845200138e-07,
+      "loss": 0.4696,
+      "reward": 2.3992427587509155,
+      "reward_std": 0.566670224070549,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.029576689936220646,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 605
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 644.1875,
-      "epoch": 0.606,
-      "grad_norm": 21.61369420565558,
-      "kl": 4.0625,
-      "learning_rate": 4.6259454195101267e-07,
-      "loss": 1.0074,
-      "reward": 2.424658179283142,
-      "reward_std": 0.7420503497123718,
-      "rewards/accuracy_reward": 0.6458333730697632,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.021522408351302147,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 311.0,
+      "epoch": 0.303,
+      "grad_norm": 12.206231699378542,
+      "kl": 1.1015625,
+      "learning_rate": 8.916720883006963e-07,
+      "loss": 0.1574,
+      "reward": 2.774844169616699,
+      "reward_std": 0.4380817115306854,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.034183867275714874,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 606
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 675.5833740234375,
-      "epoch": 0.607,
-      "grad_norm": 20.01514537964165,
-      "kl": 4.484375,
-      "learning_rate": 4.6105419682939316e-07,
-      "loss": 0.94,
-      "reward": 2.2794888019561768,
-      "reward_std": 0.6631845533847809,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.015650033485144377,
-      "rewards/tag_count_reward": 0.7604166865348816,
+      "completion_length": 292.18751525878906,
+      "epoch": 0.3035,
+      "grad_norm": 31.971619676180616,
+      "kl": 2.32421875,
+      "learning_rate": 8.911604512889434e-07,
+      "loss": 0.3176,
+      "reward": 2.726539731025696,
+      "reward_std": 0.508837565779686,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.028668691404163837,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 607
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 588.3125,
-      "epoch": 0.608,
-      "grad_norm": 15.233743739571594,
-      "kl": 3.671875,
-      "learning_rate": 4.59514935484316e-07,
-      "loss": 0.6811,
-      "reward": 2.3253376483917236,
-      "reward_std": 0.7875647842884064,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.013204073999077082,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 377.7708435058594,
+      "epoch": 0.304,
+      "grad_norm": 47.81529047217983,
+      "kl": 5.484375,
+      "learning_rate": 8.906477750432903e-07,
+      "loss": 0.7836,
+      "reward": 2.3868411779403687,
+      "reward_std": 0.6196691393852234,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.02114514308050275,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 608
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 443.0625,
-      "epoch": 0.609,
-      "grad_norm": 23.918832923297472,
-      "kl": 2.390625,
-      "learning_rate": 4.579767766711944e-07,
-      "loss": 0.5455,
-      "reward": 2.240389823913574,
-      "reward_std": 0.6115424335002899,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.02349921502172947,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 377.00001525878906,
+      "epoch": 0.3045,
+      "grad_norm": 64.95300495357336,
+      "kl": 3.546875,
+      "learning_rate": 8.901340611254378e-07,
+      "loss": 0.7444,
+      "reward": 2.6928662061691284,
+      "reward_std": 0.6587081551551819,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03630049154162407,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 609
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 501.1458435058594,
-      "epoch": 0.61,
-      "grad_norm": 21.038617061450985,
-      "kl": 4.1015625,
-      "learning_rate": 4.5643973913200837e-07,
-      "loss": 0.809,
-      "reward": 2.464809775352478,
-      "reward_std": 0.7636359930038452,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9374999701976776,
-      "rewards/repetition_penalty_reward": -0.019565261900424957,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 409.3958435058594,
+      "epoch": 0.305,
+      "grad_norm": 40.39290779152927,
+      "kl": 5.671875,
+      "learning_rate": 8.896193111002475e-07,
+      "loss": 0.4324,
+      "reward": 2.182259440422058,
+      "reward_std": 0.708200603723526,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.8541666865348816,
+      "rewards/repetition_penalty_reward": -0.0364905521273613,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 610
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 681.8125,
-      "epoch": 0.611,
-      "grad_norm": 166.2589762047824,
-      "kl": 12.0625,
-      "learning_rate": 4.549038415950751e-07,
-      "loss": 1.4371,
-      "reward": 1.9376187920570374,
-      "reward_std": 0.9399848580360413,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.8888888955116272,
-      "rewards/repetition_penalty_reward": -0.013770201243460178,
-      "rewards/tag_count_reward": 0.6875000298023224,
+      "completion_length": 382.5416717529297,
+      "epoch": 0.3055,
+      "grad_norm": 18.330556875068268,
+      "kl": 2.4296875,
+      "learning_rate": 8.891035265357371e-07,
+      "loss": 0.5719,
+      "reward": 2.4720453023910522,
+      "reward_std": 0.5455569326877594,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.024482558481395245,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 611
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 600.0625152587891,
-      "epoch": 0.612,
-      "grad_norm": 63.22955290170651,
-      "kl": 7.67578125,
-      "learning_rate": 4.5336910277482155e-07,
-      "loss": 1.1131,
-      "reward": 2.076914668083191,
-      "reward_std": 0.6576991975307465,
-      "rewards/accuracy_reward": 0.3958333432674408,
-      "rewards/reasoning_steps_reward": 0.8472222685813904,
-      "rewards/repetition_penalty_reward": -0.02030777558684349,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 463.68751525878906,
+      "epoch": 0.306,
+      "grad_norm": 20.965812940157566,
+      "kl": 3.12109375,
+      "learning_rate": 8.88586709003076e-07,
+      "loss": 0.5898,
+      "reward": 2.5473427772521973,
+      "reward_std": 0.7215102910995483,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.02904626540839672,
+      "rewards/tag_count_reward": 0.875,
       "step": 612
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 442.7291717529297,
-      "epoch": 0.613,
-      "grad_norm": 41.71612091835444,
-      "kl": 4.90625,
-      "learning_rate": 4.51835541371556e-07,
-      "loss": 0.4885,
-      "reward": 2.2476359605789185,
-      "reward_std": 0.6959330439567566,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.01972527615725994,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 277.93751525878906,
+      "epoch": 0.3065,
+      "grad_norm": 21.659936470606215,
+      "kl": 1.1328125,
+      "learning_rate": 8.8806886007658e-07,
+      "loss": 0.1303,
+      "reward": 2.7708282470703125,
+      "reward_std": 0.48425713181495667,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.026046608574688435,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 613
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 576.3750305175781,
-      "epoch": 0.614,
-      "grad_norm": 22.018133081081007,
-      "kl": 4.546875,
-      "learning_rate": 4.503031760712397e-07,
-      "loss": 0.813,
-      "reward": 2.5740822553634644,
-      "reward_std": 0.7949818074703217,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.01619557524099946,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 278.68751525878906,
+      "epoch": 0.307,
+      "grad_norm": 22.568162104962866,
+      "kl": 0.58203125,
+      "learning_rate": 8.875499813337067e-07,
+      "loss": 0.2044,
+      "reward": 2.8232574462890625,
+      "reward_std": 0.385304257273674,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.022228870075196028,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 614
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.2916870117188,
-      "epoch": 0.615,
-      "grad_norm": 20.90403864764217,
-      "kl": 4.046875,
-      "learning_rate": 4.4877202554526084e-07,
-      "loss": 0.8765,
-      "reward": 2.31480872631073,
-      "reward_std": 0.6609582602977753,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.018524575512856245,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 345.8333435058594,
+      "epoch": 0.3075,
+      "grad_norm": 13.0831888313229,
+      "kl": 0.63818359375,
+      "learning_rate": 8.87030074355051e-07,
+      "loss": 0.2361,
+      "reward": 2.7106308937072754,
+      "reward_std": 0.3397009428590536,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9652777314186096,
+      "rewards/repetition_penalty_reward": -0.030688603408634663,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 615
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 571.6666870117188,
-      "epoch": 0.616,
-      "grad_norm": 20.967646246131107,
-      "kl": 5.28125,
-      "learning_rate": 4.4724210845020494e-07,
-      "loss": 0.7339,
-      "reward": 2.366945743560791,
-      "reward_std": 0.5788512378931046,
+      "completion_length": 297.50001525878906,
+      "epoch": 0.308,
+      "grad_norm": 9.770757617195565,
+      "kl": 0.42138671875,
+      "learning_rate": 8.865091407243394e-07,
+      "loss": 0.0876,
+      "reward": 2.564429521560669,
+      "reward_std": 0.381553053855896,
       "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.014998776838183403,
-      "rewards/tag_count_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.041473302990198135,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 616
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 597.75,
-      "epoch": 0.617,
-      "grad_norm": 62.998128656761025,
-      "kl": 8.015625,
-      "learning_rate": 4.457134434276293e-07,
-      "loss": 1.0981,
-      "reward": 2.1459895372390747,
-      "reward_std": 0.7351883053779602,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9444444477558136,
-      "rewards/repetition_penalty_reward": -0.011996725108474493,
-      "rewards/tag_count_reward": 0.7552083432674408,
-      "step": 617
+      "completion_length": 433.37501525878906,
+      "epoch": 0.3085,
+      "grad_norm": 15.238026778188452,
+      "kl": 2.79296875,
+      "learning_rate": 8.859871820284261e-07,
+      "loss": 0.6701,
+      "reward": 2.4649064540863037,
+      "reward_std": 0.5701871514320374,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.029885290190577507,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 617
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 578.9583435058594,
-      "epoch": 0.618,
-      "grad_norm": 35.2952233783846,
-      "kl": 7.296875,
-      "learning_rate": 4.441860491038345e-07,
-      "loss": 1.1757,
-      "reward": 1.9162747263908386,
-      "reward_std": 0.7140241265296936,
-      "rewards/accuracy_reward": 0.2708333432674408,
-      "rewards/reasoning_steps_reward": 0.9097222685813904,
-      "rewards/repetition_penalty_reward": -0.029905788600444794,
-      "rewards/tag_count_reward": 0.7656250298023224,
+      "completion_length": 383.4166717529297,
+      "epoch": 0.309,
+      "grad_norm": 22.09258325323561,
+      "kl": 2.6484375,
+      "learning_rate": 8.85464199857288e-07,
+      "loss": 0.3808,
+      "reward": 2.376816987991333,
+      "reward_std": 0.6643776893615723,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.04158583842217922,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 618
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 412.0208435058594,
-      "epoch": 0.619,
-      "grad_norm": 20.183860058369124,
-      "kl": 3.5234375,
-      "learning_rate": 4.4265994408963867e-07,
-      "loss": 0.6553,
-      "reward": 2.6022950410842896,
-      "reward_std": 0.690807044506073,
+      "completion_length": 418.75001525878906,
+      "epoch": 0.3095,
+      "grad_norm": 18.162555409765705,
+      "kl": 2.8828125,
+      "learning_rate": 8.849401958040192e-07,
+      "loss": 0.5727,
+      "reward": 2.6696484088897705,
+      "reward_std": 0.5072681605815887,
       "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.01749663334339857,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.019587615504860878,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 619
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 551.7083587646484,
-      "epoch": 0.62,
-      "grad_norm": 30.2741526195306,
-      "kl": 3.6640625,
-      "learning_rate": 4.4113514698014953e-07,
-      "loss": 1.2213,
-      "reward": 2.3724182844161987,
-      "reward_std": 0.7322799563407898,
-      "rewards/accuracy_reward": 0.5625000149011612,
+      "completion_length": 368.75,
+      "epoch": 0.31,
+      "grad_norm": 26.7814555750565,
+      "kl": 2.5703125,
+      "learning_rate": 8.844151714648274e-07,
+      "loss": 0.5849,
+      "reward": 2.5200384855270386,
+      "reward_std": 0.5991145968437195,
+      "rewards/accuracy_reward": 0.6875000298023224,
       "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.021679001860320568,
-      "rewards/tag_count_reward": 0.8802083730697632,
+      "rewards/repetition_penalty_reward": -0.025100392289459705,
+      "rewards/tag_count_reward": 0.90625,
       "step": 620
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 477.89585876464844,
-      "epoch": 0.621,
-      "grad_norm": 23.62241565191626,
-      "kl": 4.26171875,
-      "learning_rate": 4.3961167635453876e-07,
-      "loss": 0.6591,
-      "reward": 2.3674203753471375,
-      "reward_std": 0.6224862933158875,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.012787933927029371,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 393.2083435058594,
+      "epoch": 0.3105,
+      "grad_norm": 14.7849673987086,
+      "kl": 2.59765625,
+      "learning_rate": 8.838891284390273e-07,
+      "loss": 0.5688,
+      "reward": 2.561677932739258,
+      "reward_std": 0.5532724261283875,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.033808158710598946,
+      "rewards/tag_count_reward": 0.9427083730697632,
       "step": 621
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.2500305175781,
-      "epoch": 0.622,
-      "grad_norm": 26.80506804197189,
-      "kl": 4.28125,
-      "learning_rate": 4.3808955077581546e-07,
-      "loss": 1.3034,
-      "reward": 2.1128806471824646,
-      "reward_std": 0.6486698687076569,
-      "rewards/accuracy_reward": 0.37500002048909664,
-      "rewards/reasoning_steps_reward": 0.9236111044883728,
-      "rewards/repetition_penalty_reward": -0.01906377449631691,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "completion_length": 471.14585876464844,
+      "epoch": 0.311,
+      "grad_norm": 20.53699662396806,
+      "kl": 3.734375,
+      "learning_rate": 8.833620683290375e-07,
+      "loss": 0.8009,
+      "reward": 2.2843284606933594,
+      "reward_std": 0.7857449054718018,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.8819444179534912,
+      "rewards/repetition_penalty_reward": -0.029907681047916412,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 622
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 591.5416870117188,
-      "epoch": 0.623,
-      "grad_norm": 17.029635227087233,
-      "kl": 4.71875,
-      "learning_rate": 4.365687887905988e-07,
-      "loss": 0.7684,
-      "reward": 2.364261269569397,
-      "reward_std": 0.7807703912258148,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9236111640930176,
-      "rewards/repetition_penalty_reward": -0.017683234997093678,
-      "rewards/tag_count_reward": 0.8125000298023224,
+      "completion_length": 650.0000152587891,
+      "epoch": 0.3115,
+      "grad_norm": 69.27168548723343,
+      "kl": 9.65625,
+      "learning_rate": 8.828339927403745e-07,
+      "loss": 1.1394,
+      "reward": 1.8982658386230469,
+      "reward_std": 0.9247550368309021,
+      "rewards/accuracy_reward": 0.4375000149011612,
+      "rewards/reasoning_steps_reward": 0.7361111044883728,
+      "rewards/repetition_penalty_reward": -0.020136947743594646,
+      "rewards/tag_count_reward": 0.7447916865348816,
       "step": 623
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 657.8541870117188,
-      "epoch": 0.624,
-      "grad_norm": 29.359681860338085,
-      "kl": 8.28125,
-      "learning_rate": 4.350494089288943e-07,
-      "loss": 1.1389,
-      "reward": 2.2420893907546997,
-      "reward_std": 0.8185697495937347,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.01485508494079113,
-      "rewards/tag_count_reward": 0.7083333432674408,
+      "completion_length": 845.4792175292969,
+      "epoch": 0.312,
+      "grad_norm": 31.60208939904907,
+      "kl": 7.09375,
+      "learning_rate": 8.823049032816478e-07,
+      "loss": 0.8724,
+      "reward": 1.8869645595550537,
+      "reward_std": 0.8868106305599213,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.7152778208255768,
+      "rewards/repetition_penalty_reward": -0.021021696738898754,
+      "rewards/tag_count_reward": 0.7343750298023224,
       "step": 624
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 572.8958435058594,
-      "epoch": 0.625,
-      "grad_norm": 25.429883504952183,
-      "kl": 5.6875,
-      "learning_rate": 4.3353142970386557e-07,
-      "loss": 0.8027,
-      "reward": 2.140592336654663,
-      "reward_std": 0.777917891740799,
-      "rewards/accuracy_reward": 0.4375,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.019129890482872725,
-      "rewards/tag_count_reward": 0.7708333730697632,
+      "completion_length": 501.29168701171875,
+      "epoch": 0.3125,
+      "grad_norm": 37.663529836109795,
+      "kl": 3.2734375,
+      "learning_rate": 8.817748015645558e-07,
+      "loss": 0.8728,
+      "reward": 2.3407769203186035,
+      "reward_std": 0.8801598846912384,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.86111119389534,
+      "rewards/repetition_penalty_reward": -0.025542644783854485,
+      "rewards/tag_count_reward": 0.8385416865348816,
       "step": 625
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 713.5833435058594,
-      "epoch": 0.626,
-      "grad_norm": 43.92929389982015,
-      "kl": 8.28125,
-      "learning_rate": 4.3201486961161093e-07,
-      "loss": 1.2015,
-      "reward": 1.720135509967804,
-      "reward_std": 0.6745887994766235,
-      "rewards/accuracy_reward": 0.2083333358168602,
+      "completion_length": 470.6041717529297,
+      "epoch": 0.313,
+      "grad_norm": 29.75409371364291,
+      "kl": 2.640625,
+      "learning_rate": 8.812436892038805e-07,
+      "loss": 0.7078,
+      "reward": 2.3025213479995728,
+      "reward_std": 0.6479816734790802,
+      "rewards/accuracy_reward": 0.520833358168602,
       "rewards/reasoning_steps_reward": 0.9097222685813904,
-      "rewards/repetition_penalty_reward": -0.012503455393016338,
-      "rewards/tag_count_reward": 0.6145833730697632,
+      "rewards/repetition_penalty_reward": -0.023867566138505936,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 626
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 645.3125,
-      "epoch": 0.627,
-      "grad_norm": 57.87651440683306,
-      "kl": 8.28125,
-      "learning_rate": 4.304997471309361e-07,
-      "loss": 1.0589,
-      "reward": 2.1112887859344482,
-      "reward_std": 0.8265305161476135,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9027777910232544,
-      "rewards/repetition_penalty_reward": -0.010239105205982924,
-      "rewards/tag_count_reward": 0.6979166865348816,
+      "completion_length": 459.4375,
+      "epoch": 0.3135,
+      "grad_norm": 20.212825860910232,
+      "kl": 2.25390625,
+      "learning_rate": 8.807115678174819e-07,
+      "loss": 0.5701,
+      "reward": 2.5997543334960938,
+      "reward_std": 0.5478241741657257,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.03045404888689518,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 627
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 648.2916870117188,
-      "epoch": 0.628,
-      "grad_norm": 50.008615375469205,
-      "kl": 5.703125,
-      "learning_rate": 4.2898608072313045e-07,
-      "loss": 1.1558,
-      "reward": 2.0761906504631042,
-      "reward_std": 0.8007937371730804,
-      "rewards/accuracy_reward": 0.3958333432674408,
-      "rewards/reasoning_steps_reward": 0.9027778208255768,
-      "rewards/repetition_penalty_reward": -0.02450390998274088,
-      "rewards/tag_count_reward": 0.8020833432674408,
+      "completion_length": 447.72918701171875,
+      "epoch": 0.314,
+      "grad_norm": 29.197351580965048,
+      "kl": 4.796875,
+      "learning_rate": 8.801784390262943e-07,
+      "loss": 0.8627,
+      "reward": 2.3921085596084595,
+      "reward_std": 0.6133331656455994,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.8819444477558136,
+      "rewards/repetition_penalty_reward": -0.02108590304851532,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 628
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 510.45835876464844,
-      "epoch": 0.629,
-      "grad_norm": 29.38809691299576,
-      "kl": 3.390625,
-      "learning_rate": 4.2747388883174154e-07,
-      "loss": 0.6864,
-      "reward": 2.3009127378463745,
-      "reward_std": 0.7014258801937103,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.009851326234638691,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 463.91668701171875,
+      "epoch": 0.3145,
+      "grad_norm": 44.79167908730396,
+      "kl": 4.546875,
+      "learning_rate": 8.796443044543203e-07,
+      "loss": 0.8839,
+      "reward": 2.317416787147522,
+      "reward_std": 0.7653204798698425,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.026333278976380825,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 629
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 722.8750305175781,
-      "epoch": 0.63,
-      "grad_norm": 15.997709703522245,
-      "kl": 5.765625,
-      "learning_rate": 4.2596318988235037e-07,
-      "loss": 0.7532,
-      "reward": 1.911007285118103,
-      "reward_std": 0.7613692283630371,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.8958333730697632,
-      "rewards/repetition_penalty_reward": -0.021284431219100952,
-      "rewards/tag_count_reward": 0.6822916865348816,
+      "completion_length": 412.25001525878906,
+      "epoch": 0.315,
+      "grad_norm": 44.27467343188964,
+      "kl": 3.2734375,
+      "learning_rate": 8.791091657286267e-07,
+      "loss": 0.8308,
+      "reward": 2.6593304872512817,
+      "reward_std": 0.5533206462860107,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.024697369895875454,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 630
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 539.125,
-      "epoch": 0.631,
-      "grad_norm": 21.914178804717245,
-      "kl": 3.2734375,
-      "learning_rate": 4.2445400228234687e-07,
-      "loss": 0.7349,
-      "reward": 2.5217862129211426,
-      "reward_std": 0.7204074263572693,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.018144560046494007,
-      "rewards/tag_count_reward": 0.8385416865348816,
+      "completion_length": 368.5833435058594,
+      "epoch": 0.3155,
+      "grad_norm": 30.80486315018559,
+      "kl": 2.9765625,
+      "learning_rate": 8.785730244793386e-07,
+      "loss": 0.8518,
+      "reward": 2.6147913932800293,
+      "reward_std": 0.5551804751157761,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.03451400436460972,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 631
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 582.5000152587891,
-      "epoch": 0.632,
-      "grad_norm": 23.574151149835284,
-      "kl": 4.3125,
-      "learning_rate": 4.2294634442070553e-07,
-      "loss": 0.776,
-      "reward": 2.2236552238464355,
-      "reward_std": 0.7735611796379089,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.014192146249115467,
-      "rewards/tag_count_reward": 0.7656250298023224,
+      "completion_length": 411.91668701171875,
+      "epoch": 0.316,
+      "grad_norm": 20.520000609846605,
+      "kl": 3.0859375,
+      "learning_rate": 8.780358823396352e-07,
+      "loss": 0.5721,
+      "reward": 2.6789560317993164,
+      "reward_std": 0.5304727852344513,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.039794113487005234,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 632
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 606.8125305175781,
-      "epoch": 0.633,
-      "grad_norm": 27.523457434836338,
-      "kl": 3.375,
-      "learning_rate": 4.214402346677619e-07,
-      "loss": 0.8245,
-      "reward": 2.0943539142608643,
-      "reward_std": 0.7719534635543823,
-      "rewards/accuracy_reward": 0.375,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.020229644142091274,
-      "rewards/tag_count_reward": 0.78125,
+      "completion_length": 363.5,
+      "epoch": 0.3165,
+      "grad_norm": 63.575635135783344,
+      "kl": 2.2265625,
+      "learning_rate": 8.774977409457447e-07,
+      "loss": 0.5012,
+      "reward": 2.500579595565796,
+      "reward_std": 0.2963787615299225,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.03935109078884125,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 633
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 468.04168701171875,
-      "epoch": 0.634,
-      "grad_norm": 26.042764464725323,
-      "kl": 3.5234375,
-      "learning_rate": 4.1993569137498776e-07,
-      "loss": 0.6013,
-      "reward": 2.314329147338867,
-      "reward_std": 0.48445817828178406,
-      "rewards/accuracy_reward": 0.541666679084301,
-      "rewards/reasoning_steps_reward": 0.9305555522441864,
-      "rewards/repetition_penalty_reward": -0.01205990044400096,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 379.9791717529297,
+      "epoch": 0.317,
+      "grad_norm": 12.662953099547957,
+      "kl": 1.4296875,
+      "learning_rate": 8.769586019369391e-07,
+      "loss": 0.3589,
+      "reward": 2.5281978845596313,
+      "reward_std": 0.6068725287914276,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.04645505174994469,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 634
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 585.9791717529297,
-      "epoch": 0.635,
-      "grad_norm": 110.89920440001275,
-      "kl": 5.375,
-      "learning_rate": 4.1843273287476854e-07,
-      "loss": 1.2961,
-      "reward": 2.3673198223114014,
-      "reward_std": 0.7202720046043396,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.014624850358814001,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "completion_length": 461.27085876464844,
+      "epoch": 0.3175,
+      "grad_norm": 12.945423045438853,
+      "kl": 1.072265625,
+      "learning_rate": 8.764184669555293e-07,
+      "loss": 0.7605,
+      "reward": 2.4953324794769287,
+      "reward_std": 0.6080846786499023,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.039389871060848236,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 635
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 543.1250152587891,
-      "epoch": 0.636,
-      "grad_norm": 168.91845413040815,
-      "kl": 6.78125,
-      "learning_rate": 4.1693137748017915e-07,
-      "loss": 0.9638,
-      "reward": 2.172168493270874,
-      "reward_std": 0.7465826869010925,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.01880389917641878,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 443.5208435058594,
+      "epoch": 0.318,
+      "grad_norm": 7.650991561257087,
+      "kl": 1.1640625,
+      "learning_rate": 8.758773376468604e-07,
+      "loss": 0.3799,
+      "reward": 2.4011194705963135,
+      "reward_std": 0.4720783084630966,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9166666567325592,
+      "rewards/repetition_penalty_reward": -0.025964029133319855,
+      "rewards/tag_count_reward": 0.9270833432674408,
       "step": 636
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 367.12501525878906,
-      "epoch": 0.637,
-      "grad_norm": 14.167427077773384,
-      "kl": 1.734375,
-      "learning_rate": 4.15431643484761e-07,
-      "loss": 0.2434,
-      "reward": 2.3709245920181274,
-      "reward_std": 0.41279859840869904,
-      "rewards/accuracy_reward": 0.4791666716337204,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.011019973549991846,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 408.93751525878906,
+      "epoch": 0.3185,
+      "grad_norm": 15.77947801589448,
+      "kl": 0.68359375,
+      "learning_rate": 8.753352156593055e-07,
+      "loss": 0.5988,
+      "reward": 2.430037498474121,
+      "reward_std": 0.7513845264911652,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889849185944,
+      "rewards/repetition_penalty_reward": -0.03697636350989342,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 637
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 409.0208435058594,
-      "epoch": 0.638,
-      "grad_norm": 44.523191376668905,
-      "kl": 2.34765625,
-      "learning_rate": 4.1393354916230005e-07,
-      "loss": 0.6225,
-      "reward": 2.353254556655884,
-      "reward_std": 0.5010079145431519,
-      "rewards/accuracy_reward": 0.5000000111758709,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.025217789225280285,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 375.37501525878906,
+      "epoch": 0.319,
+      "grad_norm": 13.718169867441816,
+      "kl": 0.548828125,
+      "learning_rate": 8.747921026442629e-07,
+      "loss": 0.424,
+      "reward": 2.6864466667175293,
+      "reward_std": 0.45896846055984497,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.027095147408545017,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 638
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 375.56251525878906,
-      "epoch": 0.639,
-      "grad_norm": 19.368462137693786,
-      "kl": 1.640625,
-      "learning_rate": 4.124371127666024e-07,
-      "loss": 0.4301,
-      "reward": 2.260585904121399,
-      "reward_std": 0.4507751762866974,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.017191945109516382,
-      "rewards/tag_count_reward": 0.9375000298023224,
+      "completion_length": 341.875,
+      "epoch": 0.3195,
+      "grad_norm": 17.467108774461327,
+      "kl": 0.828125,
+      "learning_rate": 8.742480002561487e-07,
+      "loss": 0.3762,
+      "reward": 2.4510825872421265,
+      "reward_std": 0.5428140312433243,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.02808418497443199,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 639
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.1666870117188,
-      "epoch": 0.64,
-      "grad_norm": 23.408643386653363,
-      "kl": 5.40625,
-      "learning_rate": 4.1094235253127374e-07,
-      "loss": 1.0619,
-      "reward": 2.11304247379303,
-      "reward_std": 0.6343182921409607,
-      "rewards/accuracy_reward": 0.3333333432674408,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.017165867146104574,
-      "rewards/tag_count_reward": 0.8385416865348816,
+      "completion_length": 345.0208435058594,
+      "epoch": 0.32,
+      "grad_norm": 9.631118470397297,
+      "kl": 1.0517578125,
+      "learning_rate": 8.737029101523929e-07,
+      "loss": 0.2993,
+      "reward": 2.2707436084747314,
+      "reward_std": 0.6229533404111862,
+      "rewards/accuracy_reward": 0.458333358168602,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.02439526654779911,
+      "rewards/tag_count_reward": 0.9270833730697632,
       "step": 640
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 338.7083435058594,
-      "epoch": 0.641,
-      "grad_norm": 26.252165148333535,
-      "kl": 2.26953125,
-      "learning_rate": 4.0944928666949527e-07,
-      "loss": 0.0696,
-      "reward": 2.4866459369659424,
-      "reward_std": 0.29595404863357544,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.020298474468290806,
-      "rewards/tag_count_reward": 0.9583333730697632,
+      "completion_length": 456.97918701171875,
+      "epoch": 0.3205,
+      "grad_norm": 18.87763752323685,
+      "kl": 1.7734375,
+      "learning_rate": 8.731568339934348e-07,
+      "loss": 0.5908,
+      "reward": 1.9133847951889038,
+      "reward_std": 0.4975127577781677,
+      "rewards/accuracy_reward": 0.18750000558793545,
+      "rewards/reasoning_steps_reward": 0.8611111640930176,
+      "rewards/repetition_penalty_reward": -0.036268092691898346,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 641
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 613.9166717529297,
-      "epoch": 0.642,
-      "grad_norm": 83.06728366905014,
-      "kl": 9.09375,
-      "learning_rate": 4.079579333738039e-07,
-      "loss": 1.4448,
-      "reward": 2.021218776702881,
-      "reward_std": 0.7841091454029083,
-      "rewards/accuracy_reward": 0.3541666679084301,
-      "rewards/reasoning_steps_reward": 0.881944477558136,
-      "rewards/repetition_penalty_reward": -0.011767241638153791,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 352.35418701171875,
+      "epoch": 0.321,
+      "grad_norm": 9.424502666967252,
+      "kl": 1.328125,
+      "learning_rate": 8.726097734427172e-07,
+      "loss": 0.1237,
+      "reward": 2.5691460371017456,
+      "reward_std": 0.49686890840530396,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.05759007856249809,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 642
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 517.4375,
-      "epoch": 0.643,
-      "grad_norm": 29.352168555788648,
-      "kl": 4.1328125,
-      "learning_rate": 4.064683108158685e-07,
-      "loss": 0.8538,
-      "reward": 2.1417415142059326,
-      "reward_std": 0.7165846824645996,
-      "rewards/accuracy_reward": 0.3333333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.019716893322765827,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 334.6666717529297,
+      "epoch": 0.3215,
+      "grad_norm": 8.243342051209725,
+      "kl": 1.203125,
+      "learning_rate": 8.72061730166681e-07,
+      "loss": 0.2726,
+      "reward": 2.5343031883239746,
+      "reward_std": 0.46366532891988754,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.024724683724343777,
+      "rewards/tag_count_reward": 0.90625,
       "step": 643
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 472.8958435058594,
-      "epoch": 0.644,
-      "grad_norm": 27.512095159522406,
-      "kl": 5.0078125,
-      "learning_rate": 4.0498043714627006e-07,
-      "loss": 0.4818,
-      "reward": 2.178748071193695,
-      "reward_std": 0.5007181763648987,
-      "rewards/accuracy_reward": 0.3958333544433117,
-      "rewards/reasoning_steps_reward": 0.9444445073604584,
-      "rewards/repetition_penalty_reward": -0.01569645293056965,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 359.8333435058594,
+      "epoch": 0.322,
+      "grad_norm": 13.81419209256365,
+      "kl": 1.3359375,
+      "learning_rate": 8.715127058347614e-07,
+      "loss": 0.4082,
+      "reward": 2.7587969303131104,
+      "reward_std": 0.44490741193294525,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.04849480465054512,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 644
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 487.25001525878906,
-      "epoch": 0.645,
-      "grad_norm": 42.088759648308134,
-      "kl": 2.484375,
-      "learning_rate": 4.034943304942796e-07,
-      "loss": 0.9438,
-      "reward": 2.6129757165908813,
-      "reward_std": 0.6440701186656952,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.02244096901267767,
-      "rewards/tag_count_reward": 0.9270833730697632,
+      "completion_length": 306.0208435058594,
+      "epoch": 0.3225,
+      "grad_norm": 7.722256504528858,
+      "kl": 1.244140625,
+      "learning_rate": 8.709627021193816e-07,
+      "loss": 0.259,
+      "reward": 2.2697885036468506,
+      "reward_std": 0.5568395256996155,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.04097543843090534,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 645
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 506.22918701171875,
-      "epoch": 0.646,
-      "grad_norm": 15.525366461402383,
-      "kl": 3.5234375,
-      "learning_rate": 4.020100089676376e-07,
-      "loss": 0.5343,
-      "reward": 2.1641972064971924,
-      "reward_std": 0.6387200355529785,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9374999701976776,
-      "rewards/repetition_penalty_reward": -0.018094514962285757,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 317.5833435058594,
+      "epoch": 0.323,
+      "grad_norm": 20.616365592900145,
+      "kl": 3.88671875,
+      "learning_rate": 8.704117206959484e-07,
+      "loss": 0.4825,
+      "reward": 2.6422855854034424,
+      "reward_std": 0.6127668023109436,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.027853261679410934,
+      "rewards/tag_count_reward": 0.9270833730697632,
       "step": 646
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 676.6875305175781,
-      "epoch": 0.647,
-      "grad_norm": 27.736362696423306,
-      "kl": 5.53125,
-      "learning_rate": 4.005274906523336e-07,
-      "loss": 1.0089,
-      "reward": 2.418774127960205,
-      "reward_std": 0.8770462274551392,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.022198081016540527,
-      "rewards/tag_count_reward": 0.8020833730697632,
+      "completion_length": 302.8958435058594,
+      "epoch": 0.3235,
+      "grad_norm": 8.074414625169473,
+      "kl": 1.1787109375,
+      "learning_rate": 8.698597632428466e-07,
+      "loss": 0.1034,
+      "reward": 2.6797112226486206,
+      "reward_std": 0.35129064321517944,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.03383052162826061,
+      "rewards/tag_count_reward": 0.984375,
       "step": 647
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 477.10418701171875,
-      "epoch": 0.648,
-      "grad_norm": 32.594519011116475,
-      "kl": 3.15625,
-      "learning_rate": 3.9904679361238526e-07,
-      "loss": 0.8488,
-      "reward": 2.172785520553589,
-      "reward_std": 0.6095466017723083,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.014714594930410385,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 305.7708435058594,
+      "epoch": 0.324,
+      "grad_norm": 16.085041721886572,
+      "kl": 1.09375,
+      "learning_rate": 8.693068314414344e-07,
+      "loss": 0.2719,
+      "reward": 2.777750849723816,
+      "reward_std": 0.49438077211380005,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.041693685576319695,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 648
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 466.7083435058594,
-      "epoch": 0.649,
-      "grad_norm": 16.053982100692465,
-      "kl": 2.6982421875,
-      "learning_rate": 3.975679358896189e-07,
-      "loss": 0.6344,
-      "reward": 2.492353320121765,
-      "reward_std": 0.48535051196813583,
-      "rewards/accuracy_reward": 0.6041666716337204,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.014591319020837545,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 385.125,
+      "epoch": 0.3245,
+      "grad_norm": 16.9777467202549,
+      "kl": 1.3984375,
+      "learning_rate": 8.687529269760379e-07,
+      "loss": 0.4593,
+      "reward": 2.6736103296279907,
+      "reward_std": 0.5818447470664978,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.022570334374904633,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 649
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 421.4583435058594,
-      "epoch": 0.65,
-      "grad_norm": 33.807152135246525,
-      "kl": 4.578125,
-      "learning_rate": 3.9609093550344907e-07,
-      "loss": 0.9576,
-      "reward": 2.364701509475708,
-      "reward_std": 0.6279256045818329,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.9513889849185944,
-      "rewards/repetition_penalty_reward": -0.01377071999013424,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "completion_length": 391.6875,
+      "epoch": 0.325,
+      "grad_norm": 16.236705823940458,
+      "kl": 0.869140625,
+      "learning_rate": 8.681980515339463e-07,
+      "loss": 0.5449,
+      "reward": 2.599353790283203,
+      "reward_std": 0.6066733002662659,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03953541815280914,
+      "rewards/tag_count_reward": 0.9583333730697632,
       "step": 650
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 430.08335876464844,
-      "epoch": 0.651,
-      "grad_norm": 54.92633243016837,
-      "kl": 4.2021484375,
-      "learning_rate": 3.946158104506594e-07,
-      "loss": 0.7082,
-      "reward": 2.6497297286987305,
-      "reward_std": 0.48640553653240204,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02214530110359192,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 360.8333435058594,
+      "epoch": 0.3255,
+      "grad_norm": 8.416276221716757,
+      "kl": 0.7939453125,
+      "learning_rate": 8.676422068054064e-07,
+      "loss": 0.163,
+      "reward": 2.606260299682617,
+      "reward_std": 0.40933507680892944,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03957311250269413,
+      "rewards/tag_count_reward": 0.9583333730697632,
       "step": 651
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 449.62501525878906,
-      "epoch": 0.652,
-      "grad_norm": 338.42582974401023,
-      "kl": 9.8125,
-      "learning_rate": 3.931425787051832e-07,
-      "loss": 1.7215,
-      "reward": 2.6626009941101074,
-      "reward_std": 0.6703141033649445,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.019690793938934803,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 283.0625,
+      "epoch": 0.326,
+      "grad_norm": 15.69003304653512,
+      "kl": 0.537109375,
+      "learning_rate": 8.670853944836176e-07,
+      "loss": 0.0603,
+      "reward": 2.657088875770569,
+      "reward_std": 0.4240477532148361,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.026939059607684612,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 652
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 304.85418701171875,
-      "epoch": 0.653,
-      "grad_norm": 17.319403464457796,
-      "kl": 1.7109375,
-      "learning_rate": 3.9167125821788416e-07,
-      "loss": 0.14,
-      "reward": 2.6247986555099487,
-      "reward_std": 0.37955524772405624,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.015826416201889515,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 282.4166717529297,
+      "epoch": 0.3265,
+      "grad_norm": 14.606960299212915,
+      "kl": 1.259765625,
+      "learning_rate": 8.665276162647267e-07,
+      "loss": 0.1115,
+      "reward": 2.3430339097976685,
+      "reward_std": 0.4984329789876938,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.0232857009395957,
+      "rewards/tag_count_reward": 0.9427083730697632,
       "step": 653
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 333.5,
-      "epoch": 0.654,
-      "grad_norm": 73.877312813324,
-      "kl": 1.705078125,
-      "learning_rate": 3.902018669163384e-07,
-      "loss": 0.4122,
-      "reward": 2.905174732208252,
-      "reward_std": 0.26645080000162125,
-      "rewards/accuracy_reward": 0.9583333730697632,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.01843650173395872,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 359.8333435058594,
+      "epoch": 0.327,
+      "grad_norm": 17.83111188587982,
+      "kl": 1.36328125,
+      "learning_rate": 8.659688738478231e-07,
+      "loss": 0.336,
+      "reward": 2.423835873603821,
+      "reward_std": 0.32535358518362045,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02408089768141508,
+      "rewards/tag_count_reward": 0.9270833730697632,
       "step": 654
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 458.52085876464844,
-      "epoch": 0.655,
-      "grad_norm": 87.77289847817504,
-      "kl": 6.890625,
-      "learning_rate": 3.8873442270461485e-07,
-      "loss": 1.0085,
-      "reward": 2.5801607370376587,
-      "reward_std": 0.7309716641902924,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.92361119389534,
-      "rewards/repetition_penalty_reward": -0.02574214804917574,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 341.375,
+      "epoch": 0.3275,
+      "grad_norm": 11.748552434251017,
+      "kl": 1.251953125,
+      "learning_rate": 8.654091689349329e-07,
+      "loss": 0.2916,
+      "reward": 2.4952768087387085,
+      "reward_std": 0.5520404577255249,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.03770934417843819,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 655
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 542.7500305175781,
-      "epoch": 0.656,
-      "grad_norm": 104.02522400010055,
-      "kl": 8.09375,
-      "learning_rate": 3.872689434630585e-07,
-      "loss": 1.0897,
-      "reward": 2.465412139892578,
-      "reward_std": 0.540164515376091,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.010282414965331554,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 436.0,
+      "epoch": 0.328,
+      "grad_norm": 12.053678746654692,
+      "kl": 1.6171875,
+      "learning_rate": 8.648485032310144e-07,
+      "loss": 0.687,
+      "reward": 2.4294373989105225,
+      "reward_std": 0.5722118020057678,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.023687631357461214,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 656
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 510.3958435058594,
-      "epoch": 0.657,
-      "grad_norm": 43.13318047502928,
-      "kl": 6.53125,
-      "learning_rate": 3.8580544704807117e-07,
-      "loss": 0.8464,
-      "reward": 2.118070363998413,
-      "reward_std": 0.7317387461662292,
-      "rewards/accuracy_reward": 0.3541666865348816,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.02255462296307087,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 437.93751525878906,
+      "epoch": 0.3285,
+      "grad_norm": 6.092908560579486,
+      "kl": 1.51171875,
+      "learning_rate": 8.642868784439527e-07,
+      "loss": 0.4096,
+      "reward": 2.373212218284607,
+      "reward_std": 0.5252508372068405,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.02435732912272215,
+      "rewards/tag_count_reward": 0.8906250298023224,
       "step": 657
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 491.0833435058594,
-      "epoch": 0.658,
-      "grad_norm": 40.435901700616085,
-      "kl": 2.390625,
-      "learning_rate": 3.843439512918949e-07,
-      "loss": 0.7348,
-      "reward": 2.421813726425171,
-      "reward_std": 0.6320691108703613,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.017422514967620373,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 469.1458435058594,
+      "epoch": 0.329,
+      "grad_norm": 7.258415740610364,
+      "kl": 1.6015625,
+      "learning_rate": 8.63724296284554e-07,
+      "loss": 0.6144,
+      "reward": 2.5958104133605957,
+      "reward_std": 0.7455363571643829,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.030925797298550606,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 658
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 508.5000305175781,
-      "epoch": 0.659,
-      "grad_norm": 13.800630290866199,
-      "kl": 2.7109375,
-      "learning_rate": 3.8288447400239443e-07,
-      "loss": 0.4474,
-      "reward": 2.462868094444275,
-      "reward_std": 0.6122282147407532,
-      "rewards/accuracy_reward": 0.6041666865348816,
+      "completion_length": 470.0208435058594,
+      "epoch": 0.3295,
+      "grad_norm": 12.390886242134757,
+      "kl": 1.51953125,
+      "learning_rate": 8.631607584665413e-07,
+      "loss": 0.4942,
+      "reward": 2.3429232835769653,
+      "reward_std": 0.5768862962722778,
+      "rewards/accuracy_reward": 0.5000000298023224,
       "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.016298661939799786,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.04249337501823902,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 659
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 857.5000305175781,
-      "epoch": 0.66,
-      "grad_norm": 26.939008770933444,
-      "kl": 8.09375,
-      "learning_rate": 3.8142703296283953e-07,
-      "loss": 0.9327,
-      "reward": 1.7486651539802551,
-      "reward_std": 0.8026353716850281,
-      "rewards/accuracy_reward": 0.2708333358168602,
-      "rewards/reasoning_steps_reward": 0.8750000298023224,
-      "rewards/repetition_penalty_reward": -0.027376560494303703,
-      "rewards/tag_count_reward": 0.6302083432674408,
+      "completion_length": 527.1458435058594,
+      "epoch": 0.33,
+      "grad_norm": 9.725057262385192,
+      "kl": 1.94140625,
+      "learning_rate": 8.625962667065487e-07,
+      "loss": 0.4938,
+      "reward": 2.334376096725464,
+      "reward_std": 0.6681367456912994,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.05277678743004799,
+      "rewards/tag_count_reward": 0.8802083432674408,
       "step": 660
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 504.16668701171875,
-      "epoch": 0.661,
-      "grad_norm": 40.18446601488726,
-      "kl": 2.203125,
-      "learning_rate": 3.7997164593168983e-07,
-      "loss": 0.8515,
-      "reward": 2.3131070137023926,
-      "reward_std": 0.6172126531600952,
-      "rewards/accuracy_reward": 0.479166679084301,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.02543477900326252,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 334.9791717529297,
+      "epoch": 0.3305,
+      "grad_norm": 7.932623599953271,
+      "kl": 1.439453125,
+      "learning_rate": 8.620308227241157e-07,
+      "loss": 0.3623,
+      "reward": 2.632012963294983,
+      "reward_std": 0.5029560029506683,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9166667461395264,
+      "rewards/repetition_penalty_reward": -0.03465373069047928,
+      "rewards/tag_count_reward": 0.9375,
       "step": 661
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 404.04168701171875,
-      "epoch": 0.662,
-      "grad_norm": 19.64806111061822,
-      "kl": 1.80078125,
-      "learning_rate": 3.785183306423767e-07,
-      "loss": 0.3147,
-      "reward": 2.165276527404785,
-      "reward_std": 0.5021775662899017,
-      "rewards/accuracy_reward": 0.2708333358168602,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.018751220777630806,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 446.8958435058594,
+      "epoch": 0.331,
+      "grad_norm": 7.989726237635147,
+      "kl": 1.38671875,
+      "learning_rate": 8.614644282416831e-07,
+      "loss": 0.5033,
+      "reward": 2.527227759361267,
+      "reward_std": 0.6098933517932892,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.04048066772520542,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 662
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 457.5416717529297,
-      "epoch": 0.663,
-      "grad_norm": 55.11379027399269,
-      "kl": 2.46875,
-      "learning_rate": 3.7706710480308835e-07,
-      "loss": 0.9405,
-      "reward": 2.3307788372039795,
-      "reward_std": 0.5423199832439423,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.014707335270941257,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 411.3333435058594,
+      "epoch": 0.3315,
+      "grad_norm": 10.307046200594176,
+      "kl": 0.96875,
+      "learning_rate": 8.608970849845862e-07,
+      "loss": 0.7632,
+      "reward": 2.6601351499557495,
+      "reward_std": 0.5646943897008896,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556416511536,
+      "rewards/repetition_penalty_reward": -0.025628680363297462,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 663
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 376.9583435058594,
-      "epoch": 0.664,
-      "grad_norm": 46.90834721507952,
-      "kl": 3.6171875,
-      "learning_rate": 3.7561798609655373e-07,
-      "loss": 0.7765,
-      "reward": 2.5231001377105713,
-      "reward_std": 0.43632885813713074,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.018566494807600975,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 404.5416717529297,
+      "epoch": 0.332,
+      "grad_norm": 16.627402867924832,
+      "kl": 2.0234375,
+      "learning_rate": 8.603287946810513e-07,
+      "loss": 0.5452,
+      "reward": 2.565666437149048,
+      "reward_std": 0.7432173788547516,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.03155587986111641,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 664
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 472.875,
-      "epoch": 0.665,
-      "grad_norm": 36.36032137924947,
-      "kl": 4.66015625,
-      "learning_rate": 3.7417099217982686e-07,
-      "loss": 0.6785,
-      "reward": 2.34568190574646,
-      "reward_std": 0.5171235352754593,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.013693147338926792,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 428.9166717529297,
+      "epoch": 0.3325,
+      "grad_norm": 10.078394182499805,
+      "kl": 1.65234375,
+      "learning_rate": 8.597595590621892e-07,
+      "loss": 0.4888,
+      "reward": 2.448451519012451,
+      "reward_std": 0.7055695652961731,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.03765958361327648,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 665
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 389.18751525878906,
-      "epoch": 0.666,
-      "grad_norm": 43.19617147385512,
-      "kl": 4.390625,
-      "learning_rate": 3.72726140684072e-07,
-      "loss": 0.7594,
-      "reward": 2.619339108467102,
-      "reward_std": 0.5449521988630295,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.0160775538533926,
+      "completion_length": 480.7916717529297,
+      "epoch": 0.333,
+      "grad_norm": 11.542957521544354,
+      "kl": 1.796875,
+      "learning_rate": 8.591893798619903e-07,
+      "loss": 0.3699,
+      "reward": 2.4829243421554565,
+      "reward_std": 0.61925408244133,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222983837128,
+      "rewards/repetition_penalty_reward": -0.04138128738850355,
       "rewards/tag_count_reward": 0.9270833432674408,
       "step": 666
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 399.2708435058594,
-      "epoch": 0.667,
-      "grad_norm": 34.057601735761196,
-      "kl": 3.2890625,
-      "learning_rate": 3.712834492143487e-07,
-      "loss": 0.504,
-      "reward": 2.495935320854187,
-      "reward_std": 0.37501704692840576,
-      "rewards/accuracy_reward": 0.6041666865348816,
+      "completion_length": 306.1666717529297,
+      "epoch": 0.3335,
+      "grad_norm": 12.408852597847376,
+      "kl": 0.822265625,
+      "learning_rate": 8.586182588173194e-07,
+      "loss": 0.2134,
+      "reward": 2.452653646469116,
+      "reward_std": 0.415747806429863,
+      "rewards/accuracy_reward": 0.5208333432674408,
       "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.01448146067559719,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "rewards/repetition_penalty_reward": -0.03172139264643192,
+      "rewards/tag_count_reward": 0.9843750298023224,
       "step": 667
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 516.7500305175781,
-      "epoch": 0.668,
-      "grad_norm": 69.82503772312313,
-      "kl": 5.609375,
-      "learning_rate": 3.6984293534939737e-07,
-      "loss": 0.9785,
-      "reward": 2.399042248725891,
-      "reward_std": 0.7354680299758911,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9513889849185944,
-      "rewards/repetition_penalty_reward": -0.015888389199972153,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 321.1666717529297,
+      "epoch": 0.334,
+      "grad_norm": 5.871197681642463,
+      "kl": 0.3994140625,
+      "learning_rate": 8.580461976679099e-07,
+      "loss": 0.2745,
+      "reward": 2.681954860687256,
+      "reward_std": 0.4819119870662689,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.0472117830067873,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 668
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 462.50001525878906,
-      "epoch": 0.669,
-      "grad_norm": 32.586128695642,
-      "kl": 4.359375,
-      "learning_rate": 3.6840461664142444e-07,
-      "loss": 0.7185,
-      "reward": 2.410151481628418,
-      "reward_std": 0.44753579795360565,
-      "rewards/accuracy_reward": 0.5625000149011612,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.016932022757828236,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 290.1458435058594,
+      "epoch": 0.3345,
+      "grad_norm": 20.400573211519653,
+      "kl": 1.712890625,
+      "learning_rate": 8.574731981563597e-07,
+      "loss": -0.0559,
+      "reward": 2.527996778488159,
+      "reward_std": 0.21311798691749573,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.037975482642650604,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 669
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 551.4375305175781,
-      "epoch": 0.67,
-      "grad_norm": 22.702344571228725,
-      "kl": 3.390625,
-      "learning_rate": 3.6696851061588994e-07,
-      "loss": 0.7687,
-      "reward": 2.3318662643432617,
-      "reward_std": 0.6909976303577423,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.017092150636017323,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 338.2083435058594,
+      "epoch": 0.335,
+      "grad_norm": 16.428046693280866,
+      "kl": 0.92578125,
+      "learning_rate": 8.568992620281243e-07,
+      "loss": 0.1399,
+      "reward": 2.6338796615600586,
+      "reward_std": 0.30435943603515625,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.05014814343303442,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 670
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 565.4791870117188,
-      "epoch": 0.671,
-      "grad_norm": 27.618008409549773,
-      "kl": 4.2265625,
-      "learning_rate": 3.655346347712922e-07,
-      "loss": 0.8779,
-      "reward": 2.0886669158935547,
-      "reward_std": 0.6510388255119324,
-      "rewards/accuracy_reward": 0.2916666716337204,
-      "rewards/reasoning_steps_reward": 0.9444445073604584,
-      "rewards/repetition_penalty_reward": -0.017235863022506237,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 283.06251525878906,
+      "epoch": 0.3355,
+      "grad_norm": 9.998010496236825,
+      "kl": 0.4912109375,
+      "learning_rate": 8.56324391031513e-07,
+      "loss": 0.1636,
+      "reward": 2.781676411628723,
+      "reward_std": 0.2892776355147362,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.04124024324119091,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 671
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 487.18751525878906,
-      "epoch": 0.672,
-      "grad_norm": 24.355953329836403,
-      "kl": 2.40625,
-      "learning_rate": 3.641030065789562e-07,
-      "loss": 0.5358,
-      "reward": 2.412584662437439,
-      "reward_std": 0.6757764220237732,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.021443082485347986,
-      "rewards/tag_count_reward": 0.9062500298023224,
-      "step": 672
-    },
-    {
+      "completion_length": 328.75001525878906,
+      "epoch": 0.336,
+      "grad_norm": 15.737714162396987,
+      "kl": 0.54296875,
+      "learning_rate": 8.557485869176825e-07,
+      "loss": 0.3096,
+      "reward": 2.5456645488739014,
+      "reward_std": 0.3961578607559204,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04114115610718727,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 672
+    },
+    {
       "clip_ratio": 0.0,
-      "completion_length": 558.0833435058594,
-      "epoch": 0.673,
-      "grad_norm": 45.77951130491503,
-      "kl": 3.328125,
-      "learning_rate": 3.6267364348281946e-07,
-      "loss": 1.0366,
-      "reward": 2.355557918548584,
-      "reward_std": 0.6195343136787415,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02465049270540476,
-      "rewards/tag_count_reward": 0.859375,
+      "completion_length": 274.3333435058594,
+      "epoch": 0.3365,
+      "grad_norm": 18.76746009667458,
+      "kl": 0.302734375,
+      "learning_rate": 8.551718514406318e-07,
+      "loss": 0.0533,
+      "reward": 2.44698166847229,
+      "reward_std": 0.15537243708968163,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03912944719195366,
+      "rewards/tag_count_reward": 1.0,
       "step": 673
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 529.1666870117188,
-      "epoch": 0.674,
-      "grad_norm": 41.84743305028243,
-      "kl": 4.65625,
-      "learning_rate": 3.612465628992203e-07,
-      "loss": 0.8446,
-      "reward": 2.305051267147064,
-      "reward_std": 0.6280795484781265,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.010920950677245855,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 259.5208435058594,
+      "epoch": 0.337,
+      "grad_norm": 6.288258024651295,
+      "kl": 1.1064453125,
+      "learning_rate": 8.545941863571973e-07,
+      "loss": 0.0915,
+      "reward": 2.7296632528305054,
+      "reward_std": 0.41032251715660095,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.029017308726906776,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 674
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 442.2083435058594,
-      "epoch": 0.675,
-      "grad_norm": 27.763872293665937,
-      "kl": 2.62890625,
-      "learning_rate": 3.5982178221668533e-07,
-      "loss": 0.7363,
-      "reward": 2.520777702331543,
-      "reward_std": 0.6029552221298218,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.019152968656271696,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 288.62501525878906,
+      "epoch": 0.3375,
+      "grad_norm": 5.025259032815617,
+      "kl": 0.3076171875,
+      "learning_rate": 8.540155934270471e-07,
+      "loss": 0.0915,
+      "reward": 2.7451287508010864,
+      "reward_std": 0.32976970821619034,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.04653800651431084,
+      "rewards/tag_count_reward": 1.0,
       "step": 675
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 626.0833435058594,
-      "epoch": 0.676,
-      "grad_norm": 146.14695704648105,
-      "kl": 11.421875,
-      "learning_rate": 3.5839931879571725e-07,
-      "loss": 1.492,
-      "reward": 1.9430179595947266,
-      "reward_std": 0.49505507946014404,
-      "rewards/accuracy_reward": 0.2291666716337204,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.027468254789710045,
-      "rewards/tag_count_reward": 0.7552083432674408,
+      "completion_length": 325.7083435058594,
+      "epoch": 0.338,
+      "grad_norm": 9.159439405693467,
+      "kl": 1.34375,
+      "learning_rate": 8.534360744126753e-07,
+      "loss": 0.4539,
+      "reward": 2.591364622116089,
+      "reward_std": 0.5737589597702026,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.047524161636829376,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 676
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 448.0000305175781,
-      "epoch": 0.677,
-      "grad_norm": 147.52034576763307,
-      "kl": 6.8095703125,
-      "learning_rate": 3.5697918996858443e-07,
-      "loss": 0.9345,
-      "reward": 2.3936002254486084,
-      "reward_std": 0.4429845064878464,
-      "rewards/accuracy_reward": 0.5208333358168602,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.010913772508502007,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 329.0833435058594,
+      "epoch": 0.3385,
+      "grad_norm": 8.458585993366595,
+      "kl": 1.11328125,
+      "learning_rate": 8.528556310793979e-07,
+      "loss": 0.4093,
+      "reward": 2.7048439979553223,
+      "reward_std": 0.45828860998153687,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.034739382565021515,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 677
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 512.9166870117188,
-      "epoch": 0.678,
-      "grad_norm": 140.08402872570105,
-      "kl": 7.609375,
-      "learning_rate": 3.555614130391079e-07,
-      "loss": 1.0663,
-      "reward": 2.3663707971572876,
-      "reward_std": 0.6558624804019928,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.02772645093500614,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 332.0208435058594,
+      "epoch": 0.339,
+      "grad_norm": 8.46266477495005,
+      "kl": 0.63671875,
+      "learning_rate": 8.522742651953456e-07,
+      "loss": 0.444,
+      "reward": 2.682199239730835,
+      "reward_std": 0.5500051826238632,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.02960638701915741,
+      "rewards/tag_count_reward": 0.96875,
       "step": 678
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 417.0,
-      "epoch": 0.679,
-      "grad_norm": 96.73929298933032,
-      "kl": 4.5390625,
-      "learning_rate": 3.5414600528245266e-07,
-      "loss": 1.0437,
-      "reward": 2.23331880569458,
-      "reward_std": 0.6037196218967438,
-      "rewards/accuracy_reward": 0.37500002048909664,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.02883412316441536,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 417.56251525878906,
+      "epoch": 0.3395,
+      "grad_norm": 10.566916898300722,
+      "kl": 1.509765625,
+      "learning_rate": 8.516919785314595e-07,
+      "loss": 0.3333,
+      "reward": 2.5015060901641846,
+      "reward_std": 0.57795649766922,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.03495226427912712,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 679
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 357.5625,
-      "epoch": 0.68,
-      "grad_norm": 21.200219203477406,
-      "kl": 2.416015625,
-      "learning_rate": 3.5273298394491515e-07,
-      "loss": 0.5258,
-      "reward": 2.639436721801758,
-      "reward_std": 0.38805626332759857,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.009868907742202282,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 333.6666717529297,
+      "epoch": 0.34,
+      "grad_norm": 11.247764389759753,
+      "kl": 0.75390625,
+      "learning_rate": 8.511087728614862e-07,
+      "loss": 0.5781,
+      "reward": 2.7087562084198,
+      "reward_std": 0.4665229171514511,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.04471610300242901,
+      "rewards/tag_count_reward": 0.96875,
       "step": 680
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 387.7708435058594,
-      "epoch": 0.681,
-      "grad_norm": 14.11167765417183,
-      "kl": 1.080078125,
-      "learning_rate": 3.513223662437147e-07,
-      "loss": 0.1954,
-      "reward": 2.5773390531539917,
-      "reward_std": 0.42192623019218445,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.02856375463306904,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 308.6458435058594,
+      "epoch": 0.3405,
+      "grad_norm": 6.290960445037793,
+      "kl": 0.47705078125,
+      "learning_rate": 8.50524649961971e-07,
+      "loss": 0.3798,
+      "reward": 2.7960026264190674,
+      "reward_std": 0.2514218669384718,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.03385855816304684,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 681
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 463.89585876464844,
-      "epoch": 0.682,
-      "grad_norm": 35.11880079033104,
-      "kl": 5.19921875,
-      "learning_rate": 3.4991416936678276e-07,
-      "loss": 0.666,
-      "reward": 2.4289733171463013,
-      "reward_std": 0.7871748507022858,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.025887836702167988,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 262.375,
+      "epoch": 0.341,
+      "grad_norm": 3.8670578486681526,
+      "kl": 0.21484375,
+      "learning_rate": 8.499396116122535e-07,
+      "loss": -0.0315,
+      "reward": 2.5164601802825928,
+      "reward_std": 0.1599935106933117,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.03215097542852163,
+      "rewards/tag_count_reward": 1.0,
       "step": 682
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 596.7500305175781,
-      "epoch": 0.683,
-      "grad_norm": 21.947154239119367,
-      "kl": 5.1796875,
-      "learning_rate": 3.4850841047255364e-07,
-      "loss": 0.7009,
-      "reward": 1.906419038772583,
-      "reward_std": 0.5553164780139923,
-      "rewards/accuracy_reward": 0.1666666716337204,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.015455981716513634,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 307.4791717529297,
+      "epoch": 0.3415,
+      "grad_norm": 5.1795240406789285,
+      "kl": 0.51318359375,
+      "learning_rate": 8.493536595944622e-07,
+      "loss": 0.2556,
+      "reward": 2.6397162675857544,
+      "reward_std": 0.3590812534093857,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.04778374172747135,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 683
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 661.0208435058594,
-      "epoch": 0.684,
-      "grad_norm": 25.598616282466537,
-      "kl": 4.953125,
-      "learning_rate": 3.471051066897562e-07,
-      "loss": 0.9532,
-      "reward": 2.2370306253433228,
-      "reward_std": 0.6496289968490601,
-      "rewards/accuracy_reward": 0.4375,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.016441638581454754,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 347.79168701171875,
+      "epoch": 0.342,
+      "grad_norm": 22.46580378549878,
+      "kl": 1.2060546875,
+      "learning_rate": 8.487667956935087e-07,
+      "loss": 0.5061,
+      "reward": 2.5328707695007324,
+      "reward_std": 0.5347878038883209,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.05740702152252197,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 684
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 464.70835876464844,
-      "epoch": 0.685,
-      "grad_norm": 49.95217737848169,
-      "kl": 1.669921875,
-      "learning_rate": 3.45704275117204e-07,
-      "loss": 0.614,
-      "reward": 2.6332918405532837,
-      "reward_std": 0.49084220826625824,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.01948598213493824,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 322.66668701171875,
+      "epoch": 0.3425,
+      "grad_norm": 25.14822551370442,
+      "kl": 1.689453125,
+      "learning_rate": 8.481790216970819e-07,
+      "loss": 0.386,
+      "reward": 2.6461331844329834,
+      "reward_std": 0.41380666196346283,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03442266769707203,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 685
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 686.3958435058594,
-      "epoch": 0.686,
-      "grad_norm": 68.70749191055191,
-      "kl": 5.984375,
-      "learning_rate": 3.4430593282358777e-07,
-      "loss": 0.9271,
-      "reward": 2.089155912399292,
-      "reward_std": 0.5732367038726807,
-      "rewards/accuracy_reward": 0.3750000223517418,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.013274761848151684,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 395.5208435058594,
+      "epoch": 0.343,
+      "grad_norm": 78.13941184528808,
+      "kl": 3.5234375,
+      "learning_rate": 8.475903393956433e-07,
+      "loss": 0.7695,
+      "reward": 2.565727114677429,
+      "reward_std": 0.7066747546195984,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.03496745601296425,
+      "rewards/tag_count_reward": 0.9270833730697632,
       "step": 686
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 432.2083435058594,
-      "epoch": 0.687,
-      "grad_norm": 32.08822044000933,
-      "kl": 1.79296875,
-      "learning_rate": 3.429100968472668e-07,
-      "loss": 0.5082,
-      "reward": 2.5415300130844116,
-      "reward_std": 0.48894260823726654,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.017497885040938854,
-      "rewards/tag_count_reward": 0.9687500298023224,
+      "completion_length": 333.4375,
+      "epoch": 0.3435,
+      "grad_norm": 10.49194552043806,
+      "kl": 0.970703125,
+      "learning_rate": 8.470007505824215e-07,
+      "loss": 0.3382,
+      "reward": 2.6929080486297607,
+      "reward_std": 0.45839959383010864,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03973084129393101,
+      "rewards/tag_count_reward": 0.96875,
       "step": 687
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 427.12501525878906,
-      "epoch": 0.688,
-      "grad_norm": 44.25186835660776,
-      "kl": 1.40625,
-      "learning_rate": 3.4151678419606233e-07,
-      "loss": 0.6032,
-      "reward": 2.457394003868103,
-      "reward_std": 0.4042099863290787,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.020036617293953896,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 340.00001525878906,
+      "epoch": 0.344,
+      "grad_norm": 24.890945025863278,
+      "kl": 1.193359375,
+      "learning_rate": 8.464102570534061e-07,
+      "loss": 0.5374,
+      "reward": 2.5198949575424194,
+      "reward_std": 0.4044055640697479,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02698008343577385,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 688
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 475.31251525878906,
-      "epoch": 0.689,
-      "grad_norm": 29.033085916995276,
-      "kl": 3.84375,
-      "learning_rate": 3.4012601184704904e-07,
-      "loss": 0.7725,
-      "reward": 2.554749846458435,
-      "reward_std": 0.610569179058075,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.023375309072434902,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 338.6666717529297,
+      "epoch": 0.3445,
+      "grad_norm": 18.61334570722872,
+      "kl": 0.78125,
+      "learning_rate": 8.458188606073431e-07,
+      "loss": 0.4962,
+      "reward": 2.715654969215393,
+      "reward_std": 0.4276605248451233,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.030872873961925507,
+      "rewards/tag_count_reward": 0.96875,
       "step": 689
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 494.12501525878906,
-      "epoch": 0.69,
-      "grad_norm": 57.99082028977846,
-      "kl": 5.2578125,
-      "learning_rate": 3.387377967463493e-07,
-      "loss": 0.5999,
-      "reward": 2.3892362117767334,
-      "reward_std": 0.5195634961128235,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9722222983837128,
-      "rewards/repetition_penalty_reward": -0.020485990215092897,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 270.2083435058594,
+      "epoch": 0.345,
+      "grad_norm": 95.37811997865627,
+      "kl": 2.6484375,
+      "learning_rate": 8.452265630457282e-07,
+      "loss": 0.161,
+      "reward": 2.8763829469680786,
+      "reward_std": 0.2372975405305624,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.026394893415272236,
+      "rewards/tag_count_reward": 1.0,
       "step": 690
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 491.29168701171875,
-      "epoch": 0.691,
-      "grad_norm": 63.801944338558975,
-      "kl": 6.546875,
-      "learning_rate": 3.3735215580892575e-07,
-      "loss": 0.8708,
-      "reward": 2.57023286819458,
-      "reward_std": 0.5704822838306427,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.020044888369739056,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 285.7708435058594,
+      "epoch": 0.3455,
+      "grad_norm": 4.725098859206571,
+      "kl": 0.3291015625,
+      "learning_rate": 8.446333661728028e-07,
+      "loss": 0.0645,
+      "reward": 2.569468140602112,
+      "reward_std": 0.23410103470087051,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.027754205279052258,
+      "rewards/tag_count_reward": 1.0,
       "step": 691
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 438.83335876464844,
-      "epoch": 0.692,
-      "grad_norm": 63.66620073917318,
-      "kl": 4.0234375,
-      "learning_rate": 3.359691059183761e-07,
-      "loss": 0.7175,
-      "reward": 2.389517307281494,
-      "reward_std": 0.6298649907112122,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.03756603505462408,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 287.62501525878906,
+      "epoch": 0.346,
+      "grad_norm": 12.97522462997103,
+      "kl": 1.1708984375,
+      "learning_rate": 8.440392717955475e-07,
+      "loss": 0.121,
+      "reward": 2.6713898181915283,
+      "reward_std": 0.5372248291969299,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.04909628629684448,
+      "rewards/tag_count_reward": 0.984375,
       "step": 692
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 559.5416870117188,
-      "epoch": 0.693,
-      "grad_norm": 261.1873431552047,
-      "kl": 10.6875,
-      "learning_rate": 3.3458866392672694e-07,
-      "loss": 1.6707,
-      "reward": 2.32285612821579,
-      "reward_std": 0.4213666617870331,
-      "rewards/accuracy_reward": 0.5000000111758709,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.019157718401402235,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 288.9583435058594,
+      "epoch": 0.3465,
+      "grad_norm": 77.5201396944965,
+      "kl": 4.1875,
+      "learning_rate": 8.434442817236765e-07,
+      "loss": 0.6809,
+      "reward": 2.826512098312378,
+      "reward_std": 0.32751043140888214,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03112685587257147,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 693
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 548.9375,
-      "epoch": 0.694,
-      "grad_norm": 30.493282980019504,
-      "kl": 5.34375,
-      "learning_rate": 3.3321084665422803e-07,
-      "loss": 1.2128,
-      "reward": 2.4720669984817505,
-      "reward_std": 0.7172423005104065,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.015780417248606682,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 396.9166717529297,
+      "epoch": 0.347,
+      "grad_norm": 65.84497722604661,
+      "kl": 4.5859375,
+      "learning_rate": 8.428483977696328e-07,
+      "loss": 0.9671,
+      "reward": 2.560770630836487,
+      "reward_std": 0.5635119080543518,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.03645172622054815,
+      "rewards/tag_count_reward": 0.9375,
       "step": 694
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.9375305175781,
-      "epoch": 0.695,
-      "grad_norm": 32.828395091508476,
-      "kl": 5.03125,
-      "learning_rate": 3.3183567088914833e-07,
-      "loss": 0.8333,
-      "reward": 2.6700429916381836,
-      "reward_std": 0.5373901128768921,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.02092924155294895,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 381.16668701171875,
+      "epoch": 0.3475,
+      "grad_norm": 18.561448766988402,
+      "kl": 2.12109375,
+      "learning_rate": 8.422516217485825e-07,
+      "loss": 0.6426,
+      "reward": 2.567765712738037,
+      "reward_std": 0.5413838922977448,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.025984477251768112,
+      "rewards/tag_count_reward": 0.96875,
       "step": 695
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 374.8333435058594,
-      "epoch": 0.696,
-      "grad_norm": 13.216131597812279,
-      "kl": 2.4765625,
-      "learning_rate": 3.3046315338757026e-07,
-      "loss": 0.3144,
-      "reward": 2.609209179878235,
-      "reward_std": 0.655095100402832,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.01926326658576727,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "completion_length": 315.6666717529297,
+      "epoch": 0.348,
+      "grad_norm": 12.793725754335805,
+      "kl": 0.51318359375,
+      "learning_rate": 8.416539554784089e-07,
+      "loss": 0.1811,
+      "reward": 2.8265974521636963,
+      "reward_std": 0.3042995482683182,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03451373428106308,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 696
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 318.6666717529297,
-      "epoch": 0.697,
-      "grad_norm": 10.118006800111877,
-      "kl": 0.5234375,
-      "learning_rate": 3.290933108731866e-07,
-      "loss": -0.0809,
-      "reward": 2.3212504386901855,
-      "reward_std": 0.5412976741790771,
-      "rewards/accuracy_reward": 0.3958333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.020763473585247993,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 312.1458435058594,
+      "epoch": 0.3485,
+      "grad_norm": 12.387839093466074,
+      "kl": 1.2890625,
+      "learning_rate": 8.410554007797068e-07,
+      "loss": 0.2936,
+      "reward": 2.6493937969207764,
+      "reward_std": 0.44927215576171875,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.022481410764157772,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 697
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 742.3541870117188,
-      "epoch": 0.698,
-      "grad_norm": 28.054949929765485,
-      "kl": 5.546875,
-      "learning_rate": 3.2772616003709616e-07,
-      "loss": 0.7595,
-      "reward": 2.193854570388794,
-      "reward_std": 0.8353064060211182,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.9375000298023224,
-      "rewards/repetition_penalty_reward": -0.035312142223119736,
-      "rewards/tag_count_reward": 0.7916666865348816,
+      "completion_length": 364.125,
+      "epoch": 0.349,
+      "grad_norm": 21.88721713904914,
+      "kl": 1.349609375,
+      "learning_rate": 8.404559594757777e-07,
+      "loss": 0.2126,
+      "reward": 2.453208088874817,
+      "reward_std": 0.3716874122619629,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03463919833302498,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 698
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 452.37501525878906,
-      "epoch": 0.699,
-      "grad_norm": 41.658117016977116,
-      "kl": 2.1875,
-      "learning_rate": 3.263617175376001e-07,
-      "loss": 0.8256,
-      "reward": 2.7566497325897217,
-      "reward_std": 0.49461202323436737,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.01418376062065363,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 319.54168701171875,
+      "epoch": 0.3495,
+      "grad_norm": 23.304107901739318,
+      "kl": 3.1298828125,
+      "learning_rate": 8.398556333926239e-07,
+      "loss": 0.396,
+      "reward": 2.7863314151763916,
+      "reward_std": 0.40553222596645355,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.040057627484202385,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 699
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 663.5000305175781,
-      "epoch": 0.7,
-      "grad_norm": 26.592270906723318,
-      "kl": 6.0625,
-      "learning_rate": 3.250000000000001e-07,
-      "loss": 0.7409,
-      "reward": 2.101960301399231,
-      "reward_std": 0.7690182626247406,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.02824808470904827,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 347.87501525878906,
+      "epoch": 0.35,
+      "grad_norm": 38.63296720283707,
+      "kl": 3.421875,
+      "learning_rate": 8.392544243589427e-07,
+      "loss": 0.9736,
+      "reward": 2.3871514797210693,
+      "reward_std": 0.5459754467010498,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.027779086492955685,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 700
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 576.25,
-      "epoch": 0.701,
-      "grad_norm": 19.522315074592626,
-      "kl": 4.1015625,
-      "learning_rate": 3.2364102401639423e-07,
-      "loss": 0.6651,
-      "reward": 2.648020029067993,
-      "reward_std": 0.46788084506988525,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.009966201148927212,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 302.5416717529297,
+      "epoch": 0.3505,
+      "grad_norm": 16.95817212965768,
+      "kl": 1.1083984375,
+      "learning_rate": 8.38652334206121e-07,
+      "loss": 0.4275,
+      "reward": 2.909200429916382,
+      "reward_std": 0.14170600473880768,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.04566095769405365,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 701
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 593.4375305175781,
-      "epoch": 0.702,
-      "grad_norm": 21.19308541158298,
-      "kl": 4.4375,
-      "learning_rate": 3.222848061454764e-07,
-      "loss": 0.8651,
-      "reward": 2.0901803970336914,
-      "reward_std": 0.6151553392410278,
-      "rewards/accuracy_reward": 0.3333333432674408,
-      "rewards/reasoning_steps_reward": 0.9444444179534912,
-      "rewards/repetition_penalty_reward": -0.020930853206664324,
-      "rewards/tag_count_reward": 0.8333333730697632,
+      "completion_length": 389.2708435058594,
+      "epoch": 0.351,
+      "grad_norm": 254.15785403417703,
+      "kl": 8.96875,
+      "learning_rate": 8.3804936476823e-07,
+      "loss": 1.3585,
+      "reward": 2.471347212791443,
+      "reward_std": 0.4235590100288391,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.03038894012570381,
+      "rewards/tag_count_reward": 0.953125,
       "step": 702
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 448.5416717529297,
-      "epoch": 0.703,
-      "grad_norm": 14.569396100243361,
-      "kl": 2.4140625,
-      "learning_rate": 3.209313629123329e-07,
-      "loss": 0.3438,
-      "reward": 2.424591898918152,
-      "reward_std": 0.4449867308139801,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.025061040185391903,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 304.0208435058594,
+      "epoch": 0.3515,
+      "grad_norm": 19.759137725479157,
+      "kl": 1.263671875,
+      "learning_rate": 8.374455178820189e-07,
+      "loss": 0.4074,
+      "reward": 2.781951069831848,
+      "reward_std": 0.3059141989797354,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.032285166904330254,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 703
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 571.1458435058594,
-      "epoch": 0.704,
-      "grad_norm": 40.600173749644895,
-      "kl": 6.4453125,
-      "learning_rate": 3.195807108082429e-07,
-      "loss": 0.9898,
-      "reward": 2.4354113340377808,
-      "reward_std": 0.6632784008979797,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.014241641853004694,
-      "rewards/tag_count_reward": 0.8385416865348816,
+      "completion_length": 400.4583435058594,
+      "epoch": 0.352,
+      "grad_norm": 87.69689094374543,
+      "kl": 7.359375,
+      "learning_rate": 8.368407953869103e-07,
+      "loss": 0.4764,
+      "reward": 2.40726637840271,
+      "reward_std": 0.5606774091720581,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.031969884410500526,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 704
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 480.1458435058594,
-      "epoch": 0.705,
-      "grad_norm": 39.53208480397733,
-      "kl": 3.4765625,
-      "learning_rate": 3.182328662904756e-07,
-      "loss": 0.7597,
-      "reward": 2.5596699714660645,
-      "reward_std": 0.6091938018798828,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9791667461395264,
-      "rewards/repetition_penalty_reward": -0.02366339974105358,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 365.6041717529297,
+      "epoch": 0.3525,
+      "grad_norm": 55.10624289780696,
+      "kl": 2.1630859375,
+      "learning_rate": 8.362351991249937e-07,
+      "loss": 0.3022,
+      "reward": 2.5830687284469604,
+      "reward_std": 0.543460875749588,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.04193143546581268,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 705
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 376.5416717529297,
-      "epoch": 0.706,
-      "grad_norm": 26.171445195886303,
-      "kl": 2.21484375,
-      "learning_rate": 3.168878457820915e-07,
-      "loss": 0.5183,
-      "reward": 2.515265941619873,
-      "reward_std": 0.4705822169780731,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.017720218747854233,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 370.1458435058594,
+      "epoch": 0.353,
+      "grad_norm": 38.93542768766942,
+      "kl": 1.20703125,
+      "learning_rate": 8.356287309410204e-07,
+      "loss": 0.4547,
+      "reward": 2.704145312309265,
+      "reward_std": 0.5362651348114014,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.037174249067902565,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 706
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 421.41668701171875,
-      "epoch": 0.707,
-      "grad_norm": 23.58547734465505,
-      "kl": 5.1875,
-      "learning_rate": 3.155456656717408e-07,
-      "loss": 0.9749,
-      "reward": 2.6543229818344116,
-      "reward_std": 0.67611363530159,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.010607585310935974,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 443.0208435058594,
+      "epoch": 0.3535,
+      "grad_norm": 32.744332762719786,
+      "kl": 0.75,
+      "learning_rate": 8.350213926823974e-07,
+      "loss": 0.44,
+      "reward": 2.5751278400421143,
+      "reward_std": 0.5337015986442566,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.04987236298620701,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 707
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 420.1458435058594,
-      "epoch": 0.708,
-      "grad_norm": 22.53505575124777,
-      "kl": 3.265625,
-      "learning_rate": 3.142063423134644e-07,
-      "loss": 0.5001,
-      "reward": 2.434324622154236,
-      "reward_std": 0.42964955419301987,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.01706432970240712,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 372.5625,
+      "epoch": 0.354,
+      "grad_norm": 42.03312426418706,
+      "kl": 0.68359375,
+      "learning_rate": 8.344131861991828e-07,
+      "loss": 0.4816,
+      "reward": 2.814732551574707,
+      "reward_std": 0.4671812057495117,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.04117002338171005,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 708
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 604.5000305175781,
-      "epoch": 0.709,
-      "grad_norm": 100.60447189912114,
-      "kl": 10.71875,
-      "learning_rate": 3.1286989202649503e-07,
-      "loss": 1.2098,
-      "reward": 2.214650869369507,
-      "reward_std": 0.7704206705093384,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 0.92361119389534,
-      "rewards/repetition_penalty_reward": -0.021460428833961487,
-      "rewards/tag_count_reward": 0.7916666865348816,
+      "completion_length": 402.8333435058594,
+      "epoch": 0.3545,
+      "grad_norm": 14.798336180101453,
+      "kl": 2.26171875,
+      "learning_rate": 8.338041133440788e-07,
+      "loss": 0.6263,
+      "reward": 2.585590362548828,
+      "reward_std": 0.5589672327041626,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.03593753091990948,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 709
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 441.5208435058594,
-      "epoch": 0.71,
-      "grad_norm": 62.885652673494626,
-      "kl": 4.6171875,
-      "learning_rate": 3.115363310950578e-07,
-      "loss": 0.6795,
-      "reward": 2.4361919164657593,
-      "reward_std": 0.6633725464344025,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.022141442634165287,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 484.1666717529297,
+      "epoch": 0.355,
+      "grad_norm": 329.43975783332286,
+      "kl": 9.34375,
+      "learning_rate": 8.331941759724268e-07,
+      "loss": 1.4746,
+      "reward": 2.464582681655884,
+      "reward_std": 0.6360695958137512,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.03541737608611584,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 710
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 505.9791717529297,
-      "epoch": 0.711,
-      "grad_norm": 91.4316669194455,
-      "kl": 6.65625,
-      "learning_rate": 3.102056757681715e-07,
-      "loss": 1.2926,
-      "reward": 2.3233230113983154,
-      "reward_std": 0.7938812673091888,
+      "completion_length": 458.4791717529297,
+      "epoch": 0.3555,
+      "grad_norm": 367.74155787665666,
+      "kl": 11.8759765625,
+      "learning_rate": 8.325833759422021e-07,
+      "loss": 1.0719,
+      "reward": 2.316919684410095,
+      "reward_std": 0.4943684861063957,
       "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.9375001192092896,
-      "rewards/repetition_penalty_reward": -0.01001034933142364,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.875,
+      "rewards/repetition_penalty_reward": -0.032038720324635506,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 711
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 497.7708435058594,
-      "epoch": 0.712,
-      "grad_norm": 49.94435657649251,
-      "kl": 5.265625,
-      "learning_rate": 3.0887794225945143e-07,
-      "loss": 0.7182,
-      "reward": 2.6127805709838867,
-      "reward_std": 0.5191345363855362,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.019164070021361113,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 483.7916717529297,
+      "epoch": 0.356,
+      "grad_norm": 47.212787106240015,
+      "kl": 4.234375,
+      "learning_rate": 8.319717151140072e-07,
+      "loss": 1.1337,
+      "reward": 2.351717948913574,
+      "reward_std": 0.6439139246940613,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.04758777469396591,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 712
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 408.9791717529297,
-      "epoch": 0.713,
-      "grad_norm": 24.539952113509884,
-      "kl": 3.46875,
-      "learning_rate": 3.075531467469116e-07,
-      "loss": 0.5752,
-      "reward": 2.349856376647949,
-      "reward_std": 0.4715754985809326,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.023407643660902977,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 324.25,
+      "epoch": 0.3565,
+      "grad_norm": 3.183005147595655,
+      "kl": 0.24560546875,
+      "learning_rate": 8.313591953510673e-07,
+      "loss": -0.005,
+      "reward": 2.9382745027542114,
+      "reward_std": 0.0876419385895133,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.04089214652776718,
+      "rewards/tag_count_reward": 1.0,
       "step": 713
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 506.7708435058594,
-      "epoch": 0.714,
-      "grad_norm": 32.90722497302991,
-      "kl": 4.640625,
-      "learning_rate": 3.062313053727671e-07,
-      "loss": 0.8249,
-      "reward": 2.2763147354125977,
-      "reward_std": 0.7002745866775513,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9513889849185944,
-      "rewards/repetition_penalty_reward": -0.013615832198411226,
-      "rewards/tag_count_reward": 0.859375,
+      "completion_length": 322.5208435058594,
+      "epoch": 0.357,
+      "grad_norm": 21.014869886703735,
+      "kl": 0.328125,
+      "learning_rate": 8.307458185192238e-07,
+      "loss": 0.259,
+      "reward": 2.6625412702560425,
+      "reward_std": 0.37747399508953094,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.035375405102968216,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 714
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 362.0833435058594,
-      "epoch": 0.715,
-      "grad_norm": 27.678515846159396,
-      "kl": 2.14453125,
-      "learning_rate": 3.0491243424323783e-07,
-      "loss": 0.3263,
-      "reward": 2.8031435012817383,
-      "reward_std": 0.45686857402324677,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.023245446383953094,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 349.4375,
+      "epoch": 0.3575,
+      "grad_norm": 20.703256092729664,
+      "kl": 0.3984375,
+      "learning_rate": 8.301315864869289e-07,
+      "loss": 0.2191,
+      "reward": 2.67917537689209,
+      "reward_std": 0.3260675296187401,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.06908857077360153,
+      "rewards/tag_count_reward": 0.984375,
       "step": 715
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 466.6666717529297,
-      "epoch": 0.716,
-      "grad_norm": 29.560245962070923,
-      "kl": 3.8515625,
-      "learning_rate": 3.0359654942835247e-07,
-      "loss": 0.8846,
-      "reward": 2.3151395320892334,
-      "reward_std": 0.7553818225860596,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.03381876181811094,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 356.5208435058594,
+      "epoch": 0.358,
+      "grad_norm": 9.846153349140325,
+      "kl": 0.90625,
+      "learning_rate": 8.295165011252396e-07,
+      "loss": 0.1475,
+      "reward": 2.4251718521118164,
+      "reward_std": 0.41564081609249115,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.04878660198301077,
+      "rewards/tag_count_reward": 0.9739583730697632,
       "step": 716
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 677.5625,
-      "epoch": 0.717,
-      "grad_norm": 28.56711039404196,
-      "kl": 6.859375,
-      "learning_rate": 3.02283666961752e-07,
-      "loss": 0.8948,
-      "reward": 2.116614818572998,
-      "reward_std": 0.61385178565979,
-      "rewards/accuracy_reward": 0.3750000223517418,
-      "rewards/reasoning_steps_reward": 0.958333432674408,
-      "rewards/repetition_penalty_reward": -0.018802037462592125,
-      "rewards/tag_count_reward": 0.8020833730697632,
+      "completion_length": 426.5625,
+      "epoch": 0.3585,
+      "grad_norm": 22.268351473633047,
+      "kl": 2.8046875,
+      "learning_rate": 8.289005643078131e-07,
+      "loss": 1.0537,
+      "reward": 2.636265277862549,
+      "reward_std": 0.5826082229614258,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.04602637514472008,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 717
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 651.3125305175781,
-      "epoch": 0.718,
-      "grad_norm": 28.83560352945601,
-      "kl": 6.46875,
-      "learning_rate": 3.0097380284049523e-07,
-      "loss": 1.0283,
-      "reward": 2.1972588300704956,
-      "reward_std": 0.6690528243780136,
-      "rewards/accuracy_reward": 0.4791666716337204,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.014546706806868315,
-      "rewards/tag_count_reward": 0.7812500298023224,
+      "completion_length": 345.04168701171875,
+      "epoch": 0.359,
+      "grad_norm": 49.628296082541894,
+      "kl": 2.771484375,
+      "learning_rate": 8.282837779108993e-07,
+      "loss": 0.5828,
+      "reward": 2.618976354598999,
+      "reward_std": 0.40643376111984253,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.044218238443136215,
+      "rewards/tag_count_reward": 0.96875,
       "step": 718
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 389.4375,
-      "epoch": 0.719,
-      "grad_norm": 10.934491674221926,
-      "kl": 1.875,
-      "learning_rate": 2.996669730248628e-07,
-      "loss": 0.2297,
-      "reward": 2.1007025241851807,
-      "reward_std": 0.48380589485168457,
-      "rewards/accuracy_reward": 0.2291666716337204,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.017353057861328125,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 376.6041717529297,
+      "epoch": 0.3595,
+      "grad_norm": 17.745954704948424,
+      "kl": 1.49609375,
+      "learning_rate": 8.276661438133368e-07,
+      "loss": 0.4342,
+      "reward": 2.6793724298477173,
+      "reward_std": 0.35265351831912994,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03937762975692749,
+      "rewards/tag_count_reward": 0.9895833730697632,
       "step": 719
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 394.1041717529297,
-      "epoch": 0.72,
-      "grad_norm": 29.82240224098748,
-      "kl": 3.19921875,
-      "learning_rate": 2.9836319343816397e-07,
-      "loss": 0.6818,
-      "reward": 2.698520064353943,
-      "reward_std": 0.6881579607725143,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.011549403425306082,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 353.66668701171875,
+      "epoch": 0.36,
+      "grad_norm": 27.380512621837557,
+      "kl": 0.939453125,
+      "learning_rate": 8.270476638965461e-07,
+      "loss": 0.5235,
+      "reward": 2.807437300682068,
+      "reward_std": 0.43915510177612305,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03978483937680721,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 720
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 467.0208435058594,
-      "epoch": 0.721,
-      "grad_norm": 25.761812478587363,
-      "kl": 3.8984375,
-      "learning_rate": 2.9706247996654134e-07,
-      "loss": 0.5726,
-      "reward": 2.4670268297195435,
-      "reward_std": 0.6666984856128693,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.0173482783138752,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 389.6875,
+      "epoch": 0.3605,
+      "grad_norm": 24.933099331429055,
+      "kl": 0.765625,
+      "learning_rate": 8.264283400445243e-07,
+      "loss": 0.4441,
+      "reward": 2.6986876726150513,
+      "reward_std": 0.5908840298652649,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.06346515193581581,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 721
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 459.50001525878906,
-      "epoch": 0.722,
-      "grad_norm": 66.05822501836576,
-      "kl": 4.71875,
-      "learning_rate": 2.9576484845877793e-07,
-      "loss": 0.7149,
-      "reward": 2.672752857208252,
-      "reward_std": 0.6598548293113708,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.02516369242221117,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 427.4583435058594,
+      "epoch": 0.361,
+      "grad_norm": 21.155555874267726,
+      "kl": 1.40625,
+      "learning_rate": 8.258081741438394e-07,
+      "loss": 0.6891,
+      "reward": 2.5299737453460693,
+      "reward_std": 0.5110133141279221,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.06377626396715641,
+      "rewards/tag_count_reward": 0.96875,
       "step": 722
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 469.5,
-      "epoch": 0.723,
-      "grad_norm": 33.50765624110453,
-      "kl": 3.296875,
-      "learning_rate": 2.944703147261046e-07,
-      "loss": 1.0137,
-      "reward": 2.1091710329055786,
-      "reward_std": 0.5909655094146729,
-      "rewards/accuracy_reward": 0.2916666716337204,
-      "rewards/reasoning_steps_reward": 0.9305555522441864,
-      "rewards/repetition_penalty_reward": -0.008884491166099906,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 358.2708435058594,
+      "epoch": 0.3615,
+      "grad_norm": 22.1107293389388,
+      "kl": 1.6025390625,
+      "learning_rate": 8.25187168083624e-07,
+      "loss": 0.3977,
+      "reward": 2.5182286500930786,
+      "reward_std": 0.3348126895725727,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.047743676230311394,
+      "rewards/tag_count_reward": 0.96875,
       "step": 723
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 475.9583435058594,
-      "epoch": 0.724,
-      "grad_norm": 32.701475311916404,
-      "kl": 4.078125,
-      "learning_rate": 2.931788945420058e-07,
-      "loss": 1.2,
-      "reward": 2.4817965030670166,
-      "reward_std": 0.5851520895957947,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.012995375785976648,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 336.125,
+      "epoch": 0.362,
+      "grad_norm": 7.252900143586577,
+      "kl": 0.908203125,
+      "learning_rate": 8.245653237555705e-07,
+      "loss": 0.2413,
+      "reward": 2.539818525314331,
+      "reward_std": 0.4269237220287323,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.04698727838695049,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 724
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 554.7083435058594,
-      "epoch": 0.725,
-      "grad_norm": 32.79842010966518,
-      "kl": 5.91796875,
-      "learning_rate": 2.918906036420294e-07,
-      "loss": 0.8106,
-      "reward": 2.432934522628784,
-      "reward_std": 0.6692648828029633,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.01671838667243719,
-      "rewards/tag_count_reward": 0.859375,
-      "step": 725
-    },
-    {
+      "completion_length": 460.52085876464844,
+      "epoch": 0.3625,
+      "grad_norm": 35.246874774684656,
+      "kl": 3.2578125,
+      "learning_rate": 8.239426430539243e-07,
+      "loss": 0.9074,
+      "reward": 2.555440068244934,
+      "reward_std": 0.6377770006656647,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.05914325639605522,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 725
+    },
+    {
       "clip_ratio": 0.0,
-      "completion_length": 572.5208435058594,
-      "epoch": 0.726,
-      "grad_norm": 34.07001019061104,
-      "kl": 4.34375,
-      "learning_rate": 2.9060545772359305e-07,
-      "loss": 0.985,
-      "reward": 2.4105674028396606,
-      "reward_std": 0.5044368803501129,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02866880688816309,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 350.3125,
+      "epoch": 0.363,
+      "grad_norm": 9.204556850773619,
+      "kl": 1.97265625,
+      "learning_rate": 8.23319127875479e-07,
+      "loss": 0.4855,
+      "reward": 2.772892713546753,
+      "reward_std": 0.43467026948928833,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.055232321843504906,
+      "rewards/tag_count_reward": 0.953125,
       "step": 726
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 485.41668701171875,
-      "epoch": 0.727,
-      "grad_norm": 22.090198544561794,
-      "kl": 3.53125,
-      "learning_rate": 2.893234724457946e-07,
-      "loss": 0.662,
-      "reward": 2.4937909841537476,
-      "reward_std": 0.6231936812400818,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.027042274363338947,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 363.9166717529297,
+      "epoch": 0.3635,
+      "grad_norm": 6.6860563949651475,
+      "kl": 1.0283203125,
+      "learning_rate": 8.226947801195699e-07,
+      "loss": 0.1283,
+      "reward": 2.6459619998931885,
+      "reward_std": 0.2583989417180419,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.057163089513778687,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 727
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 596.9375,
-      "epoch": 0.728,
-      "grad_norm": 39.74856940380198,
-      "kl": 5.5,
-      "learning_rate": 2.8804466342921987e-07,
-      "loss": 1.4125,
-      "reward": 2.3083548545837402,
-      "reward_std": 0.7532355189323425,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9513888657093048,
-      "rewards/repetition_penalty_reward": -0.02324253274127841,
-      "rewards/tag_count_reward": 0.8177083432674408,
+      "completion_length": 261.0416717529297,
+      "epoch": 0.364,
+      "grad_norm": 3.671168469586843,
+      "kl": 0.251953125,
+      "learning_rate": 8.220696016880687e-07,
+      "loss": 0.0623,
+      "reward": 2.95377254486084,
+      "reward_std": 0.04473966360092163,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.03233869280666113,
+      "rewards/tag_count_reward": 1.0,
       "step": 728
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 644.2708587646484,
-      "epoch": 0.729,
-      "grad_norm": 27.585966675541705,
-      "kl": 5.046875,
-      "learning_rate": 2.86769046255753e-07,
-      "loss": 0.8454,
-      "reward": 2.0349258184432983,
-      "reward_std": 0.7431914508342743,
-      "rewards/accuracy_reward": 0.375,
-      "rewards/reasoning_steps_reward": 0.9236111640930176,
-      "rewards/repetition_penalty_reward": -0.024102029390633106,
-      "rewards/tag_count_reward": 0.7604166865348816,
+      "completion_length": 476.3333435058594,
+      "epoch": 0.3645,
+      "grad_norm": 33.63088386822394,
+      "kl": 0.951171875,
+      "learning_rate": 8.21443594485377e-07,
+      "loss": 0.5476,
+      "reward": 2.48537540435791,
+      "reward_std": 0.7262991070747375,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9722221791744232,
+      "rewards/repetition_penalty_reward": -0.08580523729324341,
+      "rewards/tag_count_reward": 0.953125,
       "step": 729
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 416.8958435058594,
-      "epoch": 0.73,
-      "grad_norm": 38.15694283580282,
-      "kl": 1.8203125,
-      "learning_rate": 2.854966364683872e-07,
-      "loss": 0.4531,
-      "reward": 2.5842431783676147,
-      "reward_std": 0.4230002462863922,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.021659548394382,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 392.16668701171875,
+      "epoch": 0.365,
+      "grad_norm": 30.396775140349977,
+      "kl": 0.5615234375,
+      "learning_rate": 8.208167604184217e-07,
+      "loss": 0.3701,
+      "reward": 2.5822075605392456,
+      "reward_std": 0.48322977125644684,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.07404248043894768,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 730
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 522.5416870117188,
-      "epoch": 0.731,
-      "grad_norm": 19.973765241704356,
-      "kl": 3.59375,
-      "learning_rate": 2.842274495710335e-07,
-      "loss": 0.5592,
-      "reward": 2.3725064992904663,
-      "reward_std": 0.5484490990638733,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.016382555477321148,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 446.97918701171875,
+      "epoch": 0.3655,
+      "grad_norm": 39.11126689033728,
+      "kl": 1.0234375,
+      "learning_rate": 8.201891013966478e-07,
+      "loss": 0.5577,
+      "reward": 2.379446864128113,
+      "reward_std": 0.5678260624408722,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.08062271773815155,
+      "rewards/tag_count_reward": 0.953125,
       "step": 731
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 775.3125305175781,
-      "epoch": 0.732,
-      "grad_norm": 19.54218927587623,
-      "kl": 8.140625,
-      "learning_rate": 2.829615010283344e-07,
-      "loss": 1.0839,
-      "reward": 2.0963175296783447,
-      "reward_std": 0.8241714835166931,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.06340477429330349,
-      "rewards/tag_count_reward": 0.7291666865348816,
+      "completion_length": 537.5416870117188,
+      "epoch": 0.366,
+      "grad_norm": 15.068758274769461,
+      "kl": 1.6875,
+      "learning_rate": 8.195606193320136e-07,
+      "loss": 0.6393,
+      "reward": 2.3407026529312134,
+      "reward_std": 0.7115332186222076,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.07075561210513115,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 732
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 618.8958435058594,
-      "epoch": 0.733,
-      "grad_norm": 15.706934322980523,
-      "kl": 5.859375,
-      "learning_rate": 2.8169880626547283e-07,
-      "loss": 0.8039,
-      "reward": 2.311804175376892,
-      "reward_std": 0.7484097182750702,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.033682181499898434,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 421.2708435058594,
+      "epoch": 0.3665,
+      "grad_norm": 9.730085638274787,
+      "kl": 1.34375,
+      "learning_rate": 8.189313161389844e-07,
+      "loss": 0.4843,
+      "reward": 2.605613946914673,
+      "reward_std": 0.44479694962501526,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.035011230036616325,
+      "rewards/tag_count_reward": 0.953125,
       "step": 733
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 640.8541870117188,
-      "epoch": 0.734,
-      "grad_norm": 45.104210914901735,
-      "kl": 4.265625,
-      "learning_rate": 2.8043938066798645e-07,
-      "loss": 1.1408,
-      "reward": 2.3691216707229614,
-      "reward_std": 0.8749706745147705,
-      "rewards/accuracy_reward": 0.6458333730697632,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.01976742222905159,
-      "rewards/tag_count_reward": 0.8125,
+      "completion_length": 562.0000305175781,
+      "epoch": 0.367,
+      "grad_norm": 36.417464044540424,
+      "kl": 4.1171875,
+      "learning_rate": 8.183011937345271e-07,
+      "loss": 1.1908,
+      "reward": 2.5398595333099365,
+      "reward_std": 0.769398033618927,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.06430716440081596,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 734
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 556.1041717529297,
-      "epoch": 0.735,
-      "grad_norm": 25.06133634060821,
-      "kl": 4.1171875,
-      "learning_rate": 2.791832395815782e-07,
-      "loss": 0.901,
-      "reward": 2.317103087902069,
-      "reward_std": 0.6717338263988495,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.030119193717837334,
-      "rewards/tag_count_reward": 0.8333333730697632,
+      "completion_length": 439.6875,
+      "epoch": 0.3675,
+      "grad_norm": 30.83756444127917,
+      "kl": 2.984375,
+      "learning_rate": 8.176702540381036e-07,
+      "loss": 0.9181,
+      "reward": 2.7064634561538696,
+      "reward_std": 0.5320930778980255,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.03832832910120487,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 735
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 541.4166870117188,
-      "epoch": 0.736,
-      "grad_norm": 70.06964895356546,
-      "kl": 4.8828125,
-      "learning_rate": 2.7793039831193133e-07,
-      "loss": 0.6714,
-      "reward": 2.38431978225708,
-      "reward_std": 0.4595172256231308,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.9791666567325592,
-      "rewards/repetition_penalty_reward": -0.016721924766898155,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 330.7916717529297,
+      "epoch": 0.368,
+      "grad_norm": 13.454778716975406,
+      "kl": 1.0029296875,
+      "learning_rate": 8.170384989716657e-07,
+      "loss": 0.3605,
+      "reward": 2.8519721031188965,
+      "reward_std": 0.31362488865852356,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03691680170595646,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 736
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 563.3541870117188,
-      "epoch": 0.737,
-      "grad_norm": 37.75513345672942,
-      "kl": 3.3359375,
-      "learning_rate": 2.766808721245211e-07,
-      "loss": 1.1112,
-      "reward": 2.4378621578216553,
-      "reward_std": 0.59513920545578,
-      "rewards/accuracy_reward": 0.6041666716337204,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.016999011393636465,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 395.7708435058594,
+      "epoch": 0.3685,
+      "grad_norm": 13.197505616793235,
+      "kl": 2.11328125,
+      "learning_rate": 8.164059304596488e-07,
+      "loss": 0.25,
+      "reward": 2.556104898452759,
+      "reward_std": 0.37108321487903595,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.04632566310465336,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 737
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 531.875,
-      "epoch": 0.738,
-      "grad_norm": 18.56617973802945,
-      "kl": 4.125,
-      "learning_rate": 2.7543467624442956e-07,
-      "loss": 0.8568,
-      "reward": 2.246809482574463,
-      "reward_std": 0.6611096858978271,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.017079456709325314,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 365.2708435058594,
+      "epoch": 0.369,
+      "grad_norm": 9.749024221023307,
+      "kl": 2.08203125,
+      "learning_rate": 8.157725504289664e-07,
+      "loss": 0.3044,
+      "reward": 2.4535757303237915,
+      "reward_std": 0.558014452457428,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.018646665383130312,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 738
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.2708435058594,
-      "epoch": 0.739,
-      "grad_norm": 23.972621042263274,
-      "kl": 3.859375,
-      "learning_rate": 2.741918258561607e-07,
-      "loss": 1.0541,
-      "reward": 2.4821611642837524,
-      "reward_std": 0.4908117651939392,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.03520001098513603,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 479.2083435058594,
+      "epoch": 0.3695,
+      "grad_norm": 11.39083083892124,
+      "kl": 2.03125,
+      "learning_rate": 8.151383608090039e-07,
+      "loss": 0.635,
+      "reward": 2.229800820350647,
+      "reward_std": 0.5332637131214142,
+      "rewards/accuracy_reward": 0.4375000223517418,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.03929641842842102,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 739
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.1458587646484,
-      "epoch": 0.74,
-      "grad_norm": 40.196471457608176,
-      "kl": 6.765625,
-      "learning_rate": 2.729523361034538e-07,
-      "loss": 1.2053,
-      "reward": 2.3264917135238647,
-      "reward_std": 0.6097930297255516,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.022466725669801235,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 464.5625305175781,
+      "epoch": 0.37,
+      "grad_norm": 16.878206896706008,
+      "kl": 2.859375,
+      "learning_rate": 8.145033635316128e-07,
+      "loss": 0.6706,
+      "reward": 2.367822289466858,
+      "reward_std": 0.7059852480888367,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.031483279541134834,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 740
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 493.04168701171875,
-      "epoch": 0.741,
-      "grad_norm": 29.927610585853564,
-      "kl": 4.03515625,
-      "learning_rate": 2.717162220891007e-07,
-      "loss": 0.9692,
-      "reward": 2.016554594039917,
-      "reward_std": 0.4730721116065979,
-      "rewards/accuracy_reward": 0.16666667722165585,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.023376007564365864,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 343.375,
+      "epoch": 0.3705,
+      "grad_norm": 4.401882005610615,
+      "kl": 0.7861328125,
+      "learning_rate": 8.138675605311051e-07,
+      "loss": 0.1554,
+      "reward": 2.771618604660034,
+      "reward_std": 0.2334722802042961,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.0460897758603096,
+      "rewards/tag_count_reward": 0.984375,
       "step": 741
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 512.9583435058594,
-      "epoch": 0.742,
-      "grad_norm": 25.41926623083724,
-      "kl": 4.2734375,
-      "learning_rate": 2.7048349887476037e-07,
-      "loss": 1.0409,
-      "reward": 2.418197512626648,
-      "reward_std": 0.6901432275772095,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.03492756187915802,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 359.4166717529297,
+      "epoch": 0.371,
+      "grad_norm": 4.239738905969326,
+      "kl": 1.05078125,
+      "learning_rate": 8.13230953744247e-07,
+      "loss": 0.1471,
+      "reward": 2.5801262855529785,
+      "reward_std": 0.2986167371273041,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.03619310073554516,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 742
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 648.3958435058594,
-      "epoch": 0.743,
-      "grad_norm": 28.016549982017455,
-      "kl": 6.734375,
-      "learning_rate": 2.692541814807763e-07,
-      "loss": 1.1588,
-      "reward": 2.3356810808181763,
-      "reward_std": 0.7242786288261414,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.021957868244498968,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 389.3333435058594,
+      "epoch": 0.3715,
+      "grad_norm": 7.057211069031311,
+      "kl": 0.87109375,
+      "learning_rate": 8.125935451102528e-07,
+      "loss": 0.4242,
+      "reward": 2.3998607397079468,
+      "reward_std": 0.4749663770198822,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03763941116631031,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 743
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 528.0000152587891,
-      "epoch": 0.744,
-      "grad_norm": 93.21543758888738,
-      "kl": 9.375,
-      "learning_rate": 2.6802828488599294e-07,
-      "loss": 1.1425,
-      "reward": 2.218702793121338,
-      "reward_std": 0.743184506893158,
-      "rewards/accuracy_reward": 0.458333358168602,
-      "rewards/reasoning_steps_reward": 0.9375000298023224,
-      "rewards/repetition_penalty_reward": -0.02608906291425228,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 317.6458435058594,
+      "epoch": 0.372,
+      "grad_norm": 5.779641128334165,
+      "kl": 0.638671875,
+      "learning_rate": 8.119553365707802e-07,
+      "loss": 0.2267,
+      "reward": 2.8257747888565063,
+      "reward_std": 0.4051903337240219,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.026655779220163822,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 744
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 522.5833587646484,
-      "epoch": 0.745,
-      "grad_norm": 29.592772740929842,
-      "kl": 5.4375,
-      "learning_rate": 2.6680582402757324e-07,
-      "loss": 0.8376,
-      "reward": 2.1133170127868652,
-      "reward_std": 0.4543229639530182,
-      "rewards/accuracy_reward": 0.27083333395421505,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.020363647490739822,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 410.9166717529297,
+      "epoch": 0.3725,
+      "grad_norm": 3.66973899046251,
+      "kl": 0.517578125,
+      "learning_rate": 8.113163300699228e-07,
+      "loss": 0.1316,
+      "reward": 2.464886784553528,
+      "reward_std": 0.3369094356894493,
+      "rewards/accuracy_reward": 0.5625000223517418,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.038585495203733444,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 745
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 537.7083435058594,
-      "epoch": 0.746,
-      "grad_norm": 52.43685846686619,
-      "kl": 7.9375,
-      "learning_rate": 2.655868138008171e-07,
-      "loss": 1.3022,
-      "reward": 2.560097575187683,
-      "reward_std": 0.6881845593452454,
+      "completion_length": 314.81251525878906,
+      "epoch": 0.373,
+      "grad_norm": 4.867508354023929,
+      "kl": 0.4296875,
+      "learning_rate": 8.106765275542053e-07,
+      "loss": 0.1155,
+      "reward": 2.6931064128875732,
+      "reward_std": 0.40478771924972534,
       "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.01802760176360607,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0412687249481678,
+      "rewards/tag_count_reward": 0.984375,
       "step": 746
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 491.4583435058594,
-      "epoch": 0.747,
-      "grad_norm": 42.17070883402169,
-      "kl": 4.8125,
-      "learning_rate": 2.6437126905897967e-07,
-      "loss": 0.9389,
-      "reward": 2.4418286085128784,
-      "reward_std": 0.479451060295105,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.023449244908988476,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 318.7291717529297,
+      "epoch": 0.3735,
+      "grad_norm": 4.225560833503795,
+      "kl": 0.349609375,
+      "learning_rate": 8.100359309725774e-07,
+      "loss": 0.0563,
+      "reward": 2.715116500854492,
+      "reward_std": 0.3557048738002777,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03835564851760864,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 747
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 495.89585876464844,
-      "epoch": 0.748,
-      "grad_norm": 42.23947257797948,
-      "kl": 3.734375,
-      "learning_rate": 2.631592046130896e-07,
-      "loss": 0.8408,
-      "reward": 2.2848896980285645,
-      "reward_std": 0.6534326076507568,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.029346639290452003,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 364.68751525878906,
+      "epoch": 0.374,
+      "grad_norm": 5.832499305507599,
+      "kl": 0.7421875,
+      "learning_rate": 8.093945422764069e-07,
+      "loss": 0.2438,
+      "reward": 2.5478862524032593,
+      "reward_std": 0.48231200873851776,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03891943581402302,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 748
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.9166717529297,
-      "epoch": 0.749,
-      "grad_norm": 28.13462040337534,
-      "kl": 4.9921875,
-      "learning_rate": 2.6195063523177e-07,
-      "loss": 0.6576,
-      "reward": 2.5075308084487915,
-      "reward_std": 0.4372696727514267,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9444445073604584,
-      "rewards/repetition_penalty_reward": -0.020247011445462704,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 513.2916717529297,
+      "epoch": 0.3745,
+      "grad_norm": 18.59065925947558,
+      "kl": 1.67578125,
+      "learning_rate": 8.087523634194754e-07,
+      "loss": 0.7216,
+      "reward": 2.548434257507324,
+      "reward_std": 0.6094189584255219,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.04010744206607342,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 749
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 550.9791870117188,
-      "epoch": 0.75,
-      "grad_norm": 51.58149030204615,
-      "kl": 4.734375,
-      "learning_rate": 2.6074557564105724e-07,
-      "loss": 1.2626,
-      "reward": 2.304089069366455,
-      "reward_std": 0.65184286236763,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.010147055611014366,
-      "rewards/tag_count_reward": 0.8489583730697632,
+      "completion_length": 322.60418701171875,
+      "epoch": 0.375,
+      "grad_norm": 7.99311552783378,
+      "kl": 1.29296875,
+      "learning_rate": 8.081093963579707e-07,
+      "loss": 0.2852,
+      "reward": 2.577797532081604,
+      "reward_std": 0.4589303731918335,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.05241105332970619,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 750
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 414.5208435058594,
-      "epoch": 0.751,
-      "grad_norm": 25.50420977958888,
-      "kl": 2.85546875,
-      "learning_rate": 2.595440405242222e-07,
-      "loss": 0.4946,
-      "reward": 2.3329780101776123,
-      "reward_std": 0.4813043922185898,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.01424429938197136,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 367.8333435058594,
+      "epoch": 0.3755,
+      "grad_norm": 7.024161119025017,
+      "kl": 0.73779296875,
+      "learning_rate": 8.074656430504823e-07,
+      "loss": 0.1466,
+      "reward": 2.575978994369507,
+      "reward_std": 0.32069287449121475,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03339605592191219,
+      "rewards/tag_count_reward": 0.984375,
       "step": 751
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 549.7708587646484,
-      "epoch": 0.752,
-      "grad_norm": 30.257524874531967,
-      "kl": 4.296875,
-      "learning_rate": 2.583460445215911e-07,
-      "loss": 1.0795,
-      "reward": 2.672580599784851,
-      "reward_std": 0.6382189244031906,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.013183359056711197,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 360.93751525878906,
+      "epoch": 0.376,
+      "grad_norm": 13.063738129746366,
+      "kl": 0.5400390625,
+      "learning_rate": 8.068211054579943e-07,
+      "loss": 0.322,
+      "reward": 2.6451334953308105,
+      "reward_std": 0.4438292682170868,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.026741638779640198,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 752
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 409.50001525878906,
-      "epoch": 0.753,
-      "grad_norm": 42.99616520034882,
-      "kl": 3.15625,
-      "learning_rate": 2.571516022303671e-07,
-      "loss": 0.8373,
-      "reward": 2.759161591529846,
-      "reward_std": 0.7119235694408417,
-      "rewards/accuracy_reward": 0.8958333432674408,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.02382473973557353,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 420.2083435058594,
+      "epoch": 0.3765,
+      "grad_norm": 23.053449901319926,
+      "kl": 1.609375,
+      "learning_rate": 8.061757855438799e-07,
+      "loss": 0.4878,
+      "reward": 2.5872808694839478,
+      "reward_std": 0.6218395531177521,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.04292759019881487,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 753
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 536.5625305175781,
-      "epoch": 0.754,
-      "grad_norm": 31.379188639299347,
-      "kl": 5.7109375,
-      "learning_rate": 2.5596072820445254e-07,
-      "loss": 1.1123,
-      "reward": 2.451164484024048,
-      "reward_std": 0.6144693195819855,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.909722238779068,
-      "rewards/repetition_penalty_reward": -0.01584968389943242,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 393.9583435058594,
+      "epoch": 0.377,
+      "grad_norm": 18.602852015242384,
+      "kl": 1.9296875,
+      "learning_rate": 8.055296852738956e-07,
+      "loss": 0.4041,
+      "reward": 2.7404640913009644,
+      "reward_std": 0.3876974582672119,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.05120253935456276,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 754
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 489.0416717529297,
-      "epoch": 0.755,
-      "grad_norm": 45.223755625235505,
-      "kl": 4.796875,
-      "learning_rate": 2.547734369542718e-07,
-      "loss": 1.2234,
-      "reward": 2.4944331645965576,
-      "reward_std": 0.610102117061615,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.012511319015175104,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 279.62501525878906,
+      "epoch": 0.3775,
+      "grad_norm": 3.7238210341731244,
+      "kl": 0.24462890625,
+      "learning_rate": 8.048828066161747e-07,
+      "loss": 0.0714,
+      "reward": 2.776410698890686,
+      "reward_std": 0.1331208571791649,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.02914472296833992,
+      "rewards/tag_count_reward": 1.0,
       "step": 755
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 443.12501525878906,
-      "epoch": 0.756,
-      "grad_norm": 43.56858310727776,
-      "kl": 4.9921875,
-      "learning_rate": 2.5358974294659373e-07,
-      "loss": 1.0932,
-      "reward": 2.5396214723587036,
-      "reward_std": 0.7032516896724701,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.014198230113834143,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 321.1666717529297,
+      "epoch": 0.378,
+      "grad_norm": 67.36108059538589,
+      "kl": 3.19140625,
+      "learning_rate": 8.04235151541222e-07,
+      "loss": 0.2793,
+      "reward": 2.7950555086135864,
+      "reward_std": 0.21008452773094177,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03827812150120735,
+      "rewards/tag_count_reward": 1.0,
       "step": 756
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 380.25,
-      "epoch": 0.757,
-      "grad_norm": 42.833747394340975,
-      "kl": 3.90625,
-      "learning_rate": 2.5240966060435674e-07,
-      "loss": 0.4364,
-      "reward": 2.670251488685608,
-      "reward_std": 0.3203464448451996,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.013776272535324097,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "completion_length": 373.7291717529297,
+      "epoch": 0.3785,
+      "grad_norm": 84.54444780952632,
+      "kl": 2.7587890625,
+      "learning_rate": 8.035867220219071e-07,
+      "loss": 0.3899,
+      "reward": 2.434798002243042,
+      "reward_std": 0.4415217489004135,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.04263266548514366,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 757
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.7291870117188,
-      "epoch": 0.758,
-      "grad_norm": 103.32021322129948,
-      "kl": 6.125,
-      "learning_rate": 2.512332043064913e-07,
-      "loss": 1.2798,
-      "reward": 2.463586449623108,
-      "reward_std": 0.6297429800033569,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.017316261306405067,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 342.3541717529297,
+      "epoch": 0.379,
+      "grad_norm": 119.06865323156653,
+      "kl": 5.875,
+      "learning_rate": 8.029375200334587e-07,
+      "loss": 0.4683,
+      "reward": 2.722063660621643,
+      "reward_std": 0.38934317231178284,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.04356146976351738,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 758
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 603.5625305175781,
-      "epoch": 0.759,
-      "grad_norm": 43.65057021190212,
-      "kl": 6.9375,
-      "learning_rate": 2.5006038838774647e-07,
-      "loss": 0.9189,
-      "reward": 2.158547341823578,
-      "reward_std": 0.6411450803279877,
-      "rewards/accuracy_reward": 0.375,
-      "rewards/reasoning_steps_reward": 0.9444445073604584,
-      "rewards/repetition_penalty_reward": -0.009855579119175673,
-      "rewards/tag_count_reward": 0.8489583730697632,
+      "completion_length": 342.2916717529297,
+      "epoch": 0.3795,
+      "grad_norm": 27.51386663996604,
+      "kl": 2.546875,
+      "learning_rate": 8.022875475534588e-07,
+      "loss": 0.4781,
+      "reward": 2.81855309009552,
+      "reward_std": 0.3882335126399994,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.028669222258031368,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 759
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 617.6041870117188,
-      "epoch": 0.76,
-      "grad_norm": 19.541751124761845,
-      "kl": 6.046875,
-      "learning_rate": 2.488912271385139e-07,
-      "loss": 1.0373,
-      "reward": 2.354681611061096,
-      "reward_std": 0.6235771775245667,
-      "rewards/accuracy_reward": 0.6041666716337204,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.025526808109134436,
-      "rewards/tag_count_reward": 0.796875,
+      "completion_length": 388.0208435058594,
+      "epoch": 0.38,
+      "grad_norm": 75.44970102943357,
+      "kl": 3.5390625,
+      "learning_rate": 8.01636806561836e-07,
+      "loss": 0.5887,
+      "reward": 2.4815808534622192,
+      "reward_std": 0.4953022599220276,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.0427247304469347,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 760
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 486.4166717529297,
-      "epoch": 0.761,
-      "grad_norm": 39.557769432826085,
-      "kl": 2.8046875,
-      "learning_rate": 2.4772573480465445e-07,
-      "loss": 0.8268,
-      "reward": 2.4871546030044556,
-      "reward_std": 0.6703044772148132,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.024998143315315247,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 292.3958435058594,
+      "epoch": 0.3805,
+      "grad_norm": 5.1567729258708015,
+      "kl": 0.36328125,
+      "learning_rate": 8.009852990408606e-07,
+      "loss": -0.0068,
+      "reward": 2.7873027324676514,
+      "reward_std": 0.34609653055667877,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.02346120961010456,
+      "rewards/tag_count_reward": 0.984375,
       "step": 761
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.5208435058594,
-      "epoch": 0.762,
-      "grad_norm": 37.29197408055863,
-      "kl": 4.4140625,
-      "learning_rate": 2.465639255873246e-07,
-      "loss": 0.9776,
-      "reward": 2.1103891134262085,
-      "reward_std": 0.7344604879617691,
-      "rewards/accuracy_reward": 0.3958333358168602,
-      "rewards/reasoning_steps_reward": 0.9375000298023224,
-      "rewards/repetition_penalty_reward": -0.02502751164138317,
-      "rewards/tag_count_reward": 0.8020833432674408,
+      "completion_length": 298.2083435058594,
+      "epoch": 0.381,
+      "grad_norm": 5.782753089670167,
+      "kl": 0.2744140625,
+      "learning_rate": 8.003330269751372e-07,
+      "loss": 0.0454,
+      "reward": 2.6413458585739136,
+      "reward_std": 0.27150559425354004,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.05483482964336872,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 762
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 556.3333435058594,
-      "epoch": 0.763,
-      "grad_norm": 17.379113373562127,
-      "kl": 4.255859375,
-      "learning_rate": 2.454058136428027e-07,
-      "loss": 0.5297,
-      "reward": 2.3058613538742065,
-      "reward_std": 0.5592197477817535,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.017055293079465628,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 323.43751525878906,
+      "epoch": 0.3815,
+      "grad_norm": 8.790559038103488,
+      "kl": 0.712890625,
+      "learning_rate": 7.996799923515997e-07,
+      "loss": 0.1095,
+      "reward": 2.427235722541809,
+      "reward_std": 0.40729573369026184,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666269302368,
+      "rewards/repetition_penalty_reward": -0.036305982619524,
+      "rewards/tag_count_reward": 0.9427083730697632,
       "step": 763
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 404.56251525878906,
-      "epoch": 0.764,
-      "grad_norm": 18.161864885556103,
-      "kl": 1.474609375,
-      "learning_rate": 2.4425141308231765e-07,
-      "loss": 0.2815,
-      "reward": 2.35969877243042,
-      "reward_std": 0.4401056468486786,
-      "rewards/accuracy_reward": 0.4375,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.023981940932571888,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 339.37501525878906,
+      "epoch": 0.382,
+      "grad_norm": 12.201161237432581,
+      "kl": 0.9375,
+      "learning_rate": 7.990261971595048e-07,
+      "loss": 0.2632,
+      "reward": 2.6326658725738525,
+      "reward_std": 0.4427875429391861,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02879253402352333,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 764
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 388.0,
-      "epoch": 0.765,
-      "grad_norm": 35.807830777637626,
-      "kl": 2.65625,
-      "learning_rate": 2.4310073797187573e-07,
-      "loss": 0.7565,
-      "reward": 2.547232747077942,
-      "reward_std": 0.6375119686126709,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9305556416511536,
-      "rewards/repetition_penalty_reward": -0.013531276490539312,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 353.75,
+      "epoch": 0.3825,
+      "grad_norm": 22.3451597944635,
+      "kl": 1.24609375,
+      "learning_rate": 7.983716433904262e-07,
+      "loss": 0.3512,
+      "reward": 2.682904601097107,
+      "reward_std": 0.46456609666347504,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03758169710636139,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 765
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 414.5416717529297,
-      "epoch": 0.766,
-      "grad_norm": 31.125919596419223,
-      "kl": 1.38671875,
-      "learning_rate": 2.4195380233209006e-07,
-      "loss": 0.481,
-      "reward": 2.5880796909332275,
-      "reward_std": 0.547327809035778,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.021295433398336172,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 302.4583435058594,
+      "epoch": 0.383,
+      "grad_norm": 20.403277618055263,
+      "kl": 2.078125,
+      "learning_rate": 7.977163330382479e-07,
+      "loss": 0.2761,
+      "reward": 2.799572229385376,
+      "reward_std": 0.3507230281829834,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.026816830039024353,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 766
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 451.41668701171875,
-      "epoch": 0.767,
-      "grad_norm": 36.219759589516954,
-      "kl": 4.25,
-      "learning_rate": 2.408106201380097e-07,
-      "loss": 0.8591,
-      "reward": 2.333495259284973,
-      "reward_std": 0.6193402707576752,
-      "rewards/accuracy_reward": 0.583333358168602,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.015463168267160654,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 379.2291717529297,
+      "epoch": 0.3835,
+      "grad_norm": 83.31159486387506,
+      "kl": 6.796875,
+      "learning_rate": 7.970602680991592e-07,
+      "loss": 0.7361,
+      "reward": 2.5988014936447144,
+      "reward_std": 0.6159058213233948,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.026198641397058964,
+      "rewards/tag_count_reward": 0.8958333730697632,
       "step": 767
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.5833435058594,
-      "epoch": 0.768,
-      "grad_norm": 23.45639672156471,
-      "kl": 4.4765625,
-      "learning_rate": 2.3967120531894857e-07,
-      "loss": 0.9129,
-      "reward": 2.30954110622406,
-      "reward_std": 0.6106710433959961,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.016847790218889713,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 366.5625,
+      "epoch": 0.384,
+      "grad_norm": 363.6216773317555,
+      "kl": 13.65625,
+      "learning_rate": 7.964034505716476e-07,
+      "loss": 1.19,
+      "reward": 2.539207100868225,
+      "reward_std": 0.5661123096942902,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.03891796059906483,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 768
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 551.25,
-      "epoch": 0.769,
-      "grad_norm": 129.78473682307904,
-      "kl": 6.140625,
-      "learning_rate": 2.38535571758317e-07,
-      "loss": 1.3243,
-      "reward": 2.448180079460144,
-      "reward_std": 0.7538703083992004,
-      "rewards/accuracy_reward": 0.6875000149011612,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.02404231671243906,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "completion_length": 472.4791717529297,
+      "epoch": 0.3845,
+      "grad_norm": 187.4365397141768,
+      "kl": 16.546875,
+      "learning_rate": 7.957458824564931e-07,
+      "loss": 1.2768,
+      "reward": 2.338832139968872,
+      "reward_std": 0.5791323632001877,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.011862346436828375,
+      "rewards/tag_count_reward": 0.84375,
       "step": 769
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 569.8750305175781,
-      "epoch": 0.77,
-      "grad_norm": 35.971720291998714,
-      "kl": 7.40625,
-      "learning_rate": 2.374037332934512e-07,
-      "loss": 1.0446,
-      "reward": 2.1608505249023438,
-      "reward_std": 0.7265328466892242,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.028385651297867298,
-      "rewards/tag_count_reward": 0.828125,
+      "completion_length": 546.7916717529297,
+      "epoch": 0.385,
+      "grad_norm": 35.1440590310128,
+      "kl": 5.8671875,
+      "learning_rate": 7.950875657567621e-07,
+      "loss": 0.861,
+      "reward": 2.5058937072753906,
+      "reward_std": 0.5305248498916626,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.0201479597017169,
+      "rewards/tag_count_reward": 0.859375,
       "step": 770
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 474.7916717529297,
-      "epoch": 0.771,
-      "grad_norm": 78.42716292437864,
-      "kl": 4.375,
-      "learning_rate": 2.36275703715446e-07,
-      "loss": 0.6855,
-      "reward": 2.4102360010147095,
-      "reward_std": 0.627646267414093,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.016847367398440838,
-      "rewards/tag_count_reward": 0.9270833730697632,
+      "completion_length": 511.39585876464844,
+      "epoch": 0.3855,
+      "grad_norm": 78.06611028598661,
+      "kl": 6.6328125,
+      "learning_rate": 7.944285024778017e-07,
+      "loss": 0.8001,
+      "reward": 2.510785937309265,
+      "reward_std": 0.6201603412628174,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.027408569119870663,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 771
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 517.9166870117188,
-      "epoch": 0.772,
-      "grad_norm": 51.33839598106081,
-      "kl": 7.234375,
-      "learning_rate": 2.3515149676898552e-07,
-      "loss": 0.9322,
-      "reward": 2.0547146797180176,
-      "reward_std": 0.7165104746818542,
-      "rewards/accuracy_reward": 0.3333333432674408,
-      "rewards/reasoning_steps_reward": 0.8819445371627808,
-      "rewards/repetition_penalty_reward": -0.03035500179976225,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 411.1458435058594,
+      "epoch": 0.386,
+      "grad_norm": 18.27424052638579,
+      "kl": 2.765625,
+      "learning_rate": 7.93768694627233e-07,
+      "loss": 0.5689,
+      "reward": 2.3896729946136475,
+      "reward_std": 0.5375154912471771,
+      "rewards/accuracy_reward": 0.5416666716337204,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.028729761950671673,
+      "rewards/tag_count_reward": 0.8906250298023224,
       "step": 772
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 662.8541870117188,
-      "epoch": 0.773,
-      "grad_norm": 99.24123358538509,
-      "kl": 10.1875,
-      "learning_rate": 2.3403112615217693e-07,
-      "loss": 1.2164,
-      "reward": 2.322015166282654,
-      "reward_std": 0.7576204538345337,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.014790416695177555,
-      "rewards/tag_count_reward": 0.7812500298023224,
+      "completion_length": 495.625,
+      "epoch": 0.3865,
+      "grad_norm": 14.881333576700378,
+      "kl": 2.375,
+      "learning_rate": 7.931081442149448e-07,
+      "loss": 0.4987,
+      "reward": 2.577857255935669,
+      "reward_std": 0.5150826573371887,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02457323018461466,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 773
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 555.0208587646484,
-      "epoch": 0.774,
-      "grad_norm": 55.815576719277246,
-      "kl": 6.765625,
-      "learning_rate": 2.3291460551638237e-07,
-      "loss": 1.1251,
-      "reward": 2.1742671728134155,
-      "reward_std": 0.6895886063575745,
-      "rewards/accuracy_reward": 0.3958333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.020177372731268406,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "completion_length": 427.9375,
+      "epoch": 0.387,
+      "grad_norm": 11.469486615908298,
+      "kl": 1.912109375,
+      "learning_rate": 7.924468532530883e-07,
+      "loss": 0.2236,
+      "reward": 2.6311352252960205,
+      "reward_std": 0.37547628581523895,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0615730844438076,
+      "rewards/tag_count_reward": 0.9427083730697632,
       "step": 774
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 501.3333435058594,
-      "epoch": 0.775,
-      "grad_norm": 21.026258662130076,
-      "kl": 5.171875,
-      "learning_rate": 2.3180194846605364e-07,
-      "loss": 0.8011,
-      "reward": 2.2563339471817017,
-      "reward_std": 0.6224390119314194,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.011027187574654818,
-      "rewards/tag_count_reward": 0.8645833730697632,
+      "completion_length": 642.3125305175781,
+      "epoch": 0.3875,
+      "grad_norm": 47.419569435384595,
+      "kl": 5.8515625,
+      "learning_rate": 7.917848237560708e-07,
+      "loss": 0.7953,
+      "reward": 2.1859389543533325,
+      "reward_std": 0.7739139497280121,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.02933894842863083,
+      "rewards/tag_count_reward": 0.7708333432674408,
       "step": 775
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 488.3958435058594,
-      "epoch": 0.776,
-      "grad_norm": 40.72839840915295,
-      "kl": 2.578125,
-      "learning_rate": 2.306931685585657e-07,
-      "loss": 0.791,
-      "reward": 2.5573089122772217,
-      "reward_std": 0.6792595684528351,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.04164944589138031,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "completion_length": 433.8958435058594,
+      "epoch": 0.388,
+      "grad_norm": 16.151319546185952,
+      "kl": 2.9453125,
+      "learning_rate": 7.911220577405484e-07,
+      "loss": 0.5875,
+      "reward": 2.4124279022216797,
+      "reward_std": 0.5835599303245544,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.023336103186011314,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 776
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 478.1875,
-      "epoch": 0.777,
-      "grad_norm": 65.73064893150305,
-      "kl": 3.2578125,
-      "learning_rate": 2.2958827930405162e-07,
-      "loss": 1.0533,
-      "reward": 2.6781049966812134,
-      "reward_std": 0.6686598658561707,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.01633965945802629,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 493.87501525878906,
+      "epoch": 0.3885,
+      "grad_norm": 19.784079907225436,
+      "kl": 3.0546875,
+      "learning_rate": 7.904585572254218e-07,
+      "loss": 0.5988,
+      "reward": 2.564281463623047,
+      "reward_std": 0.6512129902839661,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02773252595216036,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 777
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 374.79168701171875,
-      "epoch": 0.778,
-      "grad_norm": 12.77197238349156,
-      "kl": 2.05078125,
-      "learning_rate": 2.2848729416523859e-07,
-      "loss": 0.2598,
-      "reward": 2.4885555505752563,
-      "reward_std": 0.43929314613342285,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.014916833024471998,
-      "rewards/tag_count_reward": 0.9687500298023224,
+      "completion_length": 527.8750152587891,
+      "epoch": 0.389,
+      "grad_norm": 14.102553925882571,
+      "kl": 3.291015625,
+      "learning_rate": 7.897943242318285e-07,
+      "loss": 0.6637,
+      "reward": 2.508991241455078,
+      "reward_std": 0.49256084859371185,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.03962016478180885,
+      "rewards/tag_count_reward": 0.875,
       "step": 778
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.2291870117188,
-      "epoch": 0.779,
-      "grad_norm": 27.96030488624158,
-      "kl": 6.6875,
-      "learning_rate": 2.2739022655728277e-07,
-      "loss": 0.9628,
-      "reward": 2.069424092769623,
-      "reward_std": 0.6286317706108093,
-      "rewards/accuracy_reward": 0.3125000111758709,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.012173243798315525,
-      "rewards/tag_count_reward": 0.8385416865348816,
+      "completion_length": 553.7708587646484,
+      "epoch": 0.3895,
+      "grad_norm": 21.449440079929175,
+      "kl": 4.6875,
+      "learning_rate": 7.891293607831373e-07,
+      "loss": 0.6951,
+      "reward": 2.2457956075668335,
+      "reward_std": 0.3153197094798088,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.03719065245240927,
+      "rewards/tag_count_reward": 0.8593750298023224,
       "step": 779
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 430.22918701171875,
-      "epoch": 0.78,
-      "grad_norm": 29.26274656267414,
-      "kl": 3.59375,
-      "learning_rate": 2.2629708984760706e-07,
-      "loss": 0.9216,
-      "reward": 2.4278910160064697,
-      "reward_std": 0.5140875577926636,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.018289764411747456,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 681.0833435058594,
+      "epoch": 0.39,
+      "grad_norm": 50.20111893584829,
+      "kl": 5.640625,
+      "learning_rate": 7.884636689049422e-07,
+      "loss": 0.9104,
+      "reward": 2.192242741584778,
+      "reward_std": 0.5565801113843918,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.04213242046535015,
+      "rewards/tag_count_reward": 0.7968750298023224,
       "step": 780
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.8750152587891,
-      "epoch": 0.781,
-      "grad_norm": 27.923966049196657,
-      "kl": 4.5703125,
-      "learning_rate": 2.2520789735573704e-07,
-      "loss": 1.121,
-      "reward": 2.480145573616028,
-      "reward_std": 0.6819994747638702,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.02679906040430069,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 493.3125305175781,
+      "epoch": 0.3905,
+      "grad_norm": 31.66716632745188,
+      "kl": 2.26953125,
+      "learning_rate": 7.877972506250562e-07,
+      "loss": 0.727,
+      "reward": 2.6193435192108154,
+      "reward_std": 0.655490517616272,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.03517040517181158,
+      "rewards/tag_count_reward": 0.8906250298023224,
       "step": 781
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 503.1875305175781,
-      "epoch": 0.782,
-      "grad_norm": 33.004171238913635,
-      "kl": 2.75,
-      "learning_rate": 2.2412266235313973e-07,
-      "loss": 0.7836,
-      "reward": 2.555114269256592,
-      "reward_std": 0.6732204258441925,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.04210797697305679,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 546.0833435058594,
+      "epoch": 0.391,
+      "grad_norm": 20.081813742060923,
+      "kl": 2.97265625,
+      "learning_rate": 7.871301079735049e-07,
+      "loss": 0.6893,
+      "reward": 2.3784666061401367,
+      "reward_std": 0.4472994953393936,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.02604720927774906,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 782
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 543.2291870117188,
-      "epoch": 0.783,
-      "grad_norm": 26.670810792645383,
-      "kl": 3.9765625,
-      "learning_rate": 2.230413980630609e-07,
-      "loss": 0.8157,
-      "reward": 2.3876798152923584,
-      "reward_std": 0.6481176614761353,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9305555522441864,
-      "rewards/repetition_penalty_reward": -0.02725088130682707,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 530.6458435058594,
+      "epoch": 0.3915,
+      "grad_norm": 18.755713225081045,
+      "kl": 3.94140625,
+      "learning_rate": 7.864622429825204e-07,
+      "loss": 0.5134,
+      "reward": 2.5227210521698,
+      "reward_std": 0.6135649532079697,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.04672360047698021,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 783
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 454.14585876464844,
-      "epoch": 0.784,
-      "grad_norm": 20.273867549895126,
-      "kl": 2.8828125,
-      "learning_rate": 2.2196411766036487e-07,
-      "loss": 0.3284,
-      "reward": 2.4754650592803955,
-      "reward_std": 0.5723298937082291,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03495161887258291,
-      "rewards/tag_count_reward": 0.90625,
+      "completion_length": 393.2083435058594,
+      "epoch": 0.392,
+      "grad_norm": 44.37092698629651,
+      "kl": 3.7265625,
+      "learning_rate": 7.857936576865356e-07,
+      "loss": 0.5662,
+      "reward": 2.404683828353882,
+      "reward_std": 0.31213444471359253,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.02413582894951105,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 784
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 726.125,
-      "epoch": 0.785,
-      "grad_norm": 29.95305237028232,
-      "kl": 8.2578125,
-      "learning_rate": 2.2089083427137329e-07,
-      "loss": 1.0536,
-      "reward": 2.2507245540618896,
-      "reward_std": 0.7002788186073303,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.025317177176475525,
-      "rewards/tag_count_reward": 0.7552083432674408,
+      "completion_length": 403.0208435058594,
+      "epoch": 0.3925,
+      "grad_norm": 26.04848472624046,
+      "kl": 1.279296875,
+      "learning_rate": 7.851243541221769e-07,
+      "loss": 0.4965,
+      "reward": 2.568751573562622,
+      "reward_std": 0.3721753219142556,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.023262279108166695,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 785
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 456.6875,
-      "epoch": 0.786,
-      "grad_norm": 23.330969926542757,
-      "kl": 3.6953125,
-      "learning_rate": 2.1982156097370557e-07,
-      "loss": 0.851,
-      "reward": 2.6652588844299316,
-      "reward_std": 0.6001283526420593,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.015296765603125095,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 430.6458435058594,
+      "epoch": 0.393,
+      "grad_norm": 13.25665865367494,
+      "kl": 2.01953125,
+      "learning_rate": 7.844543343282595e-07,
+      "loss": 0.4602,
+      "reward": 2.5878489017486572,
+      "reward_std": 0.3847273141145706,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9722221791744232,
+      "rewards/repetition_penalty_reward": -0.03541503753513098,
+      "rewards/tag_count_reward": 0.921875,
       "step": 786
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 439.56251525878906,
-      "epoch": 0.787,
-      "grad_norm": 41.78445904580422,
-      "kl": 3.1875,
-      "learning_rate": 2.1875631079611956e-07,
-      "loss": 0.9046,
-      "reward": 2.517007827758789,
-      "reward_std": 0.5808334052562714,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.958333432674408,
-      "rewards/repetition_penalty_reward": -0.019450515508651733,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 444.87501525878906,
+      "epoch": 0.3935,
+      "grad_norm": 22.772832737303286,
+      "kl": 4.8125,
+      "learning_rate": 7.837836003457793e-07,
+      "loss": 0.6349,
+      "reward": 2.510228753089905,
+      "reward_std": 0.5596920847892761,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04532698355615139,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 787
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 500.0208435058594,
-      "epoch": 0.788,
-      "grad_norm": 21.94070893015738,
-      "kl": 3.8828125,
-      "learning_rate": 2.1769509671835223e-07,
-      "loss": 0.8589,
-      "reward": 2.501682996749878,
-      "reward_std": 0.7997144758701324,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.022622703574597836,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 472.9583435058594,
+      "epoch": 0.394,
+      "grad_norm": 15.953872553666358,
+      "kl": 2.59375,
+      "learning_rate": 7.831121542179086e-07,
+      "loss": 0.7152,
+      "reward": 2.431522011756897,
+      "reward_std": 0.46617111563682556,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04590862803161144,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 788
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 356.10418701171875,
-      "epoch": 0.789,
-      "grad_norm": 18.917515789374068,
-      "kl": 4.484375,
-      "learning_rate": 2.166379316709625e-07,
-      "loss": 0.4355,
-      "reward": 2.2824082374572754,
-      "reward_std": 0.5984006226062775,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.916666716337204,
-      "rewards/repetition_penalty_reward": -0.019675294868648052,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 631.1041870117188,
+      "epoch": 0.3945,
+      "grad_norm": 21.209792192046844,
+      "kl": 3.51953125,
+      "learning_rate": 7.824399979899889e-07,
+      "loss": 0.4762,
+      "reward": 2.2550671100616455,
+      "reward_std": 0.3108735680580139,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06437743455171585,
+      "rewards/tag_count_reward": 0.875,
       "step": 789
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 632.2083740234375,
-      "epoch": 0.79,
-      "grad_norm": 16.91140019655358,
-      "kl": 6.6875,
-      "learning_rate": 2.1558482853517253e-07,
-      "loss": 1.2443,
-      "reward": 2.2729358673095703,
-      "reward_std": 0.853611171245575,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.8819445073604584,
-      "rewards/repetition_penalty_reward": -0.025675359182059765,
-      "rewards/tag_count_reward": 0.8333333432674408,
+      "completion_length": 340.8333435058594,
+      "epoch": 0.395,
+      "grad_norm": 33.37412557405853,
+      "kl": 1.37890625,
+      "learning_rate": 7.817671337095244e-07,
+      "loss": 0.4575,
+      "reward": 2.5797940492630005,
+      "reward_std": 0.6914741396903992,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.041733767837285995,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 790
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 376.31251525878906,
-      "epoch": 0.791,
-      "grad_norm": 20.048000622022776,
-      "kl": 1.310546875,
-      "learning_rate": 2.1453580014271203e-07,
-      "loss": 0.3778,
-      "reward": 2.2707479000091553,
-      "reward_std": 0.3341696485877037,
-      "rewards/accuracy_reward": 0.3125000074505806,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.0191827230155468,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 437.12501525878906,
+      "epoch": 0.3955,
+      "grad_norm": 16.65766077241616,
+      "kl": 1.3662109375,
+      "learning_rate": 7.810935634261764e-07,
+      "loss": 0.3751,
+      "reward": 2.6942687034606934,
+      "reward_std": 0.47277119755744934,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.04705093614757061,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 791
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 371.6458435058594,
-      "epoch": 0.792,
-      "grad_norm": 19.6502337749718,
-      "kl": 1.71484375,
-      "learning_rate": 2.134908592756607e-07,
-      "loss": 0.4382,
-      "reward": 2.6625298261642456,
-      "reward_std": 0.537158876657486,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02670634165406227,
-      "rewards/tag_count_reward": 0.9739583730697632,
+      "completion_length": 486.60418701171875,
+      "epoch": 0.396,
+      "grad_norm": 9.722828730186746,
+      "kl": 2.5546875,
+      "learning_rate": 7.804192891917571e-07,
+      "loss": 0.4686,
+      "reward": 2.0610522031784058,
+      "reward_std": 0.35085567831993103,
+      "rewards/accuracy_reward": 0.2916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.04658674821257591,
+      "rewards/tag_count_reward": 0.8645833432674408,
       "step": 792
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 515.7291870117188,
-      "epoch": 0.793,
-      "grad_norm": 28.434148981976,
-      "kl": 6.421875,
-      "learning_rate": 2.124500186662932e-07,
-      "loss": 0.8134,
-      "reward": 2.500126600265503,
-      "reward_std": 0.7610350847244263,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.013762431219220161,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 768.0,
+      "epoch": 0.3965,
+      "grad_norm": 32.12608283927452,
+      "kl": 5.296875,
+      "learning_rate": 7.797443130602226e-07,
+      "loss": 0.9858,
+      "reward": 1.9784066081047058,
+      "reward_std": 0.8380246162414551,
+      "rewards/accuracy_reward": 0.3333333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.10145452991127968,
+      "rewards/tag_count_reward": 0.7812500298023224,
       "step": 793
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 621.7500305175781,
-      "epoch": 0.794,
-      "grad_norm": 34.773569674299296,
-      "kl": 7.453125,
-      "learning_rate": 2.1141329099692406e-07,
-      "loss": 1.1273,
-      "reward": 2.032840847969055,
-      "reward_std": 0.881601870059967,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.034867677837610245,
-      "rewards/tag_count_reward": 0.7968750298023224,
+      "completion_length": 600.875,
+      "epoch": 0.397,
+      "grad_norm": 18.410765045440638,
+      "kl": 2.9765625,
+      "learning_rate": 7.79068637087667e-07,
+      "loss": 0.643,
+      "reward": 2.441210389137268,
+      "reward_std": 0.3568989485502243,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9791667461395264,
+      "rewards/repetition_penalty_reward": -0.08483139425516129,
+      "rewards/tag_count_reward": 0.8802083432674408,
       "step": 794
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 517.0000152587891,
-      "epoch": 0.795,
-      "grad_norm": 30.664837556455563,
-      "kl": 5.78125,
-      "learning_rate": 2.1038068889975259e-07,
-      "loss": 1.0339,
-      "reward": 2.5728070735931396,
-      "reward_std": 0.7294317185878754,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.024415286257863045,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 436.79168701171875,
+      "epoch": 0.3975,
+      "grad_norm": 10.317344532371504,
+      "kl": 1.23046875,
+      "learning_rate": 7.783922633323169e-07,
+      "loss": 0.2964,
+      "reward": 2.6988085508346558,
+      "reward_std": 0.543820321559906,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.07896940223872662,
+      "rewards/tag_count_reward": 0.9583333730697632,
       "step": 795
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 454.56251525878906,
-      "epoch": 0.796,
-      "grad_norm": 22.153680697625163,
-      "kl": 5.1015625,
-      "learning_rate": 2.0935222495670968e-07,
-      "loss": 1.0399,
-      "reward": 2.4211684465408325,
-      "reward_std": 0.7883006036281586,
+      "completion_length": 542.0416870117188,
+      "epoch": 0.398,
+      "grad_norm": 20.086888963335625,
+      "kl": 3.09375,
+      "learning_rate": 7.777151938545235e-07,
+      "loss": 0.6306,
+      "reward": 2.3561642169952393,
+      "reward_std": 0.7507622838020325,
       "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.01285938685759902,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.10737751796841621,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 796
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 558.0000152587891,
-      "epoch": 0.797,
-      "grad_norm": 44.01821977708549,
-      "kl": 5.890625,
-      "learning_rate": 2.0832791169930363e-07,
-      "loss": 1.2008,
-      "reward": 2.6672359704971313,
-      "reward_std": 0.5852415859699249,
+      "completion_length": 419.9375,
+      "epoch": 0.3985,
+      "grad_norm": 13.37138079802378,
+      "kl": 0.71484375,
+      "learning_rate": 7.770374307167585e-07,
+      "loss": 0.378,
+      "reward": 2.7522013187408447,
+      "reward_std": 0.4279877096414566,
       "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222983837128,
-      "rewards/repetition_penalty_reward": -0.02894441783428192,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.05335431918501854,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 797
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 579.6458435058594,
-      "epoch": 0.798,
-      "grad_norm": 23.974305974531344,
-      "kl": 5.125,
-      "learning_rate": 2.0730776160846853e-07,
-      "loss": 0.9124,
-      "reward": 2.1921111345291138,
-      "reward_std": 0.7356277704238892,
-      "rewards/accuracy_reward": 0.458333358168602,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.024902895092964172,
-      "rewards/tag_count_reward": 0.8281250298023224,
+      "completion_length": 587.2500305175781,
+      "epoch": 0.399,
+      "grad_norm": 27.70225468914978,
+      "kl": 2.796875,
+      "learning_rate": 7.763589759836058e-07,
+      "loss": 0.6997,
+      "reward": 2.363113760948181,
+      "reward_std": 0.67525914311409,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.13341398537158966,
+      "rewards/tag_count_reward": 0.8645833432674408,
       "step": 798
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 496.81251525878906,
-      "epoch": 0.799,
-      "grad_norm": 20.99627755388401,
-      "kl": 4.53125,
-      "learning_rate": 2.0629178711441115e-07,
-      "loss": 0.7604,
-      "reward": 2.4410648345947266,
-      "reward_std": 0.5636108368635178,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.010324170347303152,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 870.6666870117188,
+      "epoch": 0.3995,
+      "grad_norm": 15.870305020089033,
+      "kl": 3.0859375,
+      "learning_rate": 7.756798317217558e-07,
+      "loss": 0.8678,
+      "reward": 1.9343020915985107,
+      "reward_std": 0.7617403268814087,
+      "rewards/accuracy_reward": 0.3750000149011612,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.20458681881427765,
+      "rewards/tag_count_reward": 0.8125,
       "step": 799
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 608.5208740234375,
-      "epoch": 0.8,
-      "grad_norm": 26.72579720606878,
-      "kl": 5.25,
-      "learning_rate": 2.0528000059645995e-07,
-      "loss": 0.7956,
-      "reward": 2.119356393814087,
-      "reward_std": 0.5921457409858704,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03689361736178398,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 647.75,
+      "epoch": 0.4,
+      "grad_norm": 12.665956910806365,
+      "kl": 1.33203125,
+      "learning_rate": 7.75e-07,
+      "loss": 0.6692,
+      "reward": 2.246774673461914,
+      "reward_std": 0.6763182580471039,
+      "rewards/accuracy_reward": 0.5208333358168602,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.15079493820667267,
+      "rewards/tag_count_reward": 0.9114583730697632,
       "step": 800
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 312.0,
-      "epoch": 0.801,
-      "grad_norm": 13.404486874760053,
-      "kl": 1.23046875,
-      "learning_rate": 2.042724143829146e-07,
-      "loss": 0.077,
-      "reward": 2.3504269123077393,
-      "reward_std": 0.36202409863471985,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.014156644232571125,
-      "rewards/tag_count_reward": 0.9687500298023224,
+      "completion_length": 515.3750305175781,
+      "epoch": 0.4005,
+      "grad_norm": 11.953588265183154,
+      "kl": 0.8115234375,
+      "learning_rate": 7.743194828892235e-07,
+      "loss": 0.4331,
+      "reward": 2.6169031858444214,
+      "reward_std": 0.5906797051429749,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.10011079534888268,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 801
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 434.4583435058594,
-      "epoch": 0.802,
-      "grad_norm": 19.113136797766565,
-      "kl": 2.7421875,
-      "learning_rate": 2.032690407508949e-07,
-      "loss": 0.5925,
-      "reward": 2.4155834913253784,
-      "reward_std": 0.6261934638023376,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.020180439576506615,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 601.0625152587891,
+      "epoch": 0.401,
+      "grad_norm": 11.261392190054263,
+      "kl": 1.77734375,
+      "learning_rate": 7.736382824623999e-07,
+      "loss": 0.8205,
+      "reward": 2.2957258224487305,
+      "reward_std": 0.8579961359500885,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.909722238779068,
+      "rewards/repetition_penalty_reward": -0.15045488253235817,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 802
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 501.4583435058594,
-      "epoch": 0.803,
-      "grad_norm": 25.89659290548137,
-      "kl": 3.203125,
-      "learning_rate": 2.0226989192619204e-07,
-      "loss": 0.4875,
-      "reward": 2.301639437675476,
-      "reward_std": 0.555485874414444,
-      "rewards/accuracy_reward": 0.4791666865348816,
+      "completion_length": 513.75,
+      "epoch": 0.4015,
+      "grad_norm": 7.271737188658035,
+      "kl": 1.1640625,
+      "learning_rate": 7.729564007945834e-07,
+      "loss": 0.5417,
+      "reward": 2.590050458908081,
+      "reward_std": 0.43862822093069553,
+      "rewards/accuracy_reward": 0.8125,
       "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03690226562321186,
-      "rewards/tag_count_reward": 0.8802083730697632,
+      "rewards/repetition_penalty_reward": -0.13390799798071384,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 803
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 438.1250305175781,
-      "epoch": 0.804,
-      "grad_norm": 21.314874667338287,
-      "kl": 2.4453125,
-      "learning_rate": 2.0127498008311922e-07,
-      "loss": 0.3104,
-      "reward": 2.334763288497925,
-      "reward_std": 0.3796348571777344,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.02287570107728243,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 619.6666870117188,
+      "epoch": 0.402,
+      "grad_norm": 12.068916347239274,
+      "kl": 1.515625,
+      "learning_rate": 7.72273839962904e-07,
+      "loss": 0.6556,
+      "reward": 2.3289283514022827,
+      "reward_std": 0.6953560709953308,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.14503009244799614,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 804
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 421.6666717529297,
-      "epoch": 0.805,
-      "grad_norm": 14.366039627616365,
-      "kl": 1.8720703125,
-      "learning_rate": 2.0028431734436308e-07,
-      "loss": 0.3854,
-      "reward": 2.546318769454956,
-      "reward_std": 0.3817850574851036,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.026598164811730385,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 473.06251525878906,
+      "epoch": 0.4025,
+      "grad_norm": 7.004105215955508,
+      "kl": 1.224609375,
+      "learning_rate": 7.715906020465602e-07,
+      "loss": 0.51,
+      "reward": 2.6324926614761353,
+      "reward_std": 0.42844393849372864,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.10535471886396408,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 805
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 418.625,
-      "epoch": 0.806,
-      "grad_norm": 21.19577118539385,
-      "kl": 3.046875,
-      "learning_rate": 1.9929791578083655e-07,
-      "loss": 0.5684,
-      "reward": 2.6353834867477417,
-      "reward_std": 0.5922529548406601,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.02260266337543726,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 571.7916717529297,
+      "epoch": 0.403,
+      "grad_norm": 15.381900062308238,
+      "kl": 1.984375,
+      "learning_rate": 7.709066891268133e-07,
+      "loss": 0.5004,
+      "reward": 2.460014581680298,
+      "reward_std": 0.6100521683692932,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.14241605252027512,
+      "rewards/tag_count_reward": 0.921875,
       "step": 806
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 527.6666870117188,
-      "epoch": 0.807,
-      "grad_norm": 26.76069518759677,
-      "kl": 5.15625,
-      "learning_rate": 1.9831578741153155e-07,
-      "loss": 0.7895,
-      "reward": 2.3032405376434326,
-      "reward_std": 0.808287501335144,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9305556416511536,
-      "rewards/repetition_penalty_reward": -0.028356720693409443,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 360.31251525878906,
+      "epoch": 0.4035,
+      "grad_norm": 5.503594150237046,
+      "kl": 0.3388671875,
+      "learning_rate": 7.702221032869808e-07,
+      "loss": 0.1863,
+      "reward": 2.5507017374038696,
+      "reward_std": 0.32198452949523926,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.06388161890208721,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 807
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 381.8125,
-      "epoch": 0.808,
-      "grad_norm": 20.376838690762124,
-      "kl": 1.927734375,
-      "learning_rate": 1.9733794420337213e-07,
-      "loss": 0.268,
-      "reward": 2.428946018218994,
-      "reward_std": 0.5701454132795334,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.027651351876556873,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 360.25,
+      "epoch": 0.404,
+      "grad_norm": 11.719601907986142,
+      "kl": 0.39794921875,
+      "learning_rate": 7.695368466124296e-07,
+      "loss": 0.0649,
+      "reward": 2.8640111684799194,
+      "reward_std": 0.24921771883964539,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.052655573934316635,
+      "rewards/tag_count_reward": 1.0,
       "step": 808
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 445.62501525878906,
-      "epoch": 0.809,
-      "grad_norm": 20.816082501770886,
-      "kl": 5.140625,
-      "learning_rate": 1.9636439807106912e-07,
-      "loss": 0.8017,
-      "reward": 2.5268882513046265,
-      "reward_std": 0.7573049068450928,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.023458978161215782,
-      "rewards/tag_count_reward": 0.8489583730697632,
+      "completion_length": 489.6666717529297,
+      "epoch": 0.4045,
+      "grad_norm": 16.98625822145741,
+      "kl": 0.5205078125,
+      "learning_rate": 7.688509211905707e-07,
+      "loss": 0.6203,
+      "reward": 2.476475715637207,
+      "reward_std": 0.7080457210540771,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.08949651196599007,
+      "rewards/tag_count_reward": 0.9687500298023224,
       "step": 809
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 391.81251525878906,
-      "epoch": 0.81,
-      "grad_norm": 22.694875364982547,
-      "kl": 2.3359375,
-      "learning_rate": 1.9539516087697517e-07,
-      "loss": 0.5417,
-      "reward": 2.658096432685852,
-      "reward_std": 0.6354215741157532,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.0172510021366179,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 439.5208435058594,
+      "epoch": 0.405,
+      "grad_norm": 12.748082551046092,
+      "kl": 0.71875,
+      "learning_rate": 7.681643291108517e-07,
+      "loss": 0.3463,
+      "reward": 2.565284848213196,
+      "reward_std": 0.4373869299888611,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.07186824828386307,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 810
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 503.35418701171875,
-      "epoch": 0.811,
-      "grad_norm": 58.019855154883885,
-      "kl": 6.5625,
-      "learning_rate": 1.944302444309393e-07,
-      "loss": 1.3334,
-      "reward": 2.501904606819153,
-      "reward_std": 0.7767859101295471,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.022400940768420696,
-      "rewards/tag_count_reward": 0.8645833730697632,
+      "completion_length": 732.1875305175781,
+      "epoch": 0.4055,
+      "grad_norm": 48.03197567653632,
+      "kl": 1.46875,
+      "learning_rate": 7.67477072464751e-07,
+      "loss": 0.6842,
+      "reward": 1.9572343230247498,
+      "reward_std": 0.734717458486557,
+      "rewards/accuracy_reward": 0.4375000223517418,
+      "rewards/reasoning_steps_reward": 0.881944477558136,
+      "rewards/repetition_penalty_reward": -0.20596013963222504,
+      "rewards/tag_count_reward": 0.8437500298023224,
       "step": 811
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 516.8125,
-      "epoch": 0.812,
-      "grad_norm": 54.33176698076773,
-      "kl": 7.484375,
-      "learning_rate": 1.934696604901642e-07,
-      "loss": 0.8888,
-      "reward": 2.2534443140029907,
-      "reward_std": 0.7112808525562286,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.017389055341482162,
-      "rewards/tag_count_reward": 0.8333333730697632,
+      "completion_length": 474.5,
+      "epoch": 0.406,
+      "grad_norm": 36.55159806207502,
+      "kl": 2.560546875,
+      "learning_rate": 7.667891533457718e-07,
+      "loss": 0.3356,
+      "reward": 2.531466484069824,
+      "reward_std": 0.40393590182065964,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.09179743006825447,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 812
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 467.00001525878906,
-      "epoch": 0.813,
-      "grad_norm": 35.200107482672536,
-      "kl": 4.8515625,
-      "learning_rate": 1.9251342075906179e-07,
-      "loss": 0.9493,
-      "reward": 2.1571664810180664,
-      "reward_std": 0.6603484153747559,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9236112236976624,
-      "rewards/repetition_penalty_reward": -0.021653023548424244,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 486.1875,
+      "epoch": 0.4065,
+      "grad_norm": 501.2023762121776,
+      "kl": 10.4375,
+      "learning_rate": 7.661005738494349e-07,
+      "loss": 1.8999,
+      "reward": 2.5981922149658203,
+      "reward_std": 0.7066002190113068,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.09451625496149063,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 813
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 469.1041717529297,
-      "epoch": 0.814,
-      "grad_norm": 54.98157225337433,
-      "kl": 5.5,
-      "learning_rate": 1.915615368891117e-07,
-      "loss": 0.9186,
-      "reward": 2.1737260818481445,
-      "reward_std": 0.6402971148490906,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9166667461395264,
-      "rewards/repetition_penalty_reward": -0.013774096965789795,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 753.5000305175781,
+      "epoch": 0.407,
+      "grad_norm": 356.97977721564547,
+      "kl": 15.0,
+      "learning_rate": 7.654113360732732e-07,
+      "loss": 1.628,
+      "reward": 2.165995717048645,
+      "reward_std": 0.8618307709693909,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.19858761131763458,
+      "rewards/tag_count_reward": 0.8437500298023224,
       "step": 814
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 531.7291717529297,
-      "epoch": 0.815,
-      "grad_norm": 44.84634369429172,
-      "kl": 6.203125,
-      "learning_rate": 1.9061402047871833e-07,
-      "loss": 1.2846,
-      "reward": 2.147634506225586,
-      "reward_std": 0.592269778251648,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.019032152369618416,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 499.125,
+      "epoch": 0.4075,
+      "grad_norm": 123.33805411679445,
+      "kl": 4.54296875,
+      "learning_rate": 7.647214421168238e-07,
+      "loss": 1.0472,
+      "reward": 2.6111273765563965,
+      "reward_std": 0.6222415566444397,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.11283096671104431,
+      "rewards/tag_count_reward": 0.953125,
       "step": 815
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 393.7291717529297,
-      "epoch": 0.816,
-      "grad_norm": 23.81197579521365,
-      "kl": 4.3671875,
-      "learning_rate": 1.8967088307307e-07,
-      "loss": 0.4182,
-      "reward": 2.38164883852005,
-      "reward_std": 0.6038723587989807,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9375001192092896,
-      "rewards/repetition_penalty_reward": -0.019393039867281914,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 623.5625305175781,
+      "epoch": 0.408,
+      "grad_norm": 111.86522019096887,
+      "kl": 1.94921875,
+      "learning_rate": 7.640308940816239e-07,
+      "loss": 1.0335,
+      "reward": 2.26309335231781,
+      "reward_std": 0.8162707686424255,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.15010132268071175,
+      "rewards/tag_count_reward": 0.9062500298023224,
       "step": 816
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 435.91668701171875,
-      "epoch": 0.817,
-      "grad_norm": 25.95495924946456,
-      "kl": 3.8046875,
-      "learning_rate": 1.887321361639985e-07,
-      "loss": 1.0102,
-      "reward": 2.579134702682495,
-      "reward_std": 0.5107276141643524,
-      "rewards/accuracy_reward": 0.7083333730697632,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.026768138632178307,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 818.4166870117188,
+      "epoch": 0.4085,
+      "grad_norm": 53.005862638537856,
+      "kl": 2.6171875,
+      "learning_rate": 7.633396940712023e-07,
+      "loss": 0.7982,
+      "reward": 2.1612285375595093,
+      "reward_std": 1.055372416973114,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.21550768613815308,
+      "rewards/tag_count_reward": 0.8072916865348816,
       "step": 817
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 375.68751525878906,
-      "epoch": 0.818,
-      "grad_norm": 12.048598436192929,
-      "kl": 1.546875,
-      "learning_rate": 1.8779779118983867e-07,
-      "loss": 0.0993,
-      "reward": 2.702491521835327,
-      "reward_std": 0.4488513916730881,
-      "rewards/accuracy_reward": 0.7708333730697632,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.028411319479346275,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 524.6041717529297,
+      "epoch": 0.409,
+      "grad_norm": 35.60976729675571,
+      "kl": 2.546875,
+      "learning_rate": 7.626478441910744e-07,
+      "loss": 0.826,
+      "reward": 2.595088481903076,
+      "reward_std": 0.6593966484069824,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9374999701976776,
+      "rewards/repetition_penalty_reward": -0.1132449209690094,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 818
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 601.5416870117188,
-      "epoch": 0.819,
-      "grad_norm": 36.93153228705177,
-      "kl": 4.96875,
-      "learning_rate": 1.8686785953528922e-07,
-      "loss": 1.1533,
-      "reward": 2.5069422721862793,
-      "reward_std": 0.7914581596851349,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.02257181704044342,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 432.0833435058594,
+      "epoch": 0.4095,
+      "grad_norm": 15.736710234210651,
+      "kl": 1.71875,
+      "learning_rate": 7.619553465487344e-07,
+      "loss": 0.5061,
+      "reward": 2.743853211402893,
+      "reward_std": 0.557222306728363,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.08774404600262642,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 819
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 414.00001525878906,
-      "epoch": 0.82,
-      "grad_norm": 36.676997752826466,
-      "kl": 2.37890625,
-      "learning_rate": 1.8594235253127372e-07,
-      "loss": 0.8483,
-      "reward": 2.6289749145507812,
-      "reward_std": 0.5092899203300476,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.020330775994807482,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 654.4583435058594,
+      "epoch": 0.41,
+      "grad_norm": 55.05386297781122,
+      "kl": 2.859375,
+      "learning_rate": 7.612622032536507e-07,
+      "loss": 0.8643,
+      "reward": 2.1351473927497864,
+      "reward_std": 0.6679968535900116,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.161727674305439,
+      "rewards/tag_count_reward": 0.859375,
       "step": 820
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 424.6458435058594,
-      "epoch": 0.821,
-      "grad_norm": 25.786201228130313,
-      "kl": 1.57421875,
-      "learning_rate": 1.850212814548031e-07,
-      "loss": 0.4719,
-      "reward": 2.536083459854126,
-      "reward_std": 0.5136556923389435,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.02815271820873022,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 502.00001525878906,
+      "epoch": 0.4105,
+      "grad_norm": 13.566998114200823,
+      "kl": 0.576171875,
+      "learning_rate": 7.60568416417258e-07,
+      "loss": 0.2932,
+      "reward": 2.584618330001831,
+      "reward_std": 0.4659956693649292,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.13586794957518578,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 821
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 327.3333435058594,
-      "epoch": 0.822,
-      "grad_norm": 9.339930080359409,
-      "kl": 0.5400390625,
-      "learning_rate": 1.8410465752883758e-07,
-      "loss": 0.0109,
-      "reward": 2.5481141805648804,
-      "reward_std": 0.4344000518321991,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.026538672856986523,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 629.7500305175781,
+      "epoch": 0.411,
+      "grad_norm": 30.139911446994017,
+      "kl": 1.54296875,
+      "learning_rate": 7.59873988152951e-07,
+      "loss": 0.6395,
+      "reward": 2.022117495536804,
+      "reward_std": 0.9050773084163666,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.8750000894069672,
+      "rewards/repetition_penalty_reward": -0.1445491872727871,
+      "rewards/tag_count_reward": 0.8333333432674408,
       "step": 822
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 425.04168701171875,
-      "epoch": 0.823,
-      "grad_norm": 17.806285658594135,
-      "kl": 2.37109375,
-      "learning_rate": 1.8319249192215055e-07,
-      "loss": 0.6621,
-      "reward": 2.3081016540527344,
-      "reward_std": 0.5683871209621429,
-      "rewards/accuracy_reward": 0.4375000298023224,
-      "rewards/reasoning_steps_reward": 0.9444444477558136,
-      "rewards/repetition_penalty_reward": -0.02175955567508936,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 747.7500305175781,
+      "epoch": 0.4115,
+      "grad_norm": 33.95615169187785,
+      "kl": 1.16796875,
+      "learning_rate": 7.591789205760789e-07,
+      "loss": 0.7839,
+      "reward": 1.9424059987068176,
+      "reward_std": 0.8640447556972504,
+      "rewards/accuracy_reward": 0.33333333395421505,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.19474690407514572,
+      "rewards/tag_count_reward": 0.8593750298023224,
       "step": 823
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 488.16668701171875,
-      "epoch": 0.824,
-      "grad_norm": 30.757926718689315,
-      "kl": 4.0625,
-      "learning_rate": 1.822847957491922e-07,
-      "loss": 1.0787,
-      "reward": 2.3749157190322876,
-      "reward_std": 0.5852408409118652,
-      "rewards/accuracy_reward": 0.5416666716337204,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.01918159332126379,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 561.3541717529297,
+      "epoch": 0.412,
+      "grad_norm": 8.156901607192008,
+      "kl": 0.73828125,
+      "learning_rate": 7.584832158039378e-07,
+      "loss": 0.1677,
+      "reward": 2.5275344848632812,
+      "reward_std": 0.3939831107854843,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.12871567159891129,
+      "rewards/tag_count_reward": 0.90625,
       "step": 824
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 474.3125305175781,
-      "epoch": 0.825,
-      "grad_norm": 22.249676223818117,
-      "kl": 3.10546875,
-      "learning_rate": 1.8138158006995363e-07,
-      "loss": 0.568,
-      "reward": 2.3771095275878906,
-      "reward_std": 0.5015990436077118,
-      "rewards/accuracy_reward": 0.479166679084301,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.015251755714416504,
-      "rewards/tag_count_reward": 0.9270833730697632,
+      "completion_length": 607.4166870117188,
+      "epoch": 0.4125,
+      "grad_norm": 17.24978480660208,
+      "kl": 1.37109375,
+      "learning_rate": 7.577868759557653e-07,
+      "loss": 0.6142,
+      "reward": 2.3871285915374756,
+      "reward_std": 0.7968339771032333,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.17189929634332657,
+      "rewards/tag_count_reward": 0.8645833730697632,
       "step": 825
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 347.37501525878906,
-      "epoch": 0.826,
-      "grad_norm": 14.195193659523204,
-      "kl": 2.14453125,
-      "learning_rate": 1.804828558898332e-07,
-      "loss": 0.1007,
-      "reward": 2.5919097661972046,
-      "reward_std": 0.468966469168663,
-      "rewards/accuracy_reward": 0.6666666716337204,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.010520970448851585,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 839.1250305175781,
+      "epoch": 0.413,
+      "grad_norm": 20.131860885223457,
+      "kl": 1.71875,
+      "learning_rate": 7.570899031527332e-07,
+      "loss": 0.5612,
+      "reward": 2.0470434427261353,
+      "reward_std": 0.7597078531980515,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.24635940790176392,
+      "rewards/tag_count_reward": 0.7864583432674408,
       "step": 826
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 450.2083435058594,
-      "epoch": 0.827,
-      "grad_norm": 32.764312557604384,
-      "kl": 4.703125,
-      "learning_rate": 1.7958863415950112e-07,
-      "loss": 0.8324,
-      "reward": 2.222867786884308,
-      "reward_std": 0.6226075887680054,
-      "rewards/accuracy_reward": 0.3958333358168602,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.02192395133897662,
-      "rewards/tag_count_reward": 0.9114583730697632,
+      "completion_length": 628.4375305175781,
+      "epoch": 0.4135,
+      "grad_norm": 21.15993434326557,
+      "kl": 0.99609375,
+      "learning_rate": 7.563922995179418e-07,
+      "loss": 0.5685,
+      "reward": 2.4131293296813965,
+      "reward_std": 0.7387427687644958,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.14416248351335526,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 827
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 537.3958587646484,
-      "epoch": 0.828,
-      "grad_norm": 102.68181493594267,
-      "kl": 7.6669921875,
-      "learning_rate": 1.7869892577476722e-07,
-      "loss": 0.8764,
-      "reward": 2.513993501663208,
-      "reward_std": 0.5042714290320873,
-      "rewards/accuracy_reward": 0.7291666716337204,
-      "rewards/reasoning_steps_reward": 0.930555522441864,
-      "rewards/repetition_penalty_reward": -0.025937245693057775,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 370.8333435058594,
+      "epoch": 0.414,
+      "grad_norm": 12.527599564815661,
+      "kl": 0.3623046875,
+      "learning_rate": 7.556940671764124e-07,
+      "loss": 0.1614,
+      "reward": 2.8723835945129395,
+      "reward_std": 0.15300642838701606,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.061644500121474266,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 828
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 428.5833435058594,
-      "epoch": 0.829,
-      "grad_norm": 25.253738055443158,
-      "kl": 4.703125,
-      "learning_rate": 1.7781374157644713e-07,
-      "loss": 0.8549,
-      "reward": 2.5480579137802124,
-      "reward_std": 0.7514591813087463,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.01444217236712575,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 981.0625305175781,
+      "epoch": 0.4145,
+      "grad_norm": 29.125329174569146,
+      "kl": 1.6484375,
+      "learning_rate": 7.54995208255082e-07,
+      "loss": 0.6239,
+      "reward": 1.5005772113800049,
+      "reward_std": 0.6931151449680328,
+      "rewards/accuracy_reward": 0.10416666977107525,
+      "rewards/reasoning_steps_reward": 0.8888888955116272,
+      "rewards/repetition_penalty_reward": -0.11227016523480415,
+      "rewards/tag_count_reward": 0.6197916865348816,
       "step": 829
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 319.7708435058594,
-      "epoch": 0.83,
-      "grad_norm": 43.578316518329956,
-      "kl": 3.408203125,
-      "learning_rate": 1.7693309235023127e-07,
-      "loss": 0.2552,
-      "reward": 2.751266121864319,
-      "reward_std": 0.45656658709049225,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.014358973130583763,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 652.4375,
+      "epoch": 0.415,
+      "grad_norm": 29.203264211648943,
+      "kl": 0.794921875,
+      "learning_rate": 7.54295724882796e-07,
+      "loss": 0.6498,
+      "reward": 2.3145864009857178,
+      "reward_std": 0.7329890131950378,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.8819445073604584,
+      "rewards/repetition_penalty_reward": -0.04652494750916958,
+      "rewards/tag_count_reward": 0.8333333432674408,
       "step": 830
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 485.10418701171875,
-      "epoch": 0.831,
-      "grad_norm": 27.193058696951205,
-      "kl": 5.515625,
-      "learning_rate": 1.7605698882655233e-07,
-      "loss": 0.8536,
-      "reward": 2.412537455558777,
-      "reward_std": 0.8152902722358704,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.04058762267231941,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 405.5625,
+      "epoch": 0.4155,
+      "grad_norm": 9.688376442723252,
+      "kl": 0.556640625,
+      "learning_rate": 7.535956191903021e-07,
+      "loss": 0.1869,
+      "reward": 2.5234345197677612,
+      "reward_std": 0.46155112981796265,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.04080154746770859,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 831
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 331.0208435058594,
-      "epoch": 0.832,
-      "grad_norm": 19.688526945972043,
-      "kl": 2.2265625,
-      "learning_rate": 1.7518544168045524e-07,
-      "loss": 0.3667,
-      "reward": 2.815763831138611,
-      "reward_std": 0.3860451430082321,
-      "rewards/accuracy_reward": 0.875,
-      "rewards/reasoning_steps_reward": 0.9791667461395264,
-      "rewards/repetition_penalty_reward": -0.012361295986920595,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 313.81251525878906,
+      "epoch": 0.416,
+      "grad_norm": 15.677158884669252,
+      "kl": 0.33837890625,
+      "learning_rate": 7.528948933102438e-07,
+      "loss": 0.2505,
+      "reward": 2.81933856010437,
+      "reward_std": 0.3189728558063507,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.958333432674408,
+      "rewards/repetition_penalty_reward": -0.02441159961745143,
+      "rewards/tag_count_reward": 0.96875,
       "step": 832
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 346.56251525878906,
-      "epoch": 0.833,
-      "grad_norm": 22.298341306372343,
-      "kl": 2.765625,
-      "learning_rate": 1.743184615314671e-07,
-      "loss": 0.3607,
-      "reward": 2.7588515281677246,
-      "reward_std": 0.545660674571991,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.02066253498196602,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 471.9791717529297,
+      "epoch": 0.4165,
+      "grad_norm": 24.249907996018568,
+      "kl": 0.767578125,
+      "learning_rate": 7.521935493771534e-07,
+      "loss": 0.418,
+      "reward": 2.4950411319732666,
+      "reward_std": 0.7277962416410446,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.03620896954089403,
+      "rewards/tag_count_reward": 0.8645833730697632,
       "step": 833
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 359.31251525878906,
-      "epoch": 0.834,
-      "grad_norm": 28.797542999117116,
-      "kl": 3.25390625,
-      "learning_rate": 1.7345605894346726e-07,
-      "loss": 0.5221,
-      "reward": 2.7526845932006836,
-      "reward_std": 0.674776554107666,
-      "rewards/accuracy_reward": 0.8958333432674408,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.01641272520646453,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 396.56251525878906,
+      "epoch": 0.417,
+      "grad_norm": 9.707239836993917,
+      "kl": 0.5439453125,
+      "learning_rate": 7.514915895274463e-07,
+      "loss": 0.2468,
+      "reward": 2.515194535255432,
+      "reward_std": 0.30929915606975555,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.03862500563263893,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 834
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 430.62501525878906,
-      "epoch": 0.835,
-      "grad_norm": 43.496982955042476,
-      "kl": 3.0859375,
-      "learning_rate": 1.7259824442455923e-07,
-      "loss": 0.6982,
-      "reward": 2.299278497695923,
-      "reward_std": 0.6011901348829269,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.9027777910232544,
-      "rewards/repetition_penalty_reward": -0.00974938040599227,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 620.3333435058594,
+      "epoch": 0.4175,
+      "grad_norm": 17.508594529679065,
+      "kl": 1.17578125,
+      "learning_rate": 7.507890158994139e-07,
+      "loss": 0.5817,
+      "reward": 2.4108619689941406,
+      "reward_std": 0.8157097101211548,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.028374179266393185,
+      "rewards/tag_count_reward": 0.7864583432674408,
       "step": 835
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 519.9375152587891,
-      "epoch": 0.836,
-      "grad_norm": 23.29360992129975,
-      "kl": 3.484375,
-      "learning_rate": 1.7174502842694212e-07,
-      "loss": 1.0119,
-      "reward": 2.5549784898757935,
-      "reward_std": 0.7014127969741821,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.021410479210317135,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 676.0416870117188,
+      "epoch": 0.418,
+      "grad_norm": 10.440509283265795,
+      "kl": 1.5859375,
+      "learning_rate": 7.500858306332172e-07,
+      "loss": 0.7901,
+      "reward": 2.2338361740112305,
+      "reward_std": 0.7627847492694855,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.024844415485858917,
+      "rewards/tag_count_reward": 0.7447916865348816,
       "step": 836
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 468.81251525878906,
-      "epoch": 0.837,
-      "grad_norm": 21.45595694285051,
-      "kl": 4.984375,
-      "learning_rate": 1.7089642134678364e-07,
-      "loss": 0.8767,
-      "reward": 2.2903480529785156,
-      "reward_std": 0.6241036355495453,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.018679913599044085,
-      "rewards/tag_count_reward": 0.90625,
+      "completion_length": 726.2083435058594,
+      "epoch": 0.4185,
+      "grad_norm": 15.62225883232442,
+      "kl": 2.19921875,
+      "learning_rate": 7.493820358708809e-07,
+      "loss": 0.7195,
+      "reward": 2.36660635471344,
+      "reward_std": 0.7908368110656738,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.018810252659022808,
+      "rewards/tag_count_reward": 0.7812500298023224,
       "step": 837
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 736.0833435058594,
-      "epoch": 0.838,
-      "grad_norm": 62.125475017964185,
-      "kl": 9.53125,
-      "learning_rate": 1.7005243352409333e-07,
-      "loss": 1.49,
-      "reward": 2.3610810041427612,
-      "reward_std": 0.8324085474014282,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9444445073604584,
-      "rewards/repetition_penalty_reward": -0.020863616839051247,
-      "rewards/tag_count_reward": 0.7708333432674408,
+      "completion_length": 925.7291870117188,
+      "epoch": 0.419,
+      "grad_norm": 17.894574750635286,
+      "kl": 3.9296875,
+      "learning_rate": 7.486776337562853e-07,
+      "loss": 0.7237,
+      "reward": 1.4900822043418884,
+      "reward_std": 0.6835044920444489,
+      "rewards/accuracy_reward": 0.0833333358168602,
+      "rewards/reasoning_steps_reward": 0.847222238779068,
+      "rewards/repetition_penalty_reward": -0.023806744255125523,
+      "rewards/tag_count_reward": 0.5833333432674408,
       "step": 838
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 711.6041870117188,
-      "epoch": 0.839,
-      "grad_norm": 51.57175499154389,
-      "kl": 10.3125,
-      "learning_rate": 1.6921307524259625e-07,
-      "loss": 1.4895,
-      "reward": 2.234101891517639,
-      "reward_std": 0.9274267852306366,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.02457870915532112,
-      "rewards/tag_count_reward": 0.7239583730697632,
+      "completion_length": 497.8125305175781,
+      "epoch": 0.4195,
+      "grad_norm": 6.541740244547362,
+      "kl": 1.78515625,
+      "learning_rate": 7.479726264351618e-07,
+      "loss": 0.6408,
+      "reward": 2.4577986001968384,
+      "reward_std": 0.5397596657276154,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9513888657093048,
+      "rewards/repetition_penalty_reward": -0.030048648826777935,
+      "rewards/tag_count_reward": 0.8489583432674408,
       "step": 839
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 431.00001525878906,
-      "epoch": 0.84,
-      "grad_norm": 33.432954925810336,
-      "kl": 2.5546875,
-      "learning_rate": 1.6837835672960831e-07,
-      "loss": 0.8481,
-      "reward": 2.6463335752487183,
-      "reward_std": 0.5603546500205994,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.015124747063964605,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 691.0416870117188,
+      "epoch": 0.42,
+      "grad_norm": 12.527206541745974,
+      "kl": 2.6484375,
+      "learning_rate": 7.472670160550848e-07,
+      "loss": 0.694,
+      "reward": 1.9736045598983765,
+      "reward_std": 0.6236002445220947,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.04722887650132179,
+      "rewards/tag_count_reward": 0.75,
       "step": 840
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 512.1666717529297,
-      "epoch": 0.841,
-      "grad_norm": 66.56725483975436,
-      "kl": 5.359375,
-      "learning_rate": 1.6754828815591131e-07,
-      "loss": 1.1354,
-      "reward": 2.3149657249450684,
-      "reward_std": 0.6210701763629913,
-      "rewards/accuracy_reward": 0.479166679084301,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.01663155946880579,
-      "rewards/tag_count_reward": 0.8802083730697632,
+      "completion_length": 454.9375,
+      "epoch": 0.4205,
+      "grad_norm": 6.487598751220338,
+      "kl": 1.51953125,
+      "learning_rate": 7.46560804765466e-07,
+      "loss": 0.4798,
+      "reward": 2.6140646934509277,
+      "reward_std": 0.6524476110935211,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.03350475896149874,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 841
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 407.85418701171875,
-      "epoch": 0.842,
-      "grad_norm": 12.29872685471777,
-      "kl": 3.1630859375,
-      "learning_rate": 1.6672287963562852e-07,
-      "loss": 0.4297,
-      "reward": 2.510159969329834,
-      "reward_std": 0.602110430598259,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.01761801028624177,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 515.2708587646484,
+      "epoch": 0.421,
+      "grad_norm": 10.317644373992936,
+      "kl": 1.349609375,
+      "learning_rate": 7.458539947175473e-07,
+      "loss": 0.6031,
+      "reward": 2.220026969909668,
+      "reward_std": 0.5868920385837555,
+      "rewards/accuracy_reward": 0.4166666716337204,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.014348076190799475,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 842
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 541.5,
-      "epoch": 0.843,
-      "grad_norm": 27.148373245511962,
-      "kl": 5.609375,
-      "learning_rate": 1.659021412261026e-07,
-      "loss": 0.9403,
-      "reward": 2.315541386604309,
-      "reward_std": 0.7348792850971222,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.024736556224524975,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 379.35418701171875,
+      "epoch": 0.4215,
+      "grad_norm": 7.886525307085669,
+      "kl": 1.072265625,
+      "learning_rate": 7.45146588064395e-07,
+      "loss": 0.3578,
+      "reward": 2.5978939533233643,
+      "reward_std": 0.6561948955059052,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.04273105226457119,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 843
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 594.3750152587891,
-      "epoch": 0.844,
-      "grad_norm": 37.507439921203925,
-      "kl": 6.9140625,
-      "learning_rate": 1.6508608292777203e-07,
-      "loss": 0.4498,
-      "reward": 2.3864521980285645,
-      "reward_std": 0.3542592525482178,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.01806183159351349,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 324.1666717529297,
+      "epoch": 0.422,
+      "grad_norm": 4.37947548836921,
+      "kl": 0.30078125,
+      "learning_rate": 7.444385869608921e-07,
+      "loss": 0.0354,
+      "reward": 2.6675782203674316,
+      "reward_std": 0.17353395372629166,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03381069749593735,
+      "rewards/tag_count_reward": 1.0,
       "step": 844
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 574.5416870117188,
-      "epoch": 0.845,
-      "grad_norm": 42.479348438013346,
-      "kl": 4.875,
-      "learning_rate": 1.6427471468404952e-07,
-      "loss": 1.2607,
-      "reward": 2.191675543785095,
-      "reward_std": 0.5457549095153809,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.018394021317362785,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 372.7083435058594,
+      "epoch": 0.4225,
+      "grad_norm": 18.876751039530795,
+      "kl": 1.12890625,
+      "learning_rate": 7.437299935637328e-07,
+      "loss": 0.45,
+      "reward": 2.5024070739746094,
+      "reward_std": 0.5223306268453598,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03231525328010321,
+      "rewards/tag_count_reward": 0.9375,
       "step": 845
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 447.5,
-      "epoch": 0.846,
-      "grad_norm": 32.21272755825097,
-      "kl": 2.6328125,
-      "learning_rate": 1.6346804638120098e-07,
-      "loss": 0.6288,
-      "reward": 2.359144449234009,
-      "reward_std": 0.7404135763645172,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9236111342906952,
-      "rewards/repetition_penalty_reward": -0.017591833136975765,
-      "rewards/tag_count_reward": 0.9114583730697632,
+      "completion_length": 368.4583435058594,
+      "epoch": 0.423,
+      "grad_norm": 7.735517734052216,
+      "kl": 0.373046875,
+      "learning_rate": 7.430208100314156e-07,
+      "loss": 0.2957,
+      "reward": 2.737704277038574,
+      "reward_std": 0.41725849360227585,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.04701795056462288,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 846
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 384.18751525878906,
-      "epoch": 0.847,
-      "grad_norm": 14.416271559769205,
-      "kl": 2.5927734375,
-      "learning_rate": 1.6266608784822542e-07,
-      "loss": 0.393,
-      "reward": 2.211021304130554,
-      "reward_std": 0.4275623857975006,
+      "completion_length": 383.68751525878906,
+      "epoch": 0.4235,
+      "grad_norm": 4.3638722065625855,
+      "kl": 0.3779296875,
+      "learning_rate": 7.423110385242366e-07,
+      "loss": 0.0939,
+      "reward": 2.3006144762039185,
+      "reward_std": 0.29770078510046005,
       "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.014673332683742046,
-      "rewards/tag_count_reward": 0.90625,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.032718876376748085,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 847
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 436.5625,
-      "epoch": 0.848,
-      "grad_norm": 14.205493656115495,
-      "kl": 3.384765625,
-      "learning_rate": 1.6186884885673413e-07,
-      "loss": 0.5267,
-      "reward": 2.2585190534591675,
-      "reward_std": 0.7064023613929749,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.03314755018800497,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 337.2083435058594,
+      "epoch": 0.424,
+      "grad_norm": 9.281897298394469,
+      "kl": 0.65625,
+      "learning_rate": 7.416006812042827e-07,
+      "loss": 0.2155,
+      "reward": 2.708508014678955,
+      "reward_std": 0.40016093850135803,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06406151317059994,
+      "rewards/tag_count_reward": 0.9739583730697632,
       "step": 848
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 496.375,
-      "epoch": 0.849,
-      "grad_norm": 47.615315210946434,
-      "kl": 3.21484375,
-      "learning_rate": 1.610763391208329e-07,
-      "loss": 0.7123,
-      "reward": 2.543179988861084,
-      "reward_std": 0.6331344544887543,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.01584793906658888,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 348.37501525878906,
+      "epoch": 0.4245,
+      "grad_norm": 9.904867829218396,
+      "kl": 0.6005859375,
+      "learning_rate": 7.408897402354255e-07,
+      "loss": 0.1548,
+      "reward": 2.447229743003845,
+      "reward_std": 0.3772226721048355,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.026728657074272633,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 849
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 330.125,
-      "epoch": 0.85,
-      "grad_norm": 22.926928355727185,
-      "kl": 1.275390625,
-      "learning_rate": 1.6028856829700258e-07,
-      "loss": 0.2793,
-      "reward": 2.8741886615753174,
-      "reward_std": 0.27669021487236023,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.011228146031498909,
-      "rewards/tag_count_reward": 0.9687500298023224,
+      "completion_length": 424.04168701171875,
+      "epoch": 0.425,
+      "grad_norm": 5.890746892326936,
+      "kl": 0.86328125,
+      "learning_rate": 7.401782177833147e-07,
+      "loss": 0.4768,
+      "reward": 2.3541191816329956,
+      "reward_std": 0.5978062450885773,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.05039471574127674,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 850
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 352.6875,
-      "epoch": 0.851,
-      "grad_norm": 23.63650850429641,
-      "kl": 1.009765625,
-      "learning_rate": 1.5950554598398228e-07,
-      "loss": 0.2369,
-      "reward": 2.6674481630325317,
-      "reward_std": 0.3812839537858963,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.014843634329736233,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 361.3958435058594,
+      "epoch": 0.4255,
+      "grad_norm": 19.104666432985564,
+      "kl": 0.65234375,
+      "learning_rate": 7.394661160153709e-07,
+      "loss": 0.4226,
+      "reward": 2.7522761821746826,
+      "reward_std": 0.5294878482818604,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.020293288864195347,
+      "rewards/tag_count_reward": 0.953125,
       "step": 851
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 504.22918701171875,
-      "epoch": 0.852,
-      "grad_norm": 26.46847605504948,
-      "kl": 6.5625,
-      "learning_rate": 1.5872728172265146e-07,
-      "loss": 0.9065,
-      "reward": 2.415515661239624,
-      "reward_std": 0.7073606848716736,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9444445073604584,
-      "rewards/repetition_penalty_reward": -0.01330387475900352,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 398.9583435058594,
+      "epoch": 0.426,
+      "grad_norm": 7.305881246017198,
+      "kl": 0.7138671875,
+      "learning_rate": 7.387534371007797e-07,
+      "loss": 0.2967,
+      "reward": 2.524535655975342,
+      "reward_std": 0.29497088491916656,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.05706173926591873,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 852
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 561.5416870117188,
-      "epoch": 0.853,
-      "grad_norm": 25.999454906190575,
-      "kl": 4.828125,
-      "learning_rate": 1.579537849959148e-07,
-      "loss": 1.113,
-      "reward": 2.4573510885238647,
-      "reward_std": 0.7351089715957642,
-      "rewards/accuracy_reward": 0.6458333730697632,
+      "completion_length": 294.7291717529297,
+      "epoch": 0.4265,
+      "grad_norm": 4.07471559844156,
+      "kl": 0.3271484375,
+      "learning_rate": 7.380401832104845e-07,
+      "loss": 0.0466,
+      "reward": 2.9143292903900146,
+      "reward_std": 0.11590113118290901,
+      "rewards/accuracy_reward": 0.9791666865348816,
       "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.021815632469952106,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "rewards/repetition_penalty_reward": -0.044004119001328945,
+      "rewards/tag_count_reward": 1.0,
       "step": 853
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 511.2708435058594,
-      "epoch": 0.854,
-      "grad_norm": 32.177866129176245,
-      "kl": 5.859375,
-      "learning_rate": 1.5718506522858572e-07,
-      "loss": 0.9686,
-      "reward": 2.1545926332473755,
-      "reward_std": 0.6024680733680725,
-      "rewards/accuracy_reward": 0.3541666865348816,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.031171313486993313,
-      "rewards/tag_count_reward": 0.8802083730697632,
+      "completion_length": 460.12501525878906,
+      "epoch": 0.427,
+      "grad_norm": 14.708604251307426,
+      "kl": 1.111328125,
+      "learning_rate": 7.373263565171805e-07,
+      "loss": 0.5831,
+      "reward": 2.3425523042678833,
+      "reward_std": 0.6779049038887024,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.020295153371989727,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 854
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 464.5416717529297,
-      "epoch": 0.855,
-      "grad_norm": 35.98943778638736,
-      "kl": 2.75,
-      "learning_rate": 1.5642113178727193e-07,
-      "loss": 0.8586,
-      "reward": 2.234099566936493,
-      "reward_std": 0.4664178192615509,
-      "rewards/accuracy_reward": 0.3541666679084301,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.028053374961018562,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 337.7916717529297,
+      "epoch": 0.4275,
+      "grad_norm": 11.608809720420162,
+      "kl": 0.7373046875,
+      "learning_rate": 7.366119591953075e-07,
+      "loss": 0.1999,
+      "reward": 2.80188250541687,
+      "reward_std": 0.3770581018179655,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.03492319490760565,
+      "rewards/tag_count_reward": 0.96875,
       "step": 855
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 539.6250305175781,
-      "epoch": 0.856,
-      "grad_norm": 25.61587236309964,
-      "kl": 4.1640625,
-      "learning_rate": 1.5566199398026147e-07,
-      "loss": 1.0126,
-      "reward": 2.0876386165618896,
-      "reward_std": 0.7813932299613953,
-      "rewards/accuracy_reward": 0.3333333358168602,
-      "rewards/reasoning_steps_reward": 0.916666716337204,
-      "rewards/repetition_penalty_reward": -0.02173637691885233,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 514.4375,
+      "epoch": 0.428,
+      "grad_norm": 15.274727479273395,
+      "kl": 2.109375,
+      "learning_rate": 7.358969934210438e-07,
+      "loss": 0.4943,
+      "reward": 2.3344244956970215,
+      "reward_std": 0.38865962624549866,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.028422830160707235,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 856
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 394.68751525878906,
-      "epoch": 0.857,
-      "grad_norm": 21.20771571565506,
-      "kl": 2.45703125,
-      "learning_rate": 1.5490766105740876e-07,
-      "loss": 0.5834,
-      "reward": 2.4374274015426636,
-      "reward_std": 0.4593869596719742,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.019169961102306843,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 539.9166870117188,
+      "epoch": 0.4285,
+      "grad_norm": 21.14206646504488,
+      "kl": 2.140625,
+      "learning_rate": 7.35181461372299e-07,
+      "loss": 1.0918,
+      "reward": 2.400139808654785,
+      "reward_std": 0.6591689586639404,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9444443881511688,
+      "rewards/repetition_penalty_reward": -0.044304635375738144,
+      "rewards/tag_count_reward": 0.8750000298023224,
       "step": 857
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 338.50001525878906,
-      "epoch": 0.858,
-      "grad_norm": 11.186901565270288,
-      "kl": 2.41796875,
-      "learning_rate": 1.5415814221002265e-07,
-      "loss": 0.1621,
-      "reward": 2.769268274307251,
-      "reward_std": 0.46035242080688477,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.024134621024131775,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 492.62501525878906,
+      "epoch": 0.429,
+      "grad_norm": 17.16519433385956,
+      "kl": 1.72265625,
+      "learning_rate": 7.344653652287077e-07,
+      "loss": 0.7656,
+      "reward": 2.677896022796631,
+      "reward_std": 0.6244023740291595,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.03217345476150513,
+      "rewards/tag_count_reward": 0.8906250298023224,
       "step": 858
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 433.9166717529297,
-      "epoch": 0.859,
-      "grad_norm": 26.347652823757198,
-      "kl": 2.93359375,
-      "learning_rate": 1.5341344657075354e-07,
-      "loss": 0.7346,
-      "reward": 2.543567419052124,
-      "reward_std": 0.6669409871101379,
+      "completion_length": 538.0416717529297,
+      "epoch": 0.4295,
+      "grad_norm": 13.397818376841236,
+      "kl": 2.8046875,
+      "learning_rate": 7.337487071716232e-07,
+      "loss": 0.5746,
+      "reward": 2.45145046710968,
+      "reward_std": 0.6907146573066711,
       "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.02066876133903861,
-      "rewards/tag_count_reward": 0.9114583730697632,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.034660689532756805,
+      "rewards/tag_count_reward": 0.8541666865348816,
       "step": 859
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 308.7708435058594,
-      "epoch": 0.86,
-      "grad_norm": 24.14563880775655,
-      "kl": 1.58203125,
-      "learning_rate": 1.5267358321348285e-07,
-      "loss": 0.3053,
-      "reward": 2.8249399662017822,
-      "reward_std": 0.4262206554412842,
-      "rewards/accuracy_reward": 0.8750000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.01186564564704895,
-      "rewards/tag_count_reward": 0.96875,
+      "completion_length": 389.41668701171875,
+      "epoch": 0.43,
+      "grad_norm": 10.960948707755849,
+      "kl": 1.181640625,
+      "learning_rate": 7.330314893841101e-07,
+      "loss": 0.2798,
+      "reward": 2.515633225440979,
+      "reward_std": 0.4169411063194275,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.02603358030319214,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 860
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 364.9583435058594,
-      "epoch": 0.861,
-      "grad_norm": 30.59682645079428,
-      "kl": 2.7890625,
-      "learning_rate": 1.5193856115321224e-07,
-      "loss": 0.4332,
-      "reward": 2.4405341148376465,
-      "reward_std": 0.4027775228023529,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.01779924053698778,
-      "rewards/tag_count_reward": 0.9375000298023224,
+      "completion_length": 288.7708435058594,
+      "epoch": 0.4305,
+      "grad_norm": 3.4791756020959634,
+      "kl": 0.291015625,
+      "learning_rate": 7.323137140509381e-07,
+      "loss": 0.0551,
+      "reward": 2.6766408681869507,
+      "reward_std": 0.10699502378702164,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03169257752597332,
+      "rewards/tag_count_reward": 1.0,
       "step": 861
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 432.06251525878906,
-      "epoch": 0.862,
-      "grad_norm": 36.09673783797281,
-      "kl": 6.0625,
-      "learning_rate": 1.5120838934595337e-07,
-      "loss": 0.8199,
-      "reward": 2.335049033164978,
-      "reward_std": 0.7307652831077576,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.01738166157156229,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 492.5416717529297,
+      "epoch": 0.431,
+      "grad_norm": 184.11729136527305,
+      "kl": 8.76171875,
+      "learning_rate": 7.315953833585755e-07,
+      "loss": 0.8431,
+      "reward": 2.497468113899231,
+      "reward_std": 0.5085512399673462,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9444444179534912,
+      "rewards/repetition_penalty_reward": -0.025101464241743088,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 862
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 502.0208435058594,
-      "epoch": 0.863,
-      "grad_norm": 53.67270636598954,
-      "kl": 6.25,
-      "learning_rate": 1.5048307668861947e-07,
-      "loss": 1.0301,
-      "reward": 2.3041592836380005,
-      "reward_std": 0.6167122721672058,
-      "rewards/accuracy_reward": 0.4375000223517418,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.017021275125443935,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 547.9166870117188,
+      "epoch": 0.4315,
+      "grad_norm": 634.497752315967,
+      "kl": 19.375,
+      "learning_rate": 7.308764994951821e-07,
+      "loss": 2.7713,
+      "reward": 2.3461129665374756,
+      "reward_std": 0.7580728530883789,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.014998285099864006,
+      "rewards/tag_count_reward": 0.8750000298023224,
       "step": 863
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 502.625,
-      "epoch": 0.864,
-      "grad_norm": 40.13186559524844,
-      "kl": 6.4140625,
-      "learning_rate": 1.4976263201891613e-07,
-      "loss": 1.1658,
-      "reward": 2.634553909301758,
-      "reward_std": 0.6430298089981079,
+      "completion_length": 405.0208435058594,
+      "epoch": 0.432,
+      "grad_norm": 51.28196086116214,
+      "kl": 3.462890625,
+      "learning_rate": 7.301570646506027e-07,
+      "loss": 0.6456,
+      "reward": 2.7293022871017456,
+      "reward_std": 0.44555216282606125,
       "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9444445073604584,
-      "rewards/repetition_penalty_reward": -0.01822400465607643,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.013753433711826801,
+      "rewards/tag_count_reward": 0.9583333730697632,
       "step": 864
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 602.0625152587891,
-      "epoch": 0.865,
-      "grad_norm": 67.05207348072643,
-      "kl": 9.6796875,
-      "learning_rate": 1.4904706411523448e-07,
-      "loss": 1.2048,
-      "reward": 2.128021478652954,
-      "reward_std": 0.7044219970703125,
-      "rewards/accuracy_reward": 0.416666679084301,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.0178119083866477,
-      "rewards/tag_count_reward": 0.7916666865348816,
+      "completion_length": 292.3541717529297,
+      "epoch": 0.4325,
+      "grad_norm": 24.09046766028715,
+      "kl": 1.6337890625,
+      "learning_rate": 7.294370810163607e-07,
+      "loss": 0.192,
+      "reward": 2.6840856075286865,
+      "reward_std": 0.17013289034366608,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.013831105083227158,
+      "rewards/tag_count_reward": 0.96875,
       "step": 865
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 473.0,
-      "epoch": 0.866,
-      "grad_norm": 17.39107169503413,
-      "kl": 5.40625,
-      "learning_rate": 1.483363816965435e-07,
-      "loss": 1.0166,
-      "reward": 2.366263747215271,
-      "reward_std": 0.7000749707221985,
-      "rewards/accuracy_reward": 0.5625000149011612,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.024361333809792995,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 589.7500152587891,
+      "epoch": 0.433,
+      "grad_norm": 405.26278466771953,
+      "kl": 12.75,
+      "learning_rate": 7.287165507856512e-07,
+      "loss": 1.1717,
+      "reward": 2.1014411449432373,
+      "reward_std": 0.6244297027587891,
+      "rewards/accuracy_reward": 0.3333333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.021822815760970116,
+      "rewards/tag_count_reward": 0.8593750298023224,
       "step": 866
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 589.4166870117188,
-      "epoch": 0.867,
-      "grad_norm": 52.2532552532266,
-      "kl": 8.546875,
-      "learning_rate": 1.4763059342228434e-07,
-      "loss": 1.0051,
-      "reward": 2.1066824197769165,
-      "reward_std": 0.6221250742673874,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.009637102019041777,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 429.8333435058594,
+      "epoch": 0.4335,
+      "grad_norm": 64.46281032488784,
+      "kl": 4.21484375,
+      "learning_rate": 7.279954761533342e-07,
+      "loss": 0.5701,
+      "reward": 2.4746246337890625,
+      "reward_std": 0.7078258693218231,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.0323198651894927,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 867
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 482.3541717529297,
-      "epoch": 0.868,
-      "grad_norm": 34.11769671109303,
-      "kl": 4.921875,
-      "learning_rate": 1.469297078922642e-07,
-      "loss": 0.8941,
-      "reward": 2.2735098600387573,
-      "reward_std": 0.3796389400959015,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.92361119389534,
-      "rewards/repetition_penalty_reward": -0.01468472508713603,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 502.9583435058594,
+      "epoch": 0.434,
+      "grad_norm": 21.766406267926293,
+      "kl": 2.9140625,
+      "learning_rate": 7.27273859315928e-07,
+      "loss": 0.8102,
+      "reward": 2.275583267211914,
+      "reward_std": 0.7442755401134491,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.023027977906167507,
+      "rewards/tag_count_reward": 0.8541666865348816,
       "step": 868
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 419.0208435058594,
-      "epoch": 0.869,
-      "grad_norm": 28.34872735283239,
-      "kl": 3.4375,
-      "learning_rate": 1.4623373364655223e-07,
-      "loss": 0.9823,
-      "reward": 2.6034168004989624,
-      "reward_std": 0.6418764889240265,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.014638913795351982,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 388.0833435058594,
+      "epoch": 0.4345,
+      "grad_norm": 35.27655318443123,
+      "kl": 0.978515625,
+      "learning_rate": 7.265517024716026e-07,
+      "loss": 0.5909,
+      "reward": 2.544116973876953,
+      "reward_std": 0.6133610606193542,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.03574422746896744,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 869
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 400.56251525878906,
-      "epoch": 0.87,
-      "grad_norm": 17.236700769600052,
-      "kl": 2.61328125,
-      "learning_rate": 1.4554267916537495e-07,
-      "loss": 0.5795,
-      "reward": 2.361931324005127,
-      "reward_std": 0.5499020516872406,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 0.9375000298023224,
-      "rewards/repetition_penalty_reward": -0.018277212977409363,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 370.5416717529297,
+      "epoch": 0.435,
+      "grad_norm": 12.299061579151882,
+      "kl": 0.912109375,
+      "learning_rate": 7.258290078201731e-07,
+      "loss": 0.2876,
+      "reward": 2.400050640106201,
+      "reward_std": 0.3678786903619766,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.025296634063124657,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 870
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 570.1875152587891,
-      "epoch": 0.871,
-      "grad_norm": 27.373915984285357,
-      "kl": 5.46875,
-      "learning_rate": 1.448565528690129e-07,
-      "loss": 1.2036,
-      "reward": 2.156299114227295,
-      "reward_std": 0.615426778793335,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.92361119389534,
-      "rewards/repetition_penalty_reward": -0.027728760614991188,
-      "rewards/tag_count_reward": 0.84375,
+      "completion_length": 263.6041793823242,
+      "epoch": 0.4355,
+      "grad_norm": 3.4077521243834212,
+      "kl": 0.22705078125,
+      "learning_rate": 7.251057775630927e-07,
+      "loss": 0.0145,
+      "reward": 2.8228542804718018,
+      "reward_std": 0.17518935352563858,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.017423711717128754,
+      "rewards/tag_count_reward": 1.0,
       "step": 871
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 488.875,
-      "epoch": 0.872,
-      "grad_norm": 25.43206804197954,
-      "kl": 3.91796875,
-      "learning_rate": 1.4417536311769885e-07,
-      "loss": 0.8613,
-      "reward": 2.2447725534439087,
-      "reward_std": 0.6266459226608276,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.012171986512839794,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 358.2083435058594,
+      "epoch": 0.436,
+      "grad_norm": 30.74690332937046,
+      "kl": 0.7333984375,
+      "learning_rate": 7.243820139034464e-07,
+      "loss": 0.3554,
+      "reward": 2.778803586959839,
+      "reward_std": 0.4877840280532837,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.018071786500513554,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 872
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 385.1458435058594,
-      "epoch": 0.873,
-      "grad_norm": 17.40833999493999,
-      "kl": 2.28125,
-      "learning_rate": 1.4349911821151462e-07,
-      "loss": 0.3716,
-      "reward": 2.3771172761917114,
-      "reward_std": 0.5491065829992294,
-      "rewards/accuracy_reward": 0.4791666716337204,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.016980044543743134,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 382.04168701171875,
+      "epoch": 0.4365,
+      "grad_norm": 39.70789630630236,
+      "kl": 0.978515625,
+      "learning_rate": 7.236577190459433e-07,
+      "loss": 0.5793,
+      "reward": 2.512961983680725,
+      "reward_std": 0.49974559247493744,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.02349638007581234,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 873
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 504.31251525878906,
-      "epoch": 0.874,
-      "grad_norm": 33.796267710350925,
-      "kl": 3.75,
-      "learning_rate": 1.4282782639029128e-07,
-      "loss": 0.8301,
-      "reward": 2.4175636768341064,
-      "reward_std": 0.5949557721614838,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.02167257433757186,
-      "rewards/tag_count_reward": 0.890625,
+      "completion_length": 390.8958435058594,
+      "epoch": 0.437,
+      "grad_norm": 38.28149680062182,
+      "kl": 1.4296875,
+      "learning_rate": 7.229328951969115e-07,
+      "loss": 0.6429,
+      "reward": 2.5388529300689697,
+      "reward_std": 0.7759108543395996,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.03406373132020235,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 874
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 591.6041870117188,
-      "epoch": 0.875,
-      "grad_norm": 26.046786075949313,
-      "kl": 7.28125,
-      "learning_rate": 1.4216149583350755e-07,
-      "loss": 1.0826,
-      "reward": 2.029456853866577,
-      "reward_std": 0.6440670937299728,
-      "rewards/accuracy_reward": 0.3125000111758709,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.015682112891227007,
-      "rewards/tag_count_reward": 0.7812500298023224,
+      "completion_length": 387.2708435058594,
+      "epoch": 0.4375,
+      "grad_norm": 29.933419477194974,
+      "kl": 1.177734375,
+      "learning_rate": 7.222075445642904e-07,
+      "loss": 0.4695,
+      "reward": 2.6218247413635254,
+      "reward_std": 0.47380533814430237,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.024008896201848984,
+      "rewards/tag_count_reward": 0.9583333730697632,
       "step": 875
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 409.10418701171875,
-      "epoch": 0.876,
-      "grad_norm": 27.02250848782874,
-      "kl": 2.4375,
-      "learning_rate": 1.4150013466019114e-07,
-      "loss": 0.5262,
-      "reward": 2.6342684030532837,
-      "reward_std": 0.5182087272405624,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.023717753123492002,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 323.4791717529297,
+      "epoch": 0.438,
+      "grad_norm": 17.340458090206823,
+      "kl": 1.15283203125,
+      "learning_rate": 7.214816693576234e-07,
+      "loss": 0.3436,
+      "reward": 2.513605237007141,
+      "reward_std": 0.3129299432039261,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.024589229375123978,
+      "rewards/tag_count_reward": 0.96875,
       "step": 876
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 640.5208740234375,
-      "epoch": 0.877,
-      "grad_norm": 37.146004800834625,
-      "kl": 6.90625,
-      "learning_rate": 1.4084375092881917e-07,
-      "loss": 1.1224,
-      "reward": 2.2963308095932007,
-      "reward_std": 0.7272785305976868,
-      "rewards/accuracy_reward": 0.5833333730697632,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.024849796667695045,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 294.0208435058594,
+      "epoch": 0.4385,
+      "grad_norm": 3.2301186364234344,
+      "kl": 0.24853515625,
+      "learning_rate": 7.207552717880522e-07,
+      "loss": 0.04,
+      "reward": 2.6579480171203613,
+      "reward_std": 0.24716416746377945,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02955202106386423,
+      "rewards/tag_count_reward": 1.0,
       "step": 877
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 498.75001525878906,
-      "epoch": 0.878,
-      "grad_norm": 22.18515714015866,
-      "kl": 4.2578125,
-      "learning_rate": 1.4019235263722034e-07,
-      "loss": 0.8769,
-      "reward": 2.144127130508423,
-      "reward_std": 0.4630337953567505,
-      "rewards/accuracy_reward": 0.3125000074505806,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.015595164615660906,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 292.2291717529297,
+      "epoch": 0.439,
+      "grad_norm": 20.50363538611729,
+      "kl": 1.21435546875,
+      "learning_rate": 7.200283540683102e-07,
+      "loss": 0.1637,
+      "reward": 2.731342077255249,
+      "reward_std": 0.4017653465270996,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03254689462482929,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 878
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 486.8333435058594,
-      "epoch": 0.879,
-      "grad_norm": 33.22887528150052,
-      "kl": 4.6015625,
-      "learning_rate": 1.395459477224772e-07,
-      "loss": 0.8573,
-      "reward": 2.4029178619384766,
-      "reward_std": 0.5731684267520905,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.02763769868761301,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 319.0833435058594,
+      "epoch": 0.4395,
+      "grad_norm": 123.17102900336207,
+      "kl": 5.9921875,
+      "learning_rate": 7.193009184127145e-07,
+      "loss": 0.8709,
+      "reward": 2.4152499437332153,
+      "reward_std": 0.3710479885339737,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.0222501028329134,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 879
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 386.7291717529297,
-      "epoch": 0.88,
-      "grad_norm": 20.990104821906048,
-      "kl": 1.6796875,
-      "learning_rate": 1.3890454406082956e-07,
-      "loss": 0.5235,
-      "reward": 2.454037666320801,
-      "reward_std": 0.4934305101633072,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.01644855784252286,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 448.12501525878906,
+      "epoch": 0.44,
+      "grad_norm": 363.0002951670322,
+      "kl": 14.9375,
+      "learning_rate": 7.185729670371604e-07,
+      "loss": 2.2049,
+      "reward": 2.466444969177246,
+      "reward_std": 0.6788784861564636,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.01966631133109331,
+      "rewards/tag_count_reward": 0.8750000298023224,
       "step": 880
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 502.29168701171875,
-      "epoch": 0.881,
-      "grad_norm": 36.364235182054,
-      "kl": 5.3828125,
-      "learning_rate": 1.3826814946757888e-07,
-      "loss": 1.4712,
-      "reward": 2.4933007955551147,
-      "reward_std": 0.7101709246635437,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.015379873104393482,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 355.625,
+      "epoch": 0.4405,
+      "grad_norm": 377.3905680062747,
+      "kl": 13.265625,
+      "learning_rate": 7.17844502159114e-07,
+      "loss": 1.9299,
+      "reward": 2.5884649753570557,
+      "reward_std": 0.302852138876915,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.017437931150197983,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 881
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 520.25,
-      "epoch": 0.882,
-      "grad_norm": 60.16390554825615,
-      "kl": 6.0546875,
-      "learning_rate": 1.3763677169699217e-07,
-      "loss": 1.0879,
-      "reward": 2.3464200496673584,
-      "reward_std": 0.5781585574150085,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.019899526610970497,
-      "rewards/tag_count_reward": 0.8802083730697632,
+      "completion_length": 299.10418701171875,
+      "epoch": 0.441,
+      "grad_norm": 81.195858576975,
+      "kl": 3.01171875,
+      "learning_rate": 7.171155259976057e-07,
+      "loss": 0.8336,
+      "reward": 2.445673108100891,
+      "reward_std": 0.44037002325057983,
+      "rewards/accuracy_reward": 0.5625000223517418,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.026549202390015125,
+      "rewards/tag_count_reward": 0.9583333730697632,
       "step": 882
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 499.5625,
-      "epoch": 0.883,
-      "grad_norm": 34.74857681904962,
-      "kl": 6.4375,
-      "learning_rate": 1.370104184422085e-07,
-      "loss": 0.5873,
-      "reward": 2.2807507514953613,
-      "reward_std": 0.5849172174930573,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9166667759418488,
-      "rewards/repetition_penalty_reward": -0.01091603422537446,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 420.3541717529297,
+      "epoch": 0.4415,
+      "grad_norm": 51.10985682493742,
+      "kl": 4.3125,
+      "learning_rate": 7.163860407732231e-07,
+      "loss": 0.9492,
+      "reward": 2.43084716796875,
+      "reward_std": 0.4608851373195648,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.02575009036809206,
+      "rewards/tag_count_reward": 0.921875,
       "step": 883
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 448.31251525878906,
-      "epoch": 0.884,
-      "grad_norm": 24.793424614985756,
-      "kl": 3.4609375,
-      "learning_rate": 1.3638909733514452e-07,
-      "loss": 1.0009,
-      "reward": 2.423251986503601,
-      "reward_std": 0.6239519119262695,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.01598423975519836,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 366.79168701171875,
+      "epoch": 0.442,
+      "grad_norm": 145.68633226852327,
+      "kl": 7.15625,
+      "learning_rate": 7.156560487081051e-07,
+      "loss": 0.8389,
+      "reward": 2.534042716026306,
+      "reward_std": 0.6869227886199951,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.019776692148298025,
+      "rewards/tag_count_reward": 0.8802083730697632,
       "step": 884
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 541.0,
-      "epoch": 0.885,
-      "grad_norm": 40.71564310689436,
-      "kl": 6.109375,
-      "learning_rate": 1.3577281594640182e-07,
-      "loss": 0.9699,
-      "reward": 2.234649658203125,
-      "reward_std": 0.5764244198799133,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.01708643836900592,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 351.125,
+      "epoch": 0.4425,
+      "grad_norm": 31.81128878392762,
+      "kl": 1.453125,
+      "learning_rate": 7.149255520259338e-07,
+      "loss": 0.4362,
+      "reward": 2.469942808151245,
+      "reward_std": 0.6845696568489075,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556416511536,
+      "rewards/repetition_penalty_reward": -0.017904633656144142,
+      "rewards/tag_count_reward": 0.9114583730697632,
       "step": 885
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 400.8541717529297,
-      "epoch": 0.886,
-      "grad_norm": 26.123391575885496,
-      "kl": 3.5625,
-      "learning_rate": 1.351615817851748e-07,
-      "loss": 0.8407,
-      "reward": 2.6832687854766846,
-      "reward_std": 0.4722274839878082,
-      "rewards/accuracy_reward": 0.8333333730697632,
-      "rewards/reasoning_steps_reward": 0.9236111640930176,
-      "rewards/repetition_penalty_reward": -0.021592404693365097,
-      "rewards/tag_count_reward": 0.9479166865348816,
-      "step": 886
-    },
-    {
-      "clip_ratio": 0.0,
-      "completion_length": 527.3333587646484,
-      "epoch": 0.887,
-      "grad_norm": 20.992724819526305,
-      "kl": 5.24609375,
-      "learning_rate": 1.345554022991586e-07,
-      "loss": 1.196,
-      "reward": 2.380396842956543,
-      "reward_std": 0.6401159465312958,
-      "rewards/accuracy_reward": 0.6041666716337204,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.011964202858507633,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 276.3541717529297,
+      "epoch": 0.443,
+      "grad_norm": 4.828439869270836,
+      "kl": 0.30419921875,
+      "learning_rate": 7.141945529519288e-07,
+      "loss": 0.049,
+      "reward": 2.859205484390259,
+      "reward_std": 0.2999560683965683,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.026211323216557503,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 886
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 434.25001525878906,
+      "epoch": 0.4435,
+      "grad_norm": 18.69245813013392,
+      "kl": 1.875,
+      "learning_rate": 7.134630537128403e-07,
+      "loss": 0.4583,
+      "reward": 2.5157185792922974,
+      "reward_std": 0.4226338863372803,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.958333432674408,
+      "rewards/repetition_penalty_reward": -0.02073993068188429,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 887
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 429.3541717529297,
-      "epoch": 0.888,
-      "grad_norm": 18.876114754116053,
-      "kl": 2.08984375,
-      "learning_rate": 1.3395428487445914e-07,
-      "loss": 0.5497,
-      "reward": 2.315842390060425,
-      "reward_std": 0.41791072487831116,
-      "rewards/accuracy_reward": 0.3958333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.014018761925399303,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 262.54168701171875,
+      "epoch": 0.444,
+      "grad_norm": 8.211191313929003,
+      "kl": 0.3388671875,
+      "learning_rate": 7.127310565369415e-07,
+      "loss": 0.0778,
+      "reward": 2.9123250246047974,
+      "reward_std": 0.16250279545783997,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03559182398021221,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 888
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 433.9166717529297,
-      "epoch": 0.889,
-      "grad_norm": 20.1828507834608,
-      "kl": 4.0234375,
-      "learning_rate": 1.3335823683550237e-07,
-      "loss": 0.7386,
-      "reward": 2.623020648956299,
-      "reward_std": 0.6058576703071594,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.979166716337204,
-      "rewards/repetition_penalty_reward": -0.017604444175958633,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 381.1666717529297,
+      "epoch": 0.4445,
+      "grad_norm": 39.63561659258511,
+      "kl": 1.671875,
+      "learning_rate": 7.11998563654023e-07,
+      "loss": 0.4953,
+      "reward": 2.48183274269104,
+      "reward_std": 0.5570933073759079,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.03379237279295921,
+      "rewards/tag_count_reward": 0.9114583730697632,
       "step": 889
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 498.7500305175781,
-      "epoch": 0.89,
-      "grad_norm": 21.280761697718606,
-      "kl": 4.3125,
-      "learning_rate": 1.3276726544494571e-07,
-      "loss": 0.8813,
-      "reward": 2.311787486076355,
-      "reward_std": 0.6657915413379669,
-      "rewards/accuracy_reward": 0.5,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.019809790886938572,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 408.1875228881836,
+      "epoch": 0.445,
+      "grad_norm": 19.45400997832264,
+      "kl": 2.63671875,
+      "learning_rate": 7.11265577295385e-07,
+      "loss": 0.5174,
+      "reward": 2.653844475746155,
+      "reward_std": 0.4112878441810608,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.012822255957871675,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 890
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 408.9166717529297,
-      "epoch": 0.891,
-      "grad_norm": 15.5213911574011,
-      "kl": 3.2626953125,
-      "learning_rate": 1.3218137790358892e-07,
-      "loss": 0.3428,
-      "reward": 2.777226448059082,
-      "reward_std": 0.3445788323879242,
-      "rewards/accuracy_reward": 0.8958333432674408,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.016176452860236168,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 298.5208435058594,
+      "epoch": 0.4455,
+      "grad_norm": 40.00263159392063,
+      "kl": 1.97265625,
+      "learning_rate": 7.105320996938314e-07,
+      "loss": 0.5189,
+      "reward": 2.7283371686935425,
+      "reward_std": 0.4932422488927841,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.030343500431627035,
+      "rewards/tag_count_reward": 0.953125,
       "step": 891
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 547.1458435058594,
-      "epoch": 0.892,
-      "grad_norm": 26.69864519239987,
-      "kl": 5.2734375,
-      "learning_rate": 1.316005813502869e-07,
-      "loss": 0.735,
-      "reward": 2.280866265296936,
-      "reward_std": 0.5884718000888824,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.017744951881468296,
-      "rewards/tag_count_reward": 0.875,
+      "completion_length": 394.2083435058594,
+      "epoch": 0.446,
+      "grad_norm": 41.93561167127058,
+      "kl": 4.359375,
+      "learning_rate": 7.097981330836616e-07,
+      "loss": 1.3479,
+      "reward": 2.7040258646011353,
+      "reward_std": 0.5273069739341736,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.040765900164842606,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 892
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 429.68751525878906,
-      "epoch": 0.893,
-      "grad_norm": 16.8784751889796,
-      "kl": 2.828125,
-      "learning_rate": 1.3102488286186234e-07,
-      "loss": 0.5288,
-      "reward": 2.408605933189392,
-      "reward_std": 0.48363907635211945,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333432674408,
-      "rewards/repetition_penalty_reward": -0.023685818538069725,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 346.25001525878906,
+      "epoch": 0.4465,
+      "grad_norm": 48.76214538177417,
+      "kl": 5.28125,
+      "learning_rate": 7.090636797006657e-07,
+      "loss": 0.5066,
+      "reward": 2.608002543449402,
+      "reward_std": 0.7648839950561523,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9513889849185944,
+      "rewards/repetition_penalty_reward": -0.025678012520074844,
+      "rewards/tag_count_reward": 0.9114583730697632,
       "step": 893
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 437.5416717529297,
-      "epoch": 0.894,
-      "grad_norm": 41.80416425701382,
-      "kl": 3.0625,
-      "learning_rate": 1.3045428945301953e-07,
-      "loss": 0.8179,
-      "reward": 2.665311813354492,
-      "reward_std": 0.4915819466114044,
-      "rewards/accuracy_reward": 0.7916666865348816,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.01177146751433611,
-      "rewards/tag_count_reward": 0.9270833730697632,
+      "completion_length": 369.04168701171875,
+      "epoch": 0.447,
+      "grad_norm": 92.64315624811375,
+      "kl": 5.3515625,
+      "learning_rate": 7.083287417821157e-07,
+      "loss": 0.6629,
+      "reward": 2.784505248069763,
+      "reward_std": 0.33816368877887726,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.01063354080542922,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 894
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 358.5833435058594,
-      "epoch": 0.895,
-      "grad_norm": 47.78557902231892,
-      "kl": 2.76171875,
-      "learning_rate": 1.2988880807625927e-07,
-      "loss": 0.5157,
-      "reward": 2.5233113765716553,
-      "reward_std": 0.5640779733657837,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.025299932807683945,
-      "rewards/tag_count_reward": 0.9375000298023224,
+      "completion_length": 643.3958587646484,
+      "epoch": 0.4475,
+      "grad_norm": 223.85239628295713,
+      "kl": 16.421875,
+      "learning_rate": 7.075933215667604e-07,
+      "loss": 1.5531,
+      "reward": 2.1586782932281494,
+      "reward_std": 0.6240611672401428,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.02014122251421213,
+      "rewards/tag_count_reward": 0.7760416865348816,
       "step": 895
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 535.6666870117188,
-      "epoch": 0.896,
-      "grad_norm": 27.290533216671655,
-      "kl": 4.2578125,
-      "learning_rate": 1.2932844562179352e-07,
-      "loss": 0.6981,
-      "reward": 2.0957794189453125,
-      "reward_std": 0.5598824918270111,
-      "rewards/accuracy_reward": 0.25,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.015331756323575974,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 381.2916717529297,
+      "epoch": 0.448,
+      "grad_norm": 32.38556348235585,
+      "kl": 2.5,
+      "learning_rate": 7.068574212948169e-07,
+      "loss": 0.954,
+      "reward": 2.5282145738601685,
+      "reward_std": 0.39006973803043365,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.023868918418884277,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 896
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 384.68751525878906,
-      "epoch": 0.897,
-      "grad_norm": 8.938743613588342,
-      "kl": 1.4921875,
-      "learning_rate": 1.2877320891746201e-07,
-      "loss": 0.2616,
-      "reward": 2.667781949043274,
-      "reward_std": 0.4907715171575546,
-      "rewards/accuracy_reward": 0.75,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.026662585325539112,
-      "rewards/tag_count_reward": 0.9583333432674408,
+      "completion_length": 445.62501525878906,
+      "epoch": 0.4485,
+      "grad_norm": 29.350505167918783,
+      "kl": 5.484375,
+      "learning_rate": 7.06121043207965e-07,
+      "loss": 1.1386,
+      "reward": 2.497186064720154,
+      "reward_std": 0.7929587364196777,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.027119554579257965,
+      "rewards/tag_count_reward": 0.8645833432674408,
       "step": 897
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 413.6041717529297,
-      "epoch": 0.898,
-      "grad_norm": 44.44288512911277,
-      "kl": 1.8125,
-      "learning_rate": 1.2822310472864885e-07,
-      "loss": 0.7351,
-      "reward": 2.583755135536194,
-      "reward_std": 0.557906985282898,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.01867562346160412,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 615.4166870117188,
+      "epoch": 0.449,
+      "grad_norm": 36.911352628939454,
+      "kl": 6.421875,
+      "learning_rate": 7.053841895493406e-07,
+      "loss": 1.61,
+      "reward": 2.3755075931549072,
+      "reward_std": 0.9200084507465363,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.016853561159223318,
+      "rewards/tag_count_reward": 0.8020833730697632,
       "step": 898
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 339.2708435058594,
-      "epoch": 0.899,
-      "grad_norm": 12.549931479255974,
-      "kl": 2.23046875,
-      "learning_rate": 1.2767813975819983e-07,
-      "loss": 0.3723,
-      "reward": 2.7536635398864746,
-      "reward_std": 0.3352846037596464,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.010225379839539528,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 442.1250305175781,
+      "epoch": 0.4495,
+      "grad_norm": 22.680192929837112,
+      "kl": 4.47265625,
+      "learning_rate": 7.046468625635274e-07,
+      "loss": 0.7625,
+      "reward": 2.4903249740600586,
+      "reward_std": 0.8195319175720215,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.01661947136744857,
+      "rewards/tag_count_reward": 0.875,
       "step": 899
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 467.1250305175781,
-      "epoch": 0.9,
-      "grad_norm": 22.420504928111246,
-      "kl": 3.375,
-      "learning_rate": 1.2713832064634125e-07,
-      "loss": 0.8531,
-      "reward": 2.0020930767059326,
-      "reward_std": 0.47625498473644257,
-      "rewards/accuracy_reward": 0.1875,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.020476454868912697,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 463.6041717529297,
+      "epoch": 0.45,
+      "grad_norm": 27.272842353579144,
+      "kl": 4.59375,
+      "learning_rate": 7.039090644965509e-07,
+      "loss": 0.6818,
+      "reward": 2.319231152534485,
+      "reward_std": 0.888009250164032,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.0227828249335289,
+      "rewards/tag_count_reward": 0.8072916865348816,
       "step": 900
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 603.0833587646484,
-      "epoch": 0.901,
-      "grad_norm": 32.67875227205643,
-      "kl": 5.296875,
-      "learning_rate": 1.2660365397059856e-07,
-      "loss": 1.2488,
-      "reward": 2.2659354209899902,
-      "reward_std": 0.6388919800519943,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02573126833885908,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 449.0833435058594,
+      "epoch": 0.4505,
+      "grad_norm": 34.13999464127485,
+      "kl": 2.9140625,
+      "learning_rate": 7.031707975958726e-07,
+      "loss": 0.7008,
+      "reward": 2.466166138648987,
+      "reward_std": 0.6233952641487122,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.01994516607373953,
+      "rewards/tag_count_reward": 0.8750000298023224,
       "step": 901
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 458.7916717529297,
-      "epoch": 0.902,
-      "grad_norm": 31.91653489837797,
-      "kl": 2.703125,
-      "learning_rate": 1.260741462457165e-07,
-      "loss": 1.0144,
-      "reward": 2.4180402755737305,
-      "reward_std": 0.6183659136295319,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.017723857890814543,
-      "rewards/tag_count_reward": 0.9218750298023224,
+      "completion_length": 477.9166717529297,
+      "epoch": 0.451,
+      "grad_norm": 44.299788634003804,
+      "kl": 3.7109375,
+      "learning_rate": 7.024320641103811e-07,
+      "loss": 0.9245,
+      "reward": 2.344895601272583,
+      "reward_std": 0.6696476340293884,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.01621554931625724,
+      "rewards/tag_count_reward": 0.8541666865348816,
       "step": 902
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 526.8750305175781,
-      "epoch": 0.903,
-      "grad_norm": 29.282295286202803,
-      "kl": 6.421875,
-      "learning_rate": 1.2554980392357956e-07,
-      "loss": 1.0086,
-      "reward": 2.2422985434532166,
-      "reward_std": 0.8072032630443573,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.8888889253139496,
-      "rewards/repetition_penalty_reward": -0.0163820618763566,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 702.1458435058594,
+      "epoch": 0.4515,
+      "grad_norm": 140.07908899869074,
+      "kl": 14.5,
+      "learning_rate": 7.01692866290387e-07,
+      "loss": 1.4671,
+      "reward": 1.8353345394134521,
+      "reward_std": 0.8967846035957336,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.8750000298023224,
+      "rewards/repetition_penalty_reward": -0.013623815728351474,
+      "rewards/tag_count_reward": 0.6197916865348816,
       "step": 903
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 478.37501525878906,
-      "epoch": 0.904,
-      "grad_norm": 33.228944520213055,
-      "kl": 6.609375,
-      "learning_rate": 1.2503063339313356e-07,
-      "loss": 1.1511,
-      "reward": 2.098447561264038,
-      "reward_std": 0.6330851316452026,
-      "rewards/accuracy_reward": 0.3541666716337204,
-      "rewards/reasoning_steps_reward": 0.902777761220932,
-      "rewards/repetition_penalty_reward": -0.012663647066801786,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 422.50001525878906,
+      "epoch": 0.452,
+      "grad_norm": 52.20368268795704,
+      "kl": 5.78125,
+      "learning_rate": 7.009532063876148e-07,
+      "loss": 0.6616,
+      "reward": 2.408238172531128,
+      "reward_std": 0.5504840314388275,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.024053473956882954,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 904
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 458.87501525878906,
-      "epoch": 0.905,
-      "grad_norm": 36.169535665899474,
-      "kl": 4.171875,
-      "learning_rate": 1.2451664098030743e-07,
-      "loss": 0.7981,
-      "reward": 2.4604378938674927,
-      "reward_std": 0.7671001553535461,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9375000596046448,
-      "rewards/repetition_penalty_reward": -0.0291455565020442,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 445.0416717529297,
+      "epoch": 0.4525,
+      "grad_norm": 33.39372260946179,
+      "kl": 4.5234375,
+      "learning_rate": 7.002130866551968e-07,
+      "loss": 0.5914,
+      "reward": 2.167671024799347,
+      "reward_std": 0.30702342092990875,
+      "rewards/accuracy_reward": 0.3333333544433117,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03198174387216568,
+      "rewards/tag_count_reward": 0.8802083432674408,
       "step": 905
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 404.79168701171875,
-      "epoch": 0.906,
-      "grad_norm": 17.708470010127638,
-      "kl": 3.55078125,
-      "learning_rate": 1.2400783294793668e-07,
-      "loss": 0.9617,
-      "reward": 2.5592939853668213,
-      "reward_std": 0.7362368702888489,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.015358910895884037,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 484.9791717529297,
+      "epoch": 0.453,
+      "grad_norm": 25.372165474058104,
+      "kl": 3.1796875,
+      "learning_rate": 6.994725093476664e-07,
+      "loss": 0.5834,
+      "reward": 2.468542218208313,
+      "reward_std": 0.6370213180780411,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.01930511137470603,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 906
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 607.6250305175781,
-      "epoch": 0.907,
-      "grad_norm": 52.06808756849091,
-      "kl": 9.578125,
-      "learning_rate": 1.235042154956865e-07,
-      "loss": 1.4479,
-      "reward": 2.2677348852157593,
-      "reward_std": 0.8661213219165802,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9097222685813904,
-      "rewards/repetition_penalty_reward": -0.03261231165379286,
-      "rewards/tag_count_reward": 0.8072916865348816,
+      "completion_length": 408.97918701171875,
+      "epoch": 0.4535,
+      "grad_norm": 9.77141277310684,
+      "kl": 2.5703125,
+      "learning_rate": 6.987314767209503e-07,
+      "loss": 0.3112,
+      "reward": 2.3773725032806396,
+      "reward_std": 0.49295616149902344,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.021933134645223618,
+      "rewards/tag_count_reward": 0.8645833432674408,
       "step": 907
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 379.8958435058594,
-      "epoch": 0.908,
-      "grad_norm": 14.881992614761662,
-      "kl": 2.5,
-      "learning_rate": 1.2300579475997657e-07,
-      "loss": 0.3101,
-      "reward": 2.3961899280548096,
-      "reward_std": 0.6012940406799316,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.023949017748236656,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "completion_length": 462.1041793823242,
+      "epoch": 0.454,
+      "grad_norm": 24.983838013132367,
+      "kl": 2.546875,
+      "learning_rate": 6.979899910323624e-07,
+      "loss": 0.4343,
+      "reward": 2.223532557487488,
+      "reward_std": 0.6077858209609985,
+      "rewards/accuracy_reward": 0.45833333395421505,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.019523008493706584,
+      "rewards/tag_count_reward": 0.8541666865348816,
       "step": 908
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 573.8541870117188,
-      "epoch": 0.909,
-      "grad_norm": 96.3349004340528,
-      "kl": 8.25,
-      "learning_rate": 1.2251257681390645e-07,
-      "loss": 1.2975,
-      "reward": 2.3617441654205322,
-      "reward_std": 0.7564932107925415,
-      "rewards/accuracy_reward": 0.6041666716337204,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.01151987537741661,
-      "rewards/tag_count_reward": 0.8177083432674408,
+      "completion_length": 304.0416717529297,
+      "epoch": 0.4545,
+      "grad_norm": 28.981123137793116,
+      "kl": 1.62109375,
+      "learning_rate": 6.972480545405968e-07,
+      "loss": 0.3188,
+      "reward": 2.5748839378356934,
+      "reward_std": 0.47933661937713623,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.013657951261848211,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 909
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 523.9791717529297,
-      "epoch": 0.91,
-      "grad_norm": 36.959308180317855,
-      "kl": 6.234375,
-      "learning_rate": 1.220245676671809e-07,
-      "loss": 1.2606,
-      "reward": 2.20367431640625,
-      "reward_std": 0.9045920073986053,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9097222685813904,
-      "rewards/repetition_penalty_reward": -0.039381190203130245,
-      "rewards/tag_count_reward": 0.8125000298023224,
+      "completion_length": 440.97918701171875,
+      "epoch": 0.455,
+      "grad_norm": 25.272268744236346,
+      "kl": 5.78125,
+      "learning_rate": 6.965056695057204e-07,
+      "loss": 1.0701,
+      "reward": 2.4917463064193726,
+      "reward_std": 0.769326239824295,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9513888657093048,
+      "rewards/repetition_penalty_reward": -0.04297593608498573,
+      "rewards/tag_count_reward": 0.875,
       "step": 910
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 556.875,
-      "epoch": 0.911,
-      "grad_norm": 16.456009421461072,
-      "kl": 5.59375,
-      "learning_rate": 1.2154177326603763e-07,
-      "loss": 1.3297,
-      "reward": 2.312328815460205,
-      "reward_std": 0.7721385657787323,
-      "rewards/accuracy_reward": 0.5625000149011612,
-      "rewards/reasoning_steps_reward": 0.9236111044883728,
-      "rewards/repetition_penalty_reward": -0.022740643471479416,
-      "rewards/tag_count_reward": 0.8489583432674408,
+      "completion_length": 287.12501525878906,
+      "epoch": 0.4555,
+      "grad_norm": 30.011200731000926,
+      "kl": 2.720703125,
+      "learning_rate": 6.957628381891673e-07,
+      "loss": 0.3289,
+      "reward": 2.6884909868240356,
+      "reward_std": 0.36660417914390564,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.014634124469012022,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 911
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 409.29168701171875,
-      "epoch": 0.912,
-      "grad_norm": 18.335739930280326,
-      "kl": 2.6767578125,
-      "learning_rate": 1.2106419949317388e-07,
-      "loss": 0.6476,
-      "reward": 2.470385789871216,
-      "reward_std": 0.5638497471809387,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.01746157230809331,
-      "rewards/tag_count_reward": 0.953125,
+      "completion_length": 328.75,
+      "epoch": 0.456,
+      "grad_norm": 26.395423136562222,
+      "kl": 3.484375,
+      "learning_rate": 6.950195628537299e-07,
+      "loss": 0.7263,
+      "reward": 2.7875736951828003,
+      "reward_std": 0.4841386526823044,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.02319031674414873,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 912
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 632.8958435058594,
-      "epoch": 0.913,
-      "grad_norm": 23.440878703155796,
-      "kl": 6.265625,
-      "learning_rate": 1.2059185216767543e-07,
-      "loss": 1.0253,
-      "reward": 2.0628278851509094,
-      "reward_std": 0.9162045121192932,
-      "rewards/accuracy_reward": 0.4166666716337204,
-      "rewards/reasoning_steps_reward": 0.9027778506278992,
-      "rewards/repetition_penalty_reward": -0.04307497665286064,
-      "rewards/tag_count_reward": 0.7864583432674408,
+      "completion_length": 379.5625,
+      "epoch": 0.4565,
+      "grad_norm": 81.85144356177,
+      "kl": 6.125,
+      "learning_rate": 6.942758457635543e-07,
+      "loss": 0.9552,
+      "reward": 2.5703320503234863,
+      "reward_std": 0.3624073415994644,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222983837128,
+      "rewards/repetition_penalty_reward": -0.03209859039634466,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 913
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 622.4583435058594,
-      "epoch": 0.914,
-      "grad_norm": 42.16437419261308,
-      "kl": 6.59375,
-      "learning_rate": 1.2012473704494537e-07,
-      "loss": 1.123,
-      "reward": 1.9104456305503845,
-      "reward_std": 0.520677238702774,
-      "rewards/accuracy_reward": 0.18750000558793545,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.020110088400542736,
-      "rewards/tag_count_reward": 0.8125000298023224,
+      "completion_length": 364.25001525878906,
+      "epoch": 0.457,
+      "grad_norm": 45.15015133012363,
+      "kl": 5.046875,
+      "learning_rate": 6.935316891841315e-07,
+      "loss": 0.4927,
+      "reward": 2.6616228818893433,
+      "reward_std": 0.5104624330997467,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.018932746723294258,
+      "rewards/tag_count_reward": 0.9375,
       "step": 914
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 379.04168701171875,
-      "epoch": 0.915,
-      "grad_norm": 108.02105225090547,
-      "kl": 3.39453125,
-      "learning_rate": 1.1966285981663407e-07,
-      "loss": 0.3631,
-      "reward": 2.5404409170150757,
-      "reward_std": 0.3071303367614746,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.020322874188423157,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 392.0416717529297,
+      "epoch": 0.4575,
+      "grad_norm": 37.202986072549116,
+      "kl": 2.38671875,
+      "learning_rate": 6.927870953822915e-07,
+      "loss": 0.5714,
+      "reward": 2.446205258369446,
+      "reward_std": 0.593352735042572,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.027753093279898167,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 915
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 573.6041870117188,
-      "epoch": 0.916,
-      "grad_norm": 30.21063587734105,
-      "kl": 5.234375,
-      "learning_rate": 1.1920622611056974e-07,
-      "loss": 0.9573,
-      "reward": 1.9745216369628906,
-      "reward_std": 0.8713274002075195,
-      "rewards/accuracy_reward": 0.291666679084301,
-      "rewards/reasoning_steps_reward": 0.881944477558136,
-      "rewards/repetition_penalty_reward": -0.042839540168643,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 322.35418701171875,
+      "epoch": 0.458,
+      "grad_norm": 24.96453853888965,
+      "kl": 0.90234375,
+      "learning_rate": 6.920420666261961e-07,
+      "loss": 0.2755,
+      "reward": 2.790469765663147,
+      "reward_std": 0.28246453404426575,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.027238698676228523,
+      "rewards/tag_count_reward": 0.984375,
       "step": 916
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 568.0833587646484,
-      "epoch": 0.917,
-      "grad_norm": 43.80723570465828,
-      "kl": 4.2109375,
-      "learning_rate": 1.1875484149069004e-07,
-      "loss": 1.1291,
-      "reward": 2.45370614528656,
-      "reward_std": 0.8172959387302399,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.020252287853509188,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 345.3958435058594,
+      "epoch": 0.4585,
+      "grad_norm": 49.466728379593775,
+      "kl": 0.9541015625,
+      "learning_rate": 6.912966051853322e-07,
+      "loss": 0.5167,
+      "reward": 2.606629729270935,
+      "reward_std": 0.5284582823514938,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.025314772501587868,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 917
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 400.7083435058594,
-      "epoch": 0.918,
-      "grad_norm": 20.756190324375613,
-      "kl": 1.0966796875,
-      "learning_rate": 1.1830871145697412e-07,
-      "loss": 0.2955,
-      "reward": 2.608530044555664,
-      "reward_std": 0.4455796778202057,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.028622763231396675,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 385.18751525878906,
+      "epoch": 0.459,
+      "grad_norm": 8.401580563295221,
+      "kl": 2.28662109375,
+      "learning_rate": 6.905507133305047e-07,
+      "loss": 0.3083,
+      "reward": 2.5714030265808105,
+      "reward_std": 0.39095499366521835,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.017138768918812275,
+      "rewards/tag_count_reward": 0.921875,
       "step": 918
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 432.9791717529297,
-      "epoch": 0.919,
-      "grad_norm": 11.62915536443853,
-      "kl": 2.140625,
-      "learning_rate": 1.1786784144537563e-07,
-      "loss": 0.4186,
-      "reward": 2.5469974279403687,
-      "reward_std": 0.4763101041316986,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.015502599067986012,
-      "rewards/tag_count_reward": 0.9375000298023224,
+      "completion_length": 364.4791717529297,
+      "epoch": 0.4595,
+      "grad_norm": 30.373353543333543,
+      "kl": 1.8203125,
+      "learning_rate": 6.898043933338293e-07,
+      "loss": 0.648,
+      "reward": 2.423296332359314,
+      "reward_std": 0.4785070866346359,
+      "rewards/accuracy_reward": 0.5416666716337204,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.04892597161233425,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 919
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 507.95835876464844,
-      "epoch": 0.92,
-      "grad_norm": 15.992892533041532,
-      "kl": 3.59375,
-      "learning_rate": 1.1743223682775649e-07,
-      "loss": 0.7643,
-      "reward": 2.407273054122925,
-      "reward_std": 0.6685203611850739,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.019810385070741177,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 337.2708435058594,
+      "epoch": 0.46,
+      "grad_norm": 16.49521792000666,
+      "kl": 1.76171875,
+      "learning_rate": 6.890576474687263e-07,
+      "loss": 0.2062,
+      "reward": 2.577837347984314,
+      "reward_std": 0.36855159886181355,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.03153769485652447,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 920
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 349.81251525878906,
-      "epoch": 0.921,
-      "grad_norm": 35.49834021420722,
-      "kl": 2.08203125,
-      "learning_rate": 1.1700190291182158e-07,
-      "loss": 0.499,
-      "reward": 2.5659878253936768,
-      "reward_std": 0.5623100101947784,
-      "rewards/accuracy_reward": 0.6666666716337204,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.00866501871496439,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 345.62501525878906,
+      "epoch": 0.4605,
+      "grad_norm": 15.297924858007171,
+      "kl": 1.8203125,
+      "learning_rate": 6.883104780099133e-07,
+      "loss": 0.4431,
+      "reward": 2.5769479274749756,
+      "reward_std": 0.4878064692020416,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.046316033229231834,
+      "rewards/tag_count_reward": 0.984375,
       "step": 921
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 441.7083435058594,
-      "epoch": 0.922,
-      "grad_norm": 14.606033932622575,
-      "kl": 2.08203125,
-      "learning_rate": 1.1657684494105386e-07,
-      "loss": 0.4224,
-      "reward": 2.6236190795898438,
-      "reward_std": 0.5365782380104065,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.01526999520137906,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 318.3958435058594,
+      "epoch": 0.461,
+      "grad_norm": 17.663934976152945,
+      "kl": 1.71484375,
+      "learning_rate": 6.875628872333975e-07,
+      "loss": 0.4003,
+      "reward": 2.721220016479492,
+      "reward_std": 0.3829363286495209,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02878008782863617,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 922
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 544.25,
-      "epoch": 0.923,
-      "grad_norm": 32.63063888698561,
-      "kl": 5.8515625,
-      "learning_rate": 1.1615706809465051e-07,
-      "loss": 0.8629,
-      "reward": 2.486325263977051,
-      "reward_std": 0.6139254868030548,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.02756360173225403,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 388.12501525878906,
+      "epoch": 0.4615,
+      "grad_norm": 24.847963074766394,
+      "kl": 2.52880859375,
+      "learning_rate": 6.868148774164706e-07,
+      "loss": 0.8003,
+      "reward": 2.7672914266586304,
+      "reward_std": 0.40449826419353485,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02958354353904724,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 923
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 347.3125,
-      "epoch": 0.924,
-      "grad_norm": 12.704180107443388,
-      "kl": 1.7890625,
-      "learning_rate": 1.1574257748745986e-07,
-      "loss": 0.2841,
-      "reward": 2.5728049278259277,
-      "reward_std": 0.43000659346580505,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.024417342618107796,
-      "rewards/tag_count_reward": 0.9583333432674408,
+      "completion_length": 332.5208435058594,
+      "epoch": 0.462,
+      "grad_norm": 50.169714777079136,
+      "kl": 2.1552734375,
+      "learning_rate": 6.860664508377001e-07,
+      "loss": 0.368,
+      "reward": 2.5989573001861572,
+      "reward_std": 0.36564599722623825,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.026042713783681393,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 924
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 413.12501525878906,
-      "epoch": 0.925,
-      "grad_norm": 18.841011447480344,
-      "kl": 3.2578125,
-      "learning_rate": 1.1533337816991931e-07,
-      "loss": 0.7064,
-      "reward": 2.5212587118148804,
-      "reward_std": 0.42140892148017883,
-      "rewards/accuracy_reward": 0.6666666716337204,
-      "rewards/reasoning_steps_reward": 0.92361119389534,
-      "rewards/repetition_penalty_reward": -0.011727516539394855,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 390.3333435058594,
+      "epoch": 0.4625,
+      "grad_norm": 128.99931158793126,
+      "kl": 6.46875,
+      "learning_rate": 6.853176097769228e-07,
+      "loss": 1.1684,
+      "reward": 2.682253360748291,
+      "reward_std": 0.5880621820688248,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.03302454110234976,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 925
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 442.25,
-      "epoch": 0.926,
-      "grad_norm": 16.753619228197998,
-      "kl": 3.5546875,
-      "learning_rate": 1.1492947512799328e-07,
-      "loss": 0.6144,
-      "reward": 2.417970061302185,
-      "reward_std": 0.5506798624992371,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.02300213649868965,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "completion_length": 421.3958435058594,
+      "epoch": 0.463,
+      "grad_norm": 43.308678626367616,
+      "kl": 3.63671875,
+      "learning_rate": 6.84568356515239e-07,
+      "loss": 1.3087,
+      "reward": 2.6976706981658936,
+      "reward_std": 0.5417228192090988,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9305556416511536,
+      "rewards/repetition_penalty_reward": -0.02455159369856119,
+      "rewards/tag_count_reward": 0.9375,
       "step": 926
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 532.6250305175781,
-      "epoch": 0.927,
-      "grad_norm": 47.35016742807312,
-      "kl": 7.84375,
-      "learning_rate": 1.1453087328311299e-07,
-      "loss": 1.1422,
-      "reward": 2.0104929208755493,
-      "reward_std": 0.6963983774185181,
-      "rewards/accuracy_reward": 0.29166667722165585,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.02422943152487278,
-      "rewards/tag_count_reward": 0.8125000298023224,
+      "completion_length": 364.6875,
+      "epoch": 0.4635,
+      "grad_norm": 19.355624028175274,
+      "kl": 1.4140625,
+      "learning_rate": 6.838186933350036e-07,
+      "loss": 0.5342,
+      "reward": 2.708867311477661,
+      "reward_std": 0.4020465463399887,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.03245212323963642,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 927
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 385.4791717529297,
-      "epoch": 0.928,
-      "grad_norm": 15.095495398152355,
-      "kl": 2.072265625,
-      "learning_rate": 1.1413757749211602e-07,
-      "loss": 0.4393,
-      "reward": 2.454028844833374,
-      "reward_std": 0.44756019115448,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.01645728573203087,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 328.81251525878906,
+      "epoch": 0.464,
+      "grad_norm": 26.038208175574418,
+      "kl": 1.3828125,
+      "learning_rate": 6.83068622519821e-07,
+      "loss": 0.4745,
+      "reward": 2.8109928369522095,
+      "reward_std": 0.32376736029982567,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.017132141161710024,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 928
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 614.6041870117188,
-      "epoch": 0.929,
-      "grad_norm": 31.056679659798007,
-      "kl": 7.28125,
-      "learning_rate": 1.137495925471875e-07,
-      "loss": 1.1384,
-      "reward": 2.323223352432251,
-      "reward_std": 0.7905822396278381,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.958333432674408,
-      "rewards/repetition_penalty_reward": -0.015318467281758785,
-      "rewards/tag_count_reward": 0.8385416865348816,
+      "completion_length": 433.81251525878906,
+      "epoch": 0.4645,
+      "grad_norm": 31.19447558981725,
+      "kl": 4.1796875,
+      "learning_rate": 6.823181463545366e-07,
+      "loss": 0.8102,
+      "reward": 2.39223575592041,
+      "reward_std": 0.6007313132286072,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9374999701976776,
+      "rewards/repetition_penalty_reward": -0.04526432417333126,
+      "rewards/tag_count_reward": 0.8958333730697632,
       "step": 929
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 451.43751525878906,
-      "epoch": 0.93,
-      "grad_norm": 24.05252138370831,
-      "kl": 3.65625,
-      "learning_rate": 1.1336692317580158e-07,
-      "loss": 0.8455,
-      "reward": 2.514081120491028,
-      "reward_std": 0.6175527423620224,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.0275856489315629,
-      "rewards/tag_count_reward": 0.9375000298023224,
+      "completion_length": 354.54168701171875,
+      "epoch": 0.465,
+      "grad_norm": 16.112655951237382,
+      "kl": 1.796875,
+      "learning_rate": 6.815672671252315e-07,
+      "loss": 0.5261,
+      "reward": 2.3884493112564087,
+      "reward_std": 0.416194885969162,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.026481250301003456,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 930
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 414.43751525878906,
-      "epoch": 0.931,
-      "grad_norm": 27.33191623105398,
-      "kl": 2.1875,
-      "learning_rate": 1.1298957404066381e-07,
-      "loss": 0.3232,
-      "reward": 2.5895215272903442,
-      "reward_std": 0.4988251328468323,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.02506194543093443,
-      "rewards/tag_count_reward": 0.9479166865348816,
+      "completion_length": 308.68751525878906,
+      "epoch": 0.4655,
+      "grad_norm": 39.327147074285115,
+      "kl": 1.4365234375,
+      "learning_rate": 6.808159871192136e-07,
+      "loss": 0.3897,
+      "reward": 2.633025646209717,
+      "reward_std": 0.1877675480209291,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.04579382762312889,
+      "rewards/tag_count_reward": 0.984375,
       "step": 931
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 502.9583435058594,
-      "epoch": 0.932,
-      "grad_norm": 30.75771927110975,
-      "kl": 4.39453125,
-      "learning_rate": 1.1261754973965422e-07,
-      "loss": 0.7145,
-      "reward": 2.1934012174606323,
-      "reward_std": 0.5984681844711304,
-      "rewards/accuracy_reward": 0.3750000149011612,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.02534892037510872,
-      "rewards/tag_count_reward": 0.8645833432674408,
+      "completion_length": 435.9583435058594,
+      "epoch": 0.466,
+      "grad_norm": 22.43891344147886,
+      "kl": 3.75,
+      "learning_rate": 6.800643086250121e-07,
+      "loss": 1.0874,
+      "reward": 2.6055747270584106,
+      "reward_std": 0.6811087727546692,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.026369919069111347,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 932
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 458.54168701171875,
-      "epoch": 0.933,
-      "grad_norm": 26.112800067792758,
-      "kl": 4.203125,
-      "learning_rate": 1.1225085480577158e-07,
-      "loss": 0.8496,
-      "reward": 2.2411762475967407,
-      "reward_std": 0.7032366394996643,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.9444445371627808,
-      "rewards/repetition_penalty_reward": -0.01576836034655571,
-      "rewards/tag_count_reward": 0.8958333730697632,
+      "completion_length": 489.97918701171875,
+      "epoch": 0.4665,
+      "grad_norm": 105.78114916915985,
+      "kl": 7.40625,
+      "learning_rate": 6.793122339323705e-07,
+      "loss": 0.9496,
+      "reward": 2.352501630783081,
+      "reward_std": 0.6076760590076447,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.020762240514159203,
+      "rewards/tag_count_reward": 0.8802083730697632,
       "step": 933
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 424.37501525878906,
-      "epoch": 0.934,
-      "grad_norm": 44.09196385541136,
-      "kl": 2.703125,
-      "learning_rate": 1.1188949370707787e-07,
-      "loss": 0.9393,
-      "reward": 2.4043623208999634,
-      "reward_std": 0.672376275062561,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.010568303987383842,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 440.7708435058594,
+      "epoch": 0.467,
+      "grad_norm": 28.045265825606887,
+      "kl": 2.93017578125,
+      "learning_rate": 6.78559765332238e-07,
+      "loss": 0.4259,
+      "reward": 2.5912506580352783,
+      "reward_std": 0.45991945266723633,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.02159656397998333,
+      "rewards/tag_count_reward": 0.9114583432674408,
       "step": 934
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 421.3958435058594,
-      "epoch": 0.935,
-      "grad_norm": 19.863825505537683,
-      "kl": 2.9765625,
-      "learning_rate": 1.1153347084664419e-07,
-      "loss": 0.5384,
-      "reward": 2.456182837486267,
-      "reward_std": 0.5422267615795135,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.021247888915240765,
-      "rewards/tag_count_reward": 0.9427083432674408,
+      "completion_length": 435.00001525878906,
+      "epoch": 0.4675,
+      "grad_norm": 41.65994882506967,
+      "kl": 1.521484375,
+      "learning_rate": 6.778069051167653e-07,
+      "loss": 0.62,
+      "reward": 2.5147244930267334,
+      "reward_std": 0.556397408246994,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.0338868722319603,
+      "rewards/tag_count_reward": 0.9375,
       "step": 935
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 487.3541717529297,
-      "epoch": 0.936,
-      "grad_norm": 49.22541738740071,
-      "kl": 3.1875,
-      "learning_rate": 1.1118279056249653e-07,
-      "loss": 1.1499,
-      "reward": 2.2859641313552856,
-      "reward_std": 0.4807792901992798,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.01785538112744689,
-      "rewards/tag_count_reward": 0.9010416865348816,
-      "step": 936
-    },
-    {
-      "clip_ratio": 0.0,
-      "completion_length": 476.3958435058594,
-      "epoch": 0.937,
-      "grad_norm": 50.529982193331755,
-      "kl": 3.078125,
-      "learning_rate": 1.1083745712756364e-07,
-      "loss": 0.9369,
-      "reward": 2.6942487955093384,
-      "reward_std": 0.6124294996261597,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.019293093122541904,
-      "rewards/tag_count_reward": 0.9218750298023224,
-      "step": 937
+      "completion_length": 691.4166870117188,
+      "epoch": 0.468,
+      "grad_norm": 37.144816051148496,
+      "kl": 4.2109375,
+      "learning_rate": 6.770536555792944e-07,
+      "loss": 0.8037,
+      "reward": 2.3306804895401,
+      "reward_std": 0.7895675301551819,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.02001405507326126,
+      "rewards/tag_count_reward": 0.8229166865348816,
+      "step": 936
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 583.0625,
-      "epoch": 0.938,
-      "grad_norm": 23.005166702704713,
-      "kl": 5.2421875,
-      "learning_rate": 1.1049747474962444e-07,
-      "loss": 0.8491,
-      "reward": 2.30727219581604,
-      "reward_std": 0.6647425889968872,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778506278992,
-      "rewards/repetition_penalty_reward": -0.022589108906686306,
+      "completion_length": 526.0625305175781,
+      "epoch": 0.4685,
+      "grad_norm": 48.4930120879613,
+      "kl": 2.5390625,
+      "learning_rate": 6.763000190143545e-07,
+      "loss": 0.7913,
+      "reward": 2.2242666482925415,
+      "reward_std": 0.5582642555236816,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.036150138825178146,
       "rewards/tag_count_reward": 0.8854166865348816,
-      "step": 938
+      "step": 937
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 426.4583435058594,
-      "epoch": 0.939,
-      "grad_norm": 27.108706168312672,
-      "kl": 3.7734375,
-      "learning_rate": 1.1016284757125685e-07,
-      "loss": 0.644,
-      "reward": 2.582595705986023,
-      "reward_std": 0.6667671203613281,
+      "completion_length": 472.9583435058594,
+      "epoch": 0.469,
+      "grad_norm": 28.377979881418177,
+      "kl": 2.875,
+      "learning_rate": 6.755459977176532e-07,
+      "loss": 0.7778,
+      "reward": 2.534548759460449,
+      "reward_std": 0.6151235550642014,
       "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.023307234048843384,
-      "rewards/tag_count_reward": 0.9114583432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.02100683329626918,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 938
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 506.08335876464844,
+      "epoch": 0.4695,
+      "grad_norm": 51.22661172315456,
+      "kl": 5.640625,
+      "learning_rate": 6.747915939860701e-07,
+      "loss": 1.0882,
+      "reward": 2.568353295326233,
+      "reward_std": 0.7697390615940094,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.028869030997157097,
+      "rewards/tag_count_reward": 0.8750000298023224,
       "step": 939
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 606.9791870117188,
-      "epoch": 0.94,
-      "grad_norm": 35.88000566575318,
-      "kl": 7.203125,
-      "learning_rate": 1.0983357966978745e-07,
-      "loss": 1.2855,
-      "reward": 2.064675807952881,
-      "reward_std": 0.6436410248279572,
-      "rewards/accuracy_reward": 0.3125000149011612,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.02039373479783535,
-      "rewards/tag_count_reward": 0.828125,
+      "completion_length": 490.87501525878906,
+      "epoch": 0.47,
+      "grad_norm": 90.34107970778037,
+      "kl": 5.359375,
+      "learning_rate": 6.740368101176495e-07,
+      "loss": 1.3803,
+      "reward": 2.270451307296753,
+      "reward_std": 0.7119334787130356,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.03510438837110996,
+      "rewards/tag_count_reward": 0.8958333432674408,
       "step": 940
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 449.1041717529297,
-      "epoch": 0.941,
-      "grad_norm": 26.56932442747977,
-      "kl": 2.328125,
-      "learning_rate": 1.0950967505724175e-07,
-      "loss": 0.7233,
-      "reward": 2.446874976158142,
-      "reward_std": 0.4883129894733429,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.03229183237999678,
-      "rewards/tag_count_reward": 0.9583333730697632,
+      "completion_length": 668.5416870117188,
+      "epoch": 0.4705,
+      "grad_norm": 149.87469687137508,
+      "kl": 9.921875,
+      "learning_rate": 6.732816484115946e-07,
+      "loss": 1.7406,
+      "reward": 2.2741907835006714,
+      "reward_std": 0.8710194230079651,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.8888888955116272,
+      "rewards/repetition_penalty_reward": -0.03136484418064356,
+      "rewards/tag_count_reward": 0.8125,
       "step": 941
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 334.3541717529297,
-      "epoch": 0.942,
-      "grad_norm": 10.161335311358066,
-      "kl": 1.244140625,
-      "learning_rate": 1.0919113768029517e-07,
-      "loss": 0.0637,
-      "reward": 2.7936389446258545,
-      "reward_std": 0.438838854432106,
-      "rewards/accuracy_reward": 0.875,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.02580568566918373,
-      "rewards/tag_count_reward": 0.9791666865348816,
+      "completion_length": 346.2291717529297,
+      "epoch": 0.471,
+      "grad_norm": 28.86324528704515,
+      "kl": 3.076171875,
+      "learning_rate": 6.725261111682584e-07,
+      "loss": 0.739,
+      "reward": 2.4970829486846924,
+      "reward_std": 0.5250428169965744,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.01506982184946537,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 942
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 391.22918701171875,
-      "epoch": 0.943,
-      "grad_norm": 16.415006349713707,
-      "kl": 2.619140625,
-      "learning_rate": 1.0887797142022521e-07,
-      "loss": 0.2111,
-      "reward": 2.3882477283477783,
-      "reward_std": 0.3951050788164139,
-      "rewards/accuracy_reward": 0.458333358168602,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.018002384342253208,
+      "completion_length": 487.75001525878906,
+      "epoch": 0.4715,
+      "grad_norm": 56.50232838741636,
+      "kl": 3.095703125,
+      "learning_rate": 6.717702006891386e-07,
+      "loss": 0.723,
+      "reward": 2.7474167346954346,
+      "reward_std": 0.4104699492454529,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.02688878308981657,
       "rewards/tag_count_reward": 0.9479166865348816,
       "step": 943
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 540.5208435058594,
-      "epoch": 0.944,
-      "grad_norm": 30.14237606691049,
-      "kl": 5.453125,
-      "learning_rate": 1.0857018009286381e-07,
-      "loss": 1.0364,
-      "reward": 2.2670759558677673,
-      "reward_std": 0.7407488822937012,
-      "rewards/accuracy_reward": 0.4791666716337204,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.022854735143482685,
-      "rewards/tag_count_reward": 0.8385416865348816,
+      "completion_length": 494.35418701171875,
+      "epoch": 0.472,
+      "grad_norm": 29.170486989553982,
+      "kl": 2.484375,
+      "learning_rate": 6.710139192768694e-07,
+      "loss": 0.8221,
+      "reward": 2.548454523086548,
+      "reward_std": 0.6629728078842163,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.875,
+      "rewards/repetition_penalty_reward": -0.02446222584694624,
+      "rewards/tag_count_reward": 0.8854166865348816,
       "step": 944
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 409.10418701171875,
-      "epoch": 0.945,
-      "grad_norm": 28.58154188768852,
-      "kl": 3.96484375,
-      "learning_rate": 1.0826776744855121e-07,
-      "loss": 0.5507,
-      "reward": 2.29867160320282,
-      "reward_std": 0.5912670195102692,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.01556460838764906,
+      "completion_length": 438.97918701171875,
+      "epoch": 0.4725,
+      "grad_norm": 32.67810625226338,
+      "kl": 1.666015625,
+      "learning_rate": 6.702572692352155e-07,
+      "loss": 0.43,
+      "reward": 2.627410054206848,
+      "reward_std": 0.5064197182655334,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.027104116044938564,
       "rewards/tag_count_reward": 0.9322916865348816,
       "step": 945
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 407.93751525878906,
-      "epoch": 0.946,
-      "grad_norm": 19.249242221629466,
-      "kl": 2.23828125,
-      "learning_rate": 1.0797073717209013e-07,
-      "loss": 0.555,
-      "reward": 2.5011874437332153,
-      "reward_std": 0.40425705909729004,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9444444179534912,
-      "rewards/repetition_penalty_reward": -0.02659040503203869,
-      "rewards/tag_count_reward": 0.9583333432674408,
+      "completion_length": 395.1458435058594,
+      "epoch": 0.473,
+      "grad_norm": 17.27648804391711,
+      "kl": 1.1796875,
+      "learning_rate": 6.695002528690639e-07,
+      "loss": 0.4697,
+      "reward": 2.5221011638641357,
+      "reward_std": 0.5778738856315613,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.024773948825895786,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 946
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 505.66668701171875,
-      "epoch": 0.947,
-      "grad_norm": 73.25982352915128,
-      "kl": 6.1875,
-      "learning_rate": 1.0767909288270063e-07,
-      "loss": 1.2229,
-      "reward": 2.4074935913085938,
-      "reward_std": 0.7739560008049011,
-      "rewards/accuracy_reward": 0.6041666716337204,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.024798161815851927,
-      "rewards/tag_count_reward": 0.8906250298023224,
+      "completion_length": 393.2708435058594,
+      "epoch": 0.4735,
+      "grad_norm": 29.4628975430113,
+      "kl": 1.16015625,
+      "learning_rate": 6.687428724844179e-07,
+      "loss": 0.7055,
+      "reward": 2.7500462532043457,
+      "reward_std": 0.556076854467392,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.02078736387193203,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 947
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 349.8333435058594,
-      "epoch": 0.948,
-      "grad_norm": 24.365433217819515,
-      "kl": 2.26953125,
-      "learning_rate": 1.0739283813397639e-07,
-      "loss": 0.1764,
-      "reward": 2.4521753787994385,
-      "reward_std": 0.42572128772735596,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.025255252607166767,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 440.75,
+      "epoch": 0.474,
+      "grad_norm": 20.698807383719775,
+      "kl": 2.576171875,
+      "learning_rate": 6.679851303883891e-07,
+      "loss": 0.4381,
+      "reward": 2.62761914730072,
+      "reward_std": 0.5112544745206833,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.05293664522469044,
+      "rewards/tag_count_reward": 0.8958333730697632,
       "step": 948
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 552.0625,
-      "epoch": 0.949,
-      "grad_norm": 46.55689685946224,
-      "kl": 7.375,
-      "learning_rate": 1.0711197641384115e-07,
-      "loss": 1.0546,
-      "reward": 2.4559671878814697,
-      "reward_std": 0.6184609234333038,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9375000298023224,
-      "rewards/repetition_penalty_reward": -0.012783022597432137,
-      "rewards/tag_count_reward": 0.8437500298023224,
+      "completion_length": 693.7500305175781,
+      "epoch": 0.4745,
+      "grad_norm": 96.72626140777622,
+      "kl": 5.59375,
+      "learning_rate": 6.672270288891918e-07,
+      "loss": 1.1839,
+      "reward": 2.393503427505493,
+      "reward_std": 0.6808468699455261,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.847222238779068,
+      "rewards/repetition_penalty_reward": -0.01621889090165496,
+      "rewards/tag_count_reward": 0.8541666865348816,
       "step": 949
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 614.4166870117188,
-      "epoch": 0.95,
-      "grad_norm": 35.5385607782762,
-      "kl": 9.46875,
-      "learning_rate": 1.068365111445064e-07,
-      "loss": 1.4734,
-      "reward": 2.2201212644577026,
-      "reward_std": 0.7953130602836609,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9375000298023224,
-      "rewards/repetition_penalty_reward": -0.03508709650486708,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 373.5208435058594,
+      "epoch": 0.475,
+      "grad_norm": 19.558125629971823,
+      "kl": 1.59765625,
+      "learning_rate": 6.664685702961344e-07,
+      "loss": 0.5762,
+      "reward": 2.7208783626556396,
+      "reward_std": 0.4175289124250412,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03780231811106205,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 950
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 607.0416870117188,
-      "epoch": 0.951,
-      "grad_norm": 44.814962262317266,
-      "kl": 10.71875,
-      "learning_rate": 1.0656644568242946e-07,
-      "loss": 1.4213,
-      "reward": 2.336472511291504,
-      "reward_std": 0.7723419368267059,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9513888955116272,
-      "rewards/repetition_penalty_reward": -0.01595810428261757,
-      "rewards/tag_count_reward": 0.7760416865348816,
+      "completion_length": 386.3541717529297,
+      "epoch": 0.4755,
+      "grad_norm": 10.796574951729502,
+      "kl": 1.2109375,
+      "learning_rate": 6.657097569196133e-07,
+      "loss": 0.6176,
+      "reward": 2.8126251697540283,
+      "reward_std": 0.4413589537143707,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03633330576121807,
+      "rewards/tag_count_reward": 0.953125,
       "step": 951
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 468.31251525878906,
-      "epoch": 0.952,
-      "grad_norm": 25.864907881139306,
-      "kl": 3.4375,
-      "learning_rate": 1.063017833182728e-07,
-      "loss": 1.2212,
-      "reward": 2.3151695728302,
-      "reward_std": 0.5797133445739746,
-      "rewards/accuracy_reward": 0.4583333432674408,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.014691710472106934,
-      "rewards/tag_count_reward": 0.9062500298023224,
+      "completion_length": 511.18751525878906,
+      "epoch": 0.476,
+      "grad_norm": 19.892775367759974,
+      "kl": 1.998046875,
+      "learning_rate": 6.649505910711058e-07,
+      "loss": 0.8213,
+      "reward": 2.4712870121002197,
+      "reward_std": 0.6184780597686768,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9722221791744232,
+      "rewards/repetition_penalty_reward": -0.04781022481620312,
+      "rewards/tag_count_reward": 0.9218750298023224,
       "step": 952
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 622.8125305175781,
-      "epoch": 0.953,
-      "grad_norm": 38.335448681356254,
-      "kl": 9.015625,
-      "learning_rate": 1.0604252727686379e-07,
-      "loss": 1.011,
-      "reward": 2.2985631823539734,
-      "reward_std": 0.8530721068382263,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.024353576824069023,
-      "rewards/tag_count_reward": 0.8020833432674408,
+      "completion_length": 392.4791717529297,
+      "epoch": 0.4765,
+      "grad_norm": 4.848138327898023,
+      "kl": 0.654296875,
+      "learning_rate": 6.641910750631626e-07,
+      "loss": 0.4091,
+      "reward": 2.516338586807251,
+      "reward_std": 0.4349832981824875,
+      "rewards/accuracy_reward": 0.5833333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02011991571635008,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 953
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 426.62501525878906,
-      "epoch": 0.954,
-      "grad_norm": 44.054481520111416,
-      "kl": 3.671875,
-      "learning_rate": 1.0578868071715544e-07,
-      "loss": 1.0606,
-      "reward": 2.355110287666321,
-      "reward_std": 0.5248502939939499,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.019889703020453453,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 345.2291717529297,
+      "epoch": 0.477,
+      "grad_norm": 7.336625031482144,
+      "kl": 0.3828125,
+      "learning_rate": 6.634312112094013e-07,
+      "loss": 0.2705,
+      "reward": 2.9284573793411255,
+      "reward_std": 0.1592683894559741,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.0281402375549078,
+      "rewards/tag_count_reward": 0.984375,
       "step": 954
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 394.50001525878906,
-      "epoch": 0.955,
-      "grad_norm": 17.701509278594894,
-      "kl": 2.38671875,
-      "learning_rate": 1.0554024673218806e-07,
-      "loss": 0.4845,
-      "reward": 2.6396692991256714,
-      "reward_std": 0.647905558347702,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.020053054206073284,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 437.06251525878906,
+      "epoch": 0.4775,
+      "grad_norm": 7.131878067757179,
+      "kl": 0.966796875,
+      "learning_rate": 6.626710018244987e-07,
+      "loss": 0.475,
+      "reward": 2.6469991207122803,
+      "reward_std": 0.4648020267486572,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888657093048,
+      "rewards/repetition_penalty_reward": -0.03876485675573349,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 955
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 387.5208435058594,
-      "epoch": 0.956,
-      "grad_norm": 22.09702269576327,
-      "kl": 1.2958984375,
-      "learning_rate": 1.0529722834905125e-07,
-      "loss": 0.3797,
-      "reward": 2.641054391860962,
-      "reward_std": 0.46035097539424896,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111044883728,
-      "rewards/repetition_penalty_reward": -0.016931693069636822,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 548.125,
+      "epoch": 0.478,
+      "grad_norm": 6.425684042725763,
+      "kl": 1.1748046875,
+      "learning_rate": 6.619104492241847e-07,
+      "loss": 0.6925,
+      "reward": 2.6355875730514526,
+      "reward_std": 0.6411095261573792,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.92361119389534,
+      "rewards/repetition_penalty_reward": -0.02239851839840412,
+      "rewards/tag_count_reward": 0.921875,
       "step": 956
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 429.41668701171875,
-      "epoch": 0.957,
-      "grad_norm": 34.01762179013479,
-      "kl": 1.96875,
-      "learning_rate": 1.0505962852884739e-07,
-      "loss": 0.6316,
-      "reward": 2.5664961338043213,
-      "reward_std": 0.6158457398414612,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.965277761220932,
-      "rewards/repetition_penalty_reward": -0.02378181181848049,
-      "rewards/tag_count_reward": 0.9375000298023224,
+      "completion_length": 626.2916870117188,
+      "epoch": 0.4785,
+      "grad_norm": 16.967190075717454,
+      "kl": 1.76171875,
+      "learning_rate": 6.611495557252344e-07,
+      "loss": 0.5574,
+      "reward": 2.226294755935669,
+      "reward_std": 0.6637153029441833,
+      "rewards/accuracy_reward": 0.4375000149011612,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.03585825115442276,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 957
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 401.2916717529297,
-      "epoch": 0.958,
-      "grad_norm": 27.1346809126327,
-      "kl": 2.7109375,
-      "learning_rate": 1.0482745016665526e-07,
-      "loss": 0.7956,
-      "reward": 2.382103443145752,
-      "reward_std": 0.40817123651504517,
-      "rewards/accuracy_reward": 0.5000000298023224,
-      "rewards/reasoning_steps_reward": 0.958333432674408,
-      "rewards/repetition_penalty_reward": -0.013729824218899012,
-      "rewards/tag_count_reward": 0.9375,
+      "completion_length": 404.2291717529297,
+      "epoch": 0.479,
+      "grad_norm": 10.223600950988965,
+      "kl": 0.904296875,
+      "learning_rate": 6.603883236454612e-07,
+      "loss": 0.3123,
+      "reward": 2.5562199354171753,
+      "reward_std": 0.36893555521965027,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02884970884770155,
+      "rewards/tag_count_reward": 0.9531250298023224,
       "step": 958
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 464.29168701171875,
-      "epoch": 0.959,
-      "grad_norm": 27.609867845288534,
-      "kl": 4.34375,
-      "learning_rate": 1.0460069609149496e-07,
-      "loss": 1.1173,
-      "reward": 2.4054399728775024,
-      "reward_std": 0.4486614912748337,
-      "rewards/accuracy_reward": 0.5625,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.011226738337427378,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 564.0000305175781,
+      "epoch": 0.4795,
+      "grad_norm": 13.545049398841398,
+      "kl": 1.447265625,
+      "learning_rate": 6.596267553037102e-07,
+      "loss": 0.6647,
+      "reward": 2.6013318300247192,
+      "reward_std": 0.5714467167854309,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.021932163275778294,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 959
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 582.2500305175781,
-      "epoch": 0.96,
-      "grad_norm": 34.491132802450544,
-      "kl": 5.03515625,
-      "learning_rate": 1.0437936906629334e-07,
-      "loss": 0.922,
-      "reward": 2.4019153118133545,
-      "reward_std": 0.6779120862483978,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.021695845760405064,
-      "rewards/tag_count_reward": 0.8750000298023224,
+      "completion_length": 281.9375,
+      "epoch": 0.48,
+      "grad_norm": 3.6393715149294783,
+      "kl": 0.318359375,
+      "learning_rate": 6.588648530198504e-07,
+      "loss": 0.0146,
+      "reward": 2.820993185043335,
+      "reward_std": 0.29440733790397644,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.033173730596899986,
+      "rewards/tag_count_reward": 1.0,
       "step": 960
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 404.8125,
-      "epoch": 0.961,
-      "grad_norm": 23.15463113511402,
-      "kl": 4.71875,
-      "learning_rate": 1.0416347178785039e-07,
-      "loss": 0.7217,
-      "reward": 2.643435001373291,
-      "reward_std": 0.5447419583797455,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.021495724096894264,
-      "rewards/tag_count_reward": 0.921875,
+      "completion_length": 403.41668701171875,
+      "epoch": 0.4805,
+      "grad_norm": 12.12449463950798,
+      "kl": 0.63671875,
+      "learning_rate": 6.581026191147687e-07,
+      "loss": 0.2886,
+      "reward": 2.8694658279418945,
+      "reward_std": 0.1928266827017069,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03157583996653557,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 961
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 386.6458435058594,
-      "epoch": 0.962,
-      "grad_norm": 14.191588671125146,
-      "kl": 2.1640625,
-      "learning_rate": 1.0395300688680625e-07,
-      "loss": 0.459,
-      "reward": 2.233402729034424,
-      "reward_std": 0.43939197063446045,
-      "rewards/accuracy_reward": 0.3333333544433117,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.0165973836556077,
-      "rewards/tag_count_reward": 0.9583333432674408,
+      "completion_length": 473.35418701171875,
+      "epoch": 0.481,
+      "grad_norm": 11.78122309788569,
+      "kl": 0.736328125,
+      "learning_rate": 6.573400559103613e-07,
+      "loss": 0.5582,
+      "reward": 2.6566274166107178,
+      "reward_std": 0.5360835194587708,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.023928127251565456,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 962
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 684.1458435058594,
-      "epoch": 0.963,
-      "grad_norm": 66.27238140783334,
-      "kl": 8.921875,
-      "learning_rate": 1.0374797692760933e-07,
-      "loss": 1.2039,
-      "reward": 2.404576539993286,
-      "reward_std": 0.6691045165061951,
-      "rewards/accuracy_reward": 0.6250000298023224,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.025979137048125267,
-      "rewards/tag_count_reward": 0.8125000298023224,
+      "completion_length": 446.41668701171875,
+      "epoch": 0.4815,
+      "grad_norm": 5.271282050317202,
+      "kl": 0.7451171875,
+      "learning_rate": 6.565771657295285e-07,
+      "loss": 0.5522,
+      "reward": 2.6375869512557983,
+      "reward_std": 0.5502025783061981,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.030815905891358852,
+      "rewards/tag_count_reward": 0.9322916865348816,
       "step": 963
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 517.4375,
+      "completion_length": 652.5416717529297,
+      "epoch": 0.482,
+      "grad_norm": 15.413225325982117,
+      "kl": 2.1796875,
+      "learning_rate": 6.558139508961654e-07,
+      "loss": 0.3435,
+      "reward": 2.3981354236602783,
+      "reward_std": 0.5044302940368652,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.030684133991599083,
+      "rewards/tag_count_reward": 0.8593750298023224,
+      "step": 964
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 512.5625152587891,
+      "epoch": 0.4825,
+      "grad_norm": 12.446301073299496,
+      "kl": 2.546875,
+      "learning_rate": 6.550504137351575e-07,
+      "loss": 0.6114,
+      "reward": 2.6402413845062256,
+      "reward_std": 0.6303743720054626,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.035105928778648376,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 965
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 326.625,
+      "epoch": 0.483,
+      "grad_norm": 18.969040400669726,
+      "kl": 1.03125,
+      "learning_rate": 6.542865565723707e-07,
+      "loss": 0.2782,
+      "reward": 2.5593295097351074,
+      "reward_std": 0.25267046224325895,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0135872233659029,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 966
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 586.7291870117188,
+      "epoch": 0.4835,
+      "grad_norm": 16.445189525113065,
+      "kl": 1.46875,
+      "learning_rate": 6.53522381734647e-07,
+      "loss": 0.8025,
+      "reward": 2.4304505586624146,
+      "reward_std": 0.6917058825492859,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.029619052074849606,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 967
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 352.0416717529297,
+      "epoch": 0.484,
+      "grad_norm": 12.008822684584926,
+      "kl": 0.662109375,
+      "learning_rate": 6.527578915497951e-07,
+      "loss": 0.3397,
+      "reward": 2.639287829399109,
+      "reward_std": 0.19891707226634026,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.025642716325819492,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 968
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 434.8333435058594,
+      "epoch": 0.4845,
+      "grad_norm": 20.326619395037802,
+      "kl": 0.962890625,
+      "learning_rate": 6.519930883465847e-07,
+      "loss": 0.7163,
+      "reward": 2.6135048866271973,
+      "reward_std": 0.5543282628059387,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.02538419794291258,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 969
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 655.375,
+      "epoch": 0.485,
+      "grad_norm": 19.44522032453373,
+      "kl": 1.9609375,
+      "learning_rate": 6.512279744547392e-07,
+      "loss": 0.3818,
+      "reward": 2.0434839129447937,
+      "reward_std": 0.49595144391059875,
+      "rewards/accuracy_reward": 0.2708333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.022488368675112724,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 970
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 469.16668701171875,
+      "epoch": 0.4855,
+      "grad_norm": 24.245196285909877,
+      "kl": 1.53515625,
+      "learning_rate": 6.50462552204928e-07,
+      "loss": 0.2542,
+      "reward": 2.400037169456482,
+      "reward_std": 0.32993632555007935,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03225455339998007,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 971
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 513.3333435058594,
+      "epoch": 0.486,
+      "grad_norm": 8.729059180440508,
+      "kl": 1.6484375,
+      "learning_rate": 6.496968239287603e-07,
+      "loss": 0.6719,
+      "reward": 2.4571033716201782,
+      "reward_std": 0.5960609018802643,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.01859105657786131,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 972
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 365.9583435058594,
+      "epoch": 0.4865,
+      "grad_norm": 11.187904638989847,
+      "kl": 1.177734375,
+      "learning_rate": 6.489307919587769e-07,
+      "loss": 0.3269,
+      "reward": 2.578757405281067,
+      "reward_std": 0.5115742385387421,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.0358259379863739,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 973
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 299.37501525878906,
+      "epoch": 0.487,
+      "grad_norm": 5.4073047177912175,
+      "kl": 0.423828125,
+      "learning_rate": 6.481644586284442e-07,
+      "loss": 0.2367,
+      "reward": 2.902934432029724,
+      "reward_std": 0.275345578789711,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.018940678797662258,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 974
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 521.4791870117188,
+      "epoch": 0.4875,
+      "grad_norm": 6.345217828877253,
+      "kl": 1.3828125,
+      "learning_rate": 6.473978262721463e-07,
+      "loss": 0.5127,
+      "reward": 2.6163452863693237,
+      "reward_std": 0.4908182621002197,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.039904940873384476,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 975
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 328.47918701171875,
+      "epoch": 0.488,
+      "grad_norm": 20.969372614612894,
+      "kl": 0.859375,
+      "learning_rate": 6.466308972251785e-07,
+      "loss": 0.3447,
+      "reward": 2.4979482889175415,
+      "reward_std": 0.43351081013679504,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.01767673483118415,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 976
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 553.3541870117188,
+      "epoch": 0.4885,
+      "grad_norm": 22.992172626012245,
+      "kl": 2.705078125,
+      "learning_rate": 6.458636738237395e-07,
+      "loss": 0.2706,
+      "reward": 1.9824464321136475,
+      "reward_std": 0.34522050246596336,
+      "rewards/accuracy_reward": 0.1875,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.01928970357403159,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 977
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 351.9791717529297,
+      "epoch": 0.489,
+      "grad_norm": 13.361424348905297,
+      "kl": 1.169921875,
+      "learning_rate": 6.45096158404925e-07,
+      "loss": 0.2465,
+      "reward": 2.648337244987488,
+      "reward_std": 0.3259388506412506,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.0461073312908411,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 978
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 295.31251525878906,
+      "epoch": 0.4895,
+      "grad_norm": 7.444601060796791,
+      "kl": 0.79296875,
+      "learning_rate": 6.443283533067198e-07,
+      "loss": 0.1929,
+      "reward": 2.6383087635040283,
+      "reward_std": 0.48684653639793396,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.03703854978084564,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 979
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 394.81251525878906,
+      "epoch": 0.49,
+      "grad_norm": 16.978947124748732,
+      "kl": 2.134765625,
+      "learning_rate": 6.435602608679916e-07,
+      "loss": 0.54,
+      "reward": 2.573986530303955,
+      "reward_std": 0.4741198271512985,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.014555106405168772,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 980
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 397.5,
+      "epoch": 0.4905,
+      "grad_norm": 7.762331640955408,
+      "kl": 1.609375,
+      "learning_rate": 6.427918834284834e-07,
+      "loss": 0.3917,
+      "reward": 2.5270793437957764,
+      "reward_std": 0.3988271802663803,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9513889849185944,
+      "rewards/repetition_penalty_reward": -0.02326801000162959,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 981
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 389.5625,
+      "epoch": 0.491,
+      "grad_norm": 15.572448453302112,
+      "kl": 1.1484375,
+      "learning_rate": 6.420232233288055e-07,
+      "loss": 0.2343,
+      "reward": 2.2758413553237915,
+      "reward_std": 0.42710670828819275,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.036658719182014465,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 982
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 328.1458435058594,
+      "epoch": 0.4915,
+      "grad_norm": 8.40361032839979,
+      "kl": 0.9599609375,
+      "learning_rate": 6.412542829104306e-07,
+      "loss": 0.0975,
+      "reward": 2.634483575820923,
+      "reward_std": 0.15141713619232178,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.02350257895886898,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 983
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 373.06251525878906,
+      "epoch": 0.492,
+      "grad_norm": 27.667281989688025,
+      "kl": 0.796875,
+      "learning_rate": 6.404850645156841e-07,
+      "loss": 0.3823,
+      "reward": 2.5004520416259766,
+      "reward_std": 0.42026159167289734,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.025589827448129654,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 984
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 271.50001525878906,
+      "epoch": 0.4925,
+      "grad_norm": 5.158618170324173,
+      "kl": 0.271484375,
+      "learning_rate": 6.397155704877388e-07,
+      "loss": -0.0319,
+      "reward": 2.6765854358673096,
+      "reward_std": 0.21046356856822968,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.024803485721349716,
+      "rewards/tag_count_reward": 1.0,
+      "step": 985
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 374.375,
+      "epoch": 0.493,
+      "grad_norm": 21.884347593037273,
+      "kl": 0.80859375,
+      "learning_rate": 6.389458031706068e-07,
+      "loss": 0.431,
+      "reward": 2.621227979660034,
+      "reward_std": 0.43661460280418396,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.03154988866299391,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 986
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 330.87501525878906,
+      "epoch": 0.4935,
+      "grad_norm": 4.514766065731832,
+      "kl": 0.408203125,
+      "learning_rate": 6.381757649091329e-07,
+      "loss": 0.1863,
+      "reward": 2.54175066947937,
+      "reward_std": 0.15631027286872268,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.025957662612199783,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 987
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 426.47918701171875,
+      "epoch": 0.494,
+      "grad_norm": 8.57007282438186,
+      "kl": 1.37109375,
+      "learning_rate": 6.374054580489873e-07,
+      "loss": 0.5563,
+      "reward": 2.7405987977981567,
+      "reward_std": 0.40447968570515513,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.016345561482012272,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 988
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 459.77085876464844,
+      "epoch": 0.4945,
+      "grad_norm": 7.660866604622285,
+      "kl": 1.337890625,
+      "learning_rate": 6.366348849366583e-07,
+      "loss": 0.1801,
+      "reward": 2.6990854740142822,
+      "reward_std": 0.36674173176288605,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.04570627398788929,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 989
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 329.97918701171875,
+      "epoch": 0.495,
+      "grad_norm": 21.838061572005632,
+      "kl": 1.4375,
+      "learning_rate": 6.358640479194451e-07,
+      "loss": 0.4576,
+      "reward": 2.5701619386672974,
+      "reward_std": 0.5304689109325409,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.01664359588176012,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 990
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 327.43751525878906,
+      "epoch": 0.4955,
+      "grad_norm": 24.17528666080015,
+      "kl": 0.6298828125,
+      "learning_rate": 6.35092949345451e-07,
+      "loss": 0.3975,
+      "reward": 2.8759747743606567,
+      "reward_std": 0.2522295266389847,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.03201141953468323,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 991
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 399.31251525878906,
+      "epoch": 0.496,
+      "grad_norm": 32.1142545684353,
+      "kl": 3.203125,
+      "learning_rate": 6.343215915635761e-07,
+      "loss": 0.7948,
+      "reward": 2.630898594856262,
+      "reward_std": 0.6931562125682831,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.027087541297078133,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 992
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 471.10418701171875,
+      "epoch": 0.4965,
+      "grad_norm": 48.892823948801954,
+      "kl": 3.1953125,
+      "learning_rate": 6.335499769235098e-07,
+      "loss": 0.6139,
+      "reward": 2.738042116165161,
+      "reward_std": 0.37622474133968353,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.05015224777162075,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 993
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 330.87500762939453,
+      "epoch": 0.497,
+      "grad_norm": 28.390795641858606,
+      "kl": 1.5107421875,
+      "learning_rate": 6.327781077757241e-07,
+      "loss": 0.2853,
+      "reward": 2.7957078218460083,
+      "reward_std": 0.19093896262347698,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.025472811423242092,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 994
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 329.6666717529297,
+      "epoch": 0.4975,
+      "grad_norm": 10.891904018933554,
+      "kl": 1.03515625,
+      "learning_rate": 6.320059864714664e-07,
+      "loss": 0.1864,
+      "reward": 2.720821738243103,
+      "reward_std": 0.32492922246456146,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.02570601273328066,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 995
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 435.1041717529297,
+      "epoch": 0.498,
+      "grad_norm": 27.389591437302162,
+      "kl": 1.552734375,
+      "learning_rate": 6.31233615362752e-07,
+      "loss": 0.7395,
+      "reward": 2.655308961868286,
+      "reward_std": 0.42288788408041,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03392717707902193,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 996
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 394.4583435058594,
+      "epoch": 0.4985,
+      "grad_norm": 37.45628290835572,
+      "kl": 1.1953125,
+      "learning_rate": 6.304609968023572e-07,
+      "loss": 0.5472,
+      "reward": 2.78287935256958,
+      "reward_std": 0.4373747333884239,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.031356871128082275,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 997
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 440.3333435058594,
+      "epoch": 0.499,
+      "grad_norm": 8.83893360388916,
+      "kl": 1.12890625,
+      "learning_rate": 6.296881331438126e-07,
+      "loss": 0.3209,
+      "reward": 2.574609398841858,
+      "reward_std": 0.4172719120979309,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.019140728749334812,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 998
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 448.4166717529297,
+      "epoch": 0.4995,
+      "grad_norm": 25.468071616698786,
+      "kl": 1.109375,
+      "learning_rate": 6.289150267413942e-07,
+      "loss": 0.5873,
+      "reward": 2.646699547767639,
+      "reward_std": 0.38845836371183395,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.023439443670213223,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 999
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 591.1041717529297,
+      "epoch": 0.5,
+      "grad_norm": 9.084576194579185,
+      "kl": 2.25,
+      "learning_rate": 6.281416799501187e-07,
+      "loss": 0.5736,
+      "reward": 2.3330485820770264,
+      "reward_std": 0.5498259365558624,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.019382060505449772,
+      "rewards/tag_count_reward": 0.8802083730697632,
+      "step": 1000
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 461.8333435058594,
+      "epoch": 0.5005,
+      "grad_norm": 9.95609254967102,
+      "kl": 1.60546875,
+      "learning_rate": 6.273680951257342e-07,
+      "loss": 0.5362,
+      "reward": 2.549267292022705,
+      "reward_std": 0.6142518520355225,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.03580236993730068,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1001
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 455.0833435058594,
+      "epoch": 0.501,
+      "grad_norm": 14.049667830028957,
+      "kl": 2.2265625,
+      "learning_rate": 6.265942746247146e-07,
+      "loss": 0.7722,
+      "reward": 2.731945276260376,
+      "reward_std": 0.5192966759204865,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.021526905708014965,
+      "rewards/tag_count_reward": 0.9270833730697632,
+      "step": 1002
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 623.5625305175781,
+      "epoch": 0.5015,
+      "grad_norm": 31.177249787933718,
+      "kl": 2.55859375,
+      "learning_rate": 6.258202208042511e-07,
+      "loss": 1.1778,
+      "reward": 2.524030923843384,
+      "reward_std": 0.7691468000411987,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.02110826689749956,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1003
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 584.2083435058594,
+      "epoch": 0.502,
+      "grad_norm": 39.78085883662432,
+      "kl": 3.6015625,
+      "learning_rate": 6.25045936022246e-07,
+      "loss": 1.1624,
+      "reward": 2.3681070804595947,
+      "reward_std": 0.57652947306633,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.015573485288769007,
+      "rewards/tag_count_reward": 0.9114583730697632,
+      "step": 1004
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 541.2083587646484,
+      "epoch": 0.5025,
+      "grad_norm": 15.400304918011507,
+      "kl": 2.7578125,
+      "learning_rate": 6.242714226373049e-07,
+      "loss": 0.8093,
+      "reward": 2.1841408014297485,
+      "reward_std": 0.5138780176639557,
+      "rewards/accuracy_reward": 0.3750000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.015512102749198675,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1005
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 517.3541870117188,
+      "epoch": 0.503,
+      "grad_norm": 6.931951953481591,
+      "kl": 2.15234375,
+      "learning_rate": 6.2349668300873e-07,
+      "loss": 0.6114,
+      "reward": 2.5530234575271606,
+      "reward_std": 0.6553203761577606,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.03030997794121504,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1006
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 549.6666870117188,
+      "epoch": 0.5035,
+      "grad_norm": 22.648929704749897,
+      "kl": 1.70703125,
+      "learning_rate": 6.227217194965125e-07,
+      "loss": 0.8695,
+      "reward": 2.318639397621155,
+      "reward_std": 0.6079018115997314,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.014693964272737503,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1007
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 386.8333435058594,
+      "epoch": 0.504,
+      "grad_norm": 28.81093113427859,
+      "kl": 0.859375,
+      "learning_rate": 6.219465344613258e-07,
+      "loss": 0.4981,
+      "reward": 2.6072503328323364,
+      "reward_std": 0.45103733241558075,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.026430224999785423,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1008
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 579.2500305175781,
+      "epoch": 0.5045,
+      "grad_norm": 18.90641611996879,
+      "kl": 1.79296875,
+      "learning_rate": 6.211711302645177e-07,
+      "loss": 0.7654,
+      "reward": 2.513100743293762,
+      "reward_std": 0.6586687564849854,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.025093771517276764,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1009
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 316.3333435058594,
+      "epoch": 0.505,
+      "grad_norm": 10.521243740499507,
+      "kl": 0.7080078125,
+      "learning_rate": 6.203955092681039e-07,
+      "loss": 0.2849,
+      "reward": 2.616108775138855,
+      "reward_std": 0.26797990035265684,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.031460804864764214,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1010
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 762.6875457763672,
+      "epoch": 0.5055,
+      "grad_norm": 17.712711157663584,
+      "kl": 2.359375,
+      "learning_rate": 6.196196738347607e-07,
+      "loss": 0.5658,
+      "reward": 2.252875328063965,
+      "reward_std": 0.44946494698524475,
+      "rewards/accuracy_reward": 0.4375000149011612,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.0370553620159626,
+      "rewards/tag_count_reward": 0.8593750298023224,
+      "step": 1011
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 358.9791717529297,
+      "epoch": 0.506,
+      "grad_norm": 6.724680601267686,
+      "kl": 0.7841796875,
+      "learning_rate": 6.188436263278172e-07,
+      "loss": 0.371,
+      "reward": 2.8078267574310303,
+      "reward_std": 0.3918469473719597,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.018562145065516233,
+      "rewards/tag_count_reward": 0.9583333730697632,
+      "step": 1012
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 381.6666717529297,
+      "epoch": 0.5065,
+      "grad_norm": 7.753482099144785,
+      "kl": 1.005859375,
+      "learning_rate": 6.180673691112486e-07,
+      "loss": 0.7111,
+      "reward": 2.758282780647278,
+      "reward_std": 0.42837944626808167,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.02643944276496768,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1013
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 618.2500305175781,
+      "epoch": 0.507,
+      "grad_norm": 14.72862436999275,
+      "kl": 1.84765625,
+      "learning_rate": 6.172909045496694e-07,
+      "loss": 1.0001,
+      "reward": 2.36434543132782,
+      "reward_std": 0.6381807327270508,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.0245436392724514,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1014
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 824.8750457763672,
+      "epoch": 0.5075,
+      "grad_norm": 16.05316707574026,
+      "kl": 1.919921875,
+      "learning_rate": 6.165142350083249e-07,
+      "loss": 0.9197,
+      "reward": 2.327476441860199,
+      "reward_std": 0.8264204859733582,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.8750000298023224,
+      "rewards/repetition_penalty_reward": -0.016273547895252705,
+      "rewards/tag_count_reward": 0.78125,
+      "step": 1015
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 662.8333435058594,
+      "epoch": 0.508,
+      "grad_norm": 14.986219495301736,
+      "kl": 1.734375,
+      "learning_rate": 6.157373628530852e-07,
+      "loss": 0.6306,
+      "reward": 2.3806835412979126,
+      "reward_std": 0.4617319107055664,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.018622069619596004,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1016
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 493.97918701171875,
+      "epoch": 0.5085,
+      "grad_norm": 10.603280718054956,
+      "kl": 1.19921875,
+      "learning_rate": 6.149602904504378e-07,
+      "loss": 0.6167,
+      "reward": 2.702357053756714,
+      "reward_std": 0.46163134276866913,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.026809771545231342,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1017
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 505.5,
+      "epoch": 0.509,
+      "grad_norm": 12.432373079567979,
+      "kl": 1.107421875,
+      "learning_rate": 6.141830201674802e-07,
+      "loss": 0.7902,
+      "reward": 2.693175435066223,
+      "reward_std": 0.5161421597003937,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03425509575754404,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1018
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 632.8750305175781,
+      "epoch": 0.5095,
+      "grad_norm": 7.175100274761201,
+      "kl": 1.376953125,
+      "learning_rate": 6.134055543719121e-07,
+      "loss": 0.4146,
+      "reward": 2.4220484495162964,
+      "reward_std": 0.5273237824440002,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.027604437433183193,
+      "rewards/tag_count_reward": 0.8593750298023224,
+      "step": 1019
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 456.6250305175781,
+      "epoch": 0.51,
+      "grad_norm": 16.588343745029032,
+      "kl": 0.994140625,
+      "learning_rate": 6.126278954320294e-07,
+      "loss": 0.7095,
+      "reward": 2.6970959901809692,
+      "reward_std": 0.5613195151090622,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.023390086367726326,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1020
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 734.9375,
+      "epoch": 0.5105,
+      "grad_norm": 9.757877385128149,
+      "kl": 1.38671875,
+      "learning_rate": 6.118500457167159e-07,
+      "loss": 0.8497,
+      "reward": 2.0375831723213196,
+      "reward_std": 0.6933330148458481,
+      "rewards/accuracy_reward": 0.3750000149011612,
+      "rewards/reasoning_steps_reward": 0.8402778208255768,
+      "rewards/repetition_penalty_reward": -0.021444641053676605,
+      "rewards/tag_count_reward": 0.8437500298023224,
+      "step": 1021
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 400.6250305175781,
+      "epoch": 0.511,
+      "grad_norm": 3.581679020231462,
+      "kl": 0.5986328125,
+      "learning_rate": 6.11072007595437e-07,
+      "loss": 0.2243,
+      "reward": 2.635998249053955,
+      "reward_std": 0.3600511699914932,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.020251845009624958,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1022
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 532.3125305175781,
+      "epoch": 0.5115,
+      "grad_norm": 17.57086871577716,
+      "kl": 0.810546875,
+      "learning_rate": 6.102937834382315e-07,
+      "loss": 0.7239,
+      "reward": 2.707748532295227,
+      "reward_std": 0.5111279785633087,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.021418226417154074,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1023
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 446.0000305175781,
+      "epoch": 0.512,
+      "grad_norm": 5.091174343110264,
+      "kl": 0.666015625,
+      "learning_rate": 6.095153756157051e-07,
+      "loss": 0.4922,
+      "reward": 2.5856151580810547,
+      "reward_std": 0.5574827194213867,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.027232157066464424,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1024
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 425.56251525878906,
+      "epoch": 0.5125,
+      "grad_norm": 7.051286670184644,
+      "kl": 0.54296875,
+      "learning_rate": 6.087367864990232e-07,
+      "loss": 0.4512,
+      "reward": 2.401527166366577,
+      "reward_std": 0.45996415615081787,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.027292468585073948,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1025
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 481.41668701171875,
+      "epoch": 0.513,
+      "grad_norm": 6.804881212406539,
+      "kl": 0.7265625,
+      "learning_rate": 6.079580184599032e-07,
+      "loss": 0.752,
+      "reward": 2.53912889957428,
+      "reward_std": 0.4916386604309082,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9513888657093048,
+      "rewards/repetition_penalty_reward": -0.03205180913209915,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1026
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 428.25001525878906,
+      "epoch": 0.5135,
+      "grad_norm": 4.0838002502110955,
+      "kl": 0.689453125,
+      "learning_rate": 6.071790738706078e-07,
+      "loss": 0.7677,
+      "reward": 2.633172869682312,
+      "reward_std": 0.5927000939846039,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.030021720565855503,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1027
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 445.8750305175781,
+      "epoch": 0.514,
+      "grad_norm": 6.82558184595021,
+      "kl": 0.6650390625,
+      "learning_rate": 6.06399955103937e-07,
+      "loss": 0.3746,
+      "reward": 2.3708232641220093,
+      "reward_std": 0.47719016671180725,
+      "rewards/accuracy_reward": 0.5000000223517418,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.021537834778428078,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1028
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 612.4375305175781,
+      "epoch": 0.5145,
+      "grad_norm": 29.593011906377633,
+      "kl": 1.89453125,
+      "learning_rate": 6.056206645332217e-07,
+      "loss": 0.6587,
+      "reward": 2.507744789123535,
+      "reward_std": 0.5372722446918488,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.03218586929142475,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1029
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 447.0208435058594,
+      "epoch": 0.515,
+      "grad_norm": 14.40828703378696,
+      "kl": 1.123046875,
+      "learning_rate": 6.048412045323164e-07,
+      "loss": 0.3732,
+      "reward": 2.6217926740646362,
+      "reward_std": 0.3849295526742935,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.03966595232486725,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1030
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 443.4583435058594,
+      "epoch": 0.5155,
+      "grad_norm": 14.1549537273764,
+      "kl": 1.20703125,
+      "learning_rate": 6.040615774755911e-07,
+      "loss": 0.8186,
+      "reward": 2.3135095834732056,
+      "reward_std": 0.4170246571302414,
+      "rewards/accuracy_reward": 0.45833333395421505,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.01461556926369667,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1031
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 668.5625,
+      "epoch": 0.516,
+      "grad_norm": 16.64560517730956,
+      "kl": 1.71875,
+      "learning_rate": 6.032817857379256e-07,
+      "loss": 0.5198,
+      "reward": 2.264461040496826,
+      "reward_std": 0.5811220407485962,
+      "rewards/accuracy_reward": 0.5416666716337204,
+      "rewards/reasoning_steps_reward": 0.9166666567325592,
+      "rewards/repetition_penalty_reward": -0.02720571681857109,
+      "rewards/tag_count_reward": 0.8333333432674408,
+      "step": 1032
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.93751525878906,
+      "epoch": 0.5165,
+      "grad_norm": 4.122201207331607,
+      "kl": 0.443359375,
+      "learning_rate": 6.025018316946999e-07,
+      "loss": 0.2138,
+      "reward": 2.7509995698928833,
+      "reward_std": 0.42070019245147705,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.026778437197208405,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1033
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 630.0833740234375,
+      "epoch": 0.517,
+      "grad_norm": 5.176497247797343,
+      "kl": 1.1875,
+      "learning_rate": 6.017217177217899e-07,
+      "loss": 0.7832,
+      "reward": 2.3645520210266113,
+      "reward_std": 0.8221311867237091,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.02954527549445629,
+      "rewards/tag_count_reward": 0.859375,
+      "step": 1034
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 402.08335876464844,
+      "epoch": 0.5175,
+      "grad_norm": 4.289877365631067,
+      "kl": 0.6337890625,
+      "learning_rate": 6.009414461955581e-07,
+      "loss": 0.3456,
+      "reward": 2.8156780004501343,
+      "reward_std": 0.25007878383621573,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.012447083368897438,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1035
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 511.6666717529297,
+      "epoch": 0.518,
+      "grad_norm": 4.534984555827674,
+      "kl": 0.7265625,
+      "learning_rate": 6.001610194928464e-07,
+      "loss": 0.6627,
+      "reward": 2.6687086820602417,
+      "reward_std": 0.5741814076900482,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.02052734326571226,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1036
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.4166717529297,
+      "epoch": 0.5185,
+      "grad_norm": 9.036303143255743,
+      "kl": 0.3173828125,
+      "learning_rate": 5.993804399909703e-07,
+      "loss": 0.2845,
+      "reward": 2.868683695793152,
+      "reward_std": 0.25010958313941956,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.023677408695220947,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1037
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 581.3125305175781,
+      "epoch": 0.519,
+      "grad_norm": 4.473097097190504,
+      "kl": 1.04296875,
+      "learning_rate": 5.985997100677103e-07,
+      "loss": 0.3827,
+      "reward": 2.478750705718994,
+      "reward_std": 0.30521145928651094,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.01083279075101018,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1038
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 448.91668701171875,
+      "epoch": 0.5195,
+      "grad_norm": 4.330270457188464,
+      "kl": 0.677734375,
+      "learning_rate": 5.97818832101305e-07,
+      "loss": 0.4467,
+      "reward": 2.632758140563965,
+      "reward_std": 0.37518632411956787,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.025228080339729786,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1039
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 432.4166717529297,
+      "epoch": 0.52,
+      "grad_norm": 5.413314396002232,
+      "kl": 0.7490234375,
+      "learning_rate": 5.97037808470444e-07,
+      "loss": 0.4434,
+      "reward": 2.658997416496277,
+      "reward_std": 0.548340767621994,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.021558281034231186,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1040
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 462.7083435058594,
+      "epoch": 0.5205,
+      "grad_norm": 13.173521990092542,
+      "kl": 0.990234375,
+      "learning_rate": 5.96256641554261e-07,
+      "loss": 0.9078,
+      "reward": 2.477555751800537,
+      "reward_std": 0.6211674511432648,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889849185944,
+      "rewards/repetition_penalty_reward": -0.025916431099176407,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1041
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 747.7500305175781,
+      "epoch": 0.521,
+      "grad_norm": 5.994356817443346,
+      "kl": 1.3681640625,
+      "learning_rate": 5.954753337323259e-07,
+      "loss": 0.3688,
+      "reward": 2.3351893424987793,
+      "reward_std": 0.5483206920325756,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.875,
+      "rewards/repetition_penalty_reward": -0.024185655638575554,
+      "rewards/tag_count_reward": 0.796875,
+      "step": 1042
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 641.5833587646484,
+      "epoch": 0.5215,
+      "grad_norm": 6.65678571854715,
+      "kl": 1.65234375,
+      "learning_rate": 5.946938873846375e-07,
+      "loss": 0.6598,
+      "reward": 2.1638104915618896,
+      "reward_std": 0.6675990223884583,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.8541667461395264,
+      "rewards/repetition_penalty_reward": -0.02889794297516346,
+      "rewards/tag_count_reward": 0.8177083432674408,
+      "step": 1043
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 581.5625,
+      "epoch": 0.522,
+      "grad_norm": 7.592842509620014,
+      "kl": 1.189453125,
+      "learning_rate": 5.939123048916173e-07,
+      "loss": 0.836,
+      "reward": 2.385079026222229,
+      "reward_std": 0.90406933426857,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.8680556118488312,
+      "rewards/repetition_penalty_reward": -0.019434815272688866,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1044
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 786.4583740234375,
+      "epoch": 0.5225,
+      "grad_norm": 4.691007195698727,
+      "kl": 1.4140625,
+      "learning_rate": 5.931305886341008e-07,
+      "loss": 0.7366,
+      "reward": 2.2028392553329468,
+      "reward_std": 0.9715997576713562,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.8541666865348816,
+      "rewards/repetition_penalty_reward": -0.015910686925053596,
+      "rewards/tag_count_reward": 0.7604166865348816,
+      "step": 1045
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 484.85418701171875,
+      "epoch": 0.523,
+      "grad_norm": 11.332963488425568,
+      "kl": 0.888671875,
+      "learning_rate": 5.923487409933315e-07,
+      "loss": 0.4905,
+      "reward": 2.6106256246566772,
+      "reward_std": 0.4605254530906677,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.016110544558614492,
+      "rewards/tag_count_reward": 0.9114583730697632,
+      "step": 1046
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 620.1250152587891,
+      "epoch": 0.5235,
+      "grad_norm": 7.355316334467517,
+      "kl": 1.28125,
+      "learning_rate": 5.915667643509528e-07,
+      "loss": 0.5121,
+      "reward": 2.435990571975708,
+      "reward_std": 0.720505028963089,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.03449564054608345,
+      "rewards/tag_count_reward": 0.8385416865348816,
+      "step": 1047
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 493.79168701171875,
+      "epoch": 0.524,
+      "grad_norm": 10.744727961689891,
+      "kl": 0.791015625,
+      "learning_rate": 5.907846610890011e-07,
+      "loss": 0.5435,
+      "reward": 2.488598942756653,
+      "reward_std": 0.5222236812114716,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.028762370347976685,
+      "rewards/tag_count_reward": 0.9270833730697632,
+      "step": 1048
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 643.3958435058594,
+      "epoch": 0.5245,
+      "grad_norm": 10.473792543164542,
+      "kl": 1.25,
+      "learning_rate": 5.900024335898987e-07,
+      "loss": 0.7268,
+      "reward": 2.29390549659729,
+      "reward_std": 0.8730533719062805,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.03769184276461601,
+      "rewards/tag_count_reward": 0.8385416865348816,
+      "step": 1049
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 499.35418701171875,
+      "epoch": 0.525,
+      "grad_norm": 12.588187330698355,
+      "kl": 1.1171875,
+      "learning_rate": 5.892200842364462e-07,
+      "loss": 0.5724,
+      "reward": 2.6492995023727417,
+      "reward_std": 0.7308410704135895,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.03299206867814064,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1050
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 393.375,
+      "epoch": 0.5255,
+      "grad_norm": 10.44138233428014,
+      "kl": 0.4931640625,
+      "learning_rate": 5.884376154118154e-07,
+      "loss": 0.2827,
+      "reward": 2.626540184020996,
+      "reward_std": 0.44100892543792725,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.017557154409587383,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1051
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 692.6458435058594,
+      "epoch": 0.526,
+      "grad_norm": 18.556893989812647,
+      "kl": 2.2890625,
+      "learning_rate": 5.87655029499542e-07,
+      "loss": 0.7235,
+      "reward": 2.1944429874420166,
+      "reward_std": 0.7475170493125916,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.02604318968951702,
+      "rewards/tag_count_reward": 0.8385416865348816,
+      "step": 1052
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 561.8125,
+      "epoch": 0.5265,
+      "grad_norm": 23.386680578226347,
+      "kl": 2.37890625,
+      "learning_rate": 5.868723288835184e-07,
+      "loss": 0.6783,
+      "reward": 2.347724437713623,
+      "reward_std": 0.5882080346345901,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.013386863050982356,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1053
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 507.18751525878906,
+      "epoch": 0.527,
+      "grad_norm": 25.81517984790388,
+      "kl": 1.8828125,
+      "learning_rate": 5.860895159479864e-07,
+      "loss": 0.7704,
+      "reward": 2.459168314933777,
+      "reward_std": 0.5893406569957733,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.030415082350373268,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1054
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 358.00001525878906,
+      "epoch": 0.5275,
+      "grad_norm": 5.939357848487736,
+      "kl": 0.48828125,
+      "learning_rate": 5.853065930775303e-07,
+      "loss": 0.2008,
+      "reward": 2.8095940351486206,
+      "reward_std": 0.29858143627643585,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.023739230819046497,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1055
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 587.3125,
+      "epoch": 0.528,
+      "grad_norm": 29.053870250832556,
+      "kl": 2.244140625,
+      "learning_rate": 5.845235626570683e-07,
+      "loss": 0.4827,
+      "reward": 2.430117607116699,
+      "reward_std": 0.5931105017662048,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.03689653240144253,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1056
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 464.2916717529297,
+      "epoch": 0.5285,
+      "grad_norm": 19.31179372615611,
+      "kl": 1.26953125,
+      "learning_rate": 5.837404270718475e-07,
+      "loss": 0.7061,
+      "reward": 2.5217255353927612,
+      "reward_std": 0.4243352711200714,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.04424683563411236,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1057
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 326.8541717529297,
+      "epoch": 0.529,
+      "grad_norm": 6.273677914137758,
+      "kl": 0.4892578125,
+      "learning_rate": 5.829571887074343e-07,
+      "loss": 0.1964,
+      "reward": 2.713360071182251,
+      "reward_std": 0.12399672530591488,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03490386623889208,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1058
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 351.7291717529297,
+      "epoch": 0.5295,
+      "grad_norm": 4.222827898707743,
+      "kl": 0.4443359375,
+      "learning_rate": 5.821738499497086e-07,
+      "loss": 0.1775,
+      "reward": 2.596918225288391,
+      "reward_std": 0.3194064572453499,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03329028841108084,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1059
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 304.06251525878906,
+      "epoch": 0.53,
+      "grad_norm": 3.4491270319460914,
+      "kl": 0.31640625,
+      "learning_rate": 5.813904131848564e-07,
+      "loss": 0.0394,
+      "reward": 2.959268808364868,
+      "reward_std": 0.0850947042927146,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.019898047670722008,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1060
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 392.60418701171875,
+      "epoch": 0.5305,
+      "grad_norm": 7.953501592426025,
+      "kl": 0.4990234375,
+      "learning_rate": 5.806068807993617e-07,
+      "loss": 0.4339,
+      "reward": 2.659141421318054,
+      "reward_std": 0.45532485842704773,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.024886406026780605,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1061
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 335.18751525878906,
+      "epoch": 0.531,
+      "grad_norm": 3.524239803768179,
+      "kl": 0.40625,
+      "learning_rate": 5.798232551800002e-07,
+      "loss": 0.0785,
+      "reward": 2.477108120918274,
+      "reward_std": 0.38651110231876373,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03330867923796177,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1062
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 310.3541717529297,
+      "epoch": 0.5315,
+      "grad_norm": 3.4813021990603783,
+      "kl": 0.353515625,
+      "learning_rate": 5.790395387138311e-07,
+      "loss": 0.0409,
+      "reward": 2.969285249710083,
+      "reward_std": 0.014413285069167614,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.030714819207787514,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1063
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 387.9791717529297,
+      "epoch": 0.532,
+      "grad_norm": 7.496016496118182,
+      "kl": 0.8828125,
+      "learning_rate": 5.78255733788191e-07,
+      "loss": 0.6521,
+      "reward": 2.826792359352112,
+      "reward_std": 0.36261652410030365,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.029110469855368137,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1064
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 370.5833435058594,
+      "epoch": 0.5325,
+      "grad_norm": 5.973629081248443,
+      "kl": 0.701171875,
+      "learning_rate": 5.774718427906856e-07,
+      "loss": 0.3884,
+      "reward": 2.569111704826355,
+      "reward_std": 0.5015820562839508,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9513888657093048,
+      "rewards/repetition_penalty_reward": -0.01769387163221836,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1065
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.5416717529297,
+      "epoch": 0.533,
+      "grad_norm": 3.6653468809368026,
+      "kl": 0.345703125,
+      "learning_rate": 5.766878681091828e-07,
+      "loss": 0.0043,
+      "reward": 2.7937185764312744,
+      "reward_std": 0.12711532320827246,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.039614940993487835,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1066
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 330.7916717529297,
+      "epoch": 0.5335,
+      "grad_norm": 5.676230222355891,
+      "kl": 0.3671875,
+      "learning_rate": 5.759038121318052e-07,
+      "loss": 0.1888,
+      "reward": 2.7420976161956787,
+      "reward_std": 0.3609135150909424,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.026999651454389095,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1067
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 309.00001525878906,
+      "epoch": 0.534,
+      "grad_norm": 2.9464749451257464,
+      "kl": 0.33203125,
+      "learning_rate": 5.751196772469237e-07,
+      "loss": 0.035,
+      "reward": 2.555404543876648,
+      "reward_std": 0.19347204267978668,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.027928968891501427,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1068
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 346.6041717529297,
+      "epoch": 0.5345,
+      "grad_norm": 4.718857747074456,
+      "kl": 0.4462890625,
+      "learning_rate": 5.743354658431489e-07,
+      "loss": 0.2027,
+      "reward": 2.7974963188171387,
+      "reward_std": 0.351472407579422,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.015003710053861141,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1069
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 368.68751525878906,
+      "epoch": 0.535,
+      "grad_norm": 7.916520582997023,
+      "kl": 0.783203125,
+      "learning_rate": 5.735511803093248e-07,
+      "loss": 0.7907,
+      "reward": 2.7259095907211304,
+      "reward_std": 0.5795675814151764,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.018882008269429207,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1070
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 383.5416717529297,
+      "epoch": 0.5355,
+      "grad_norm": 89.56603759912137,
+      "kl": 1.296875,
+      "learning_rate": 5.727668230345209e-07,
+      "loss": 0.6934,
+      "reward": 2.24530291557312,
+      "reward_std": 0.495767742395401,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.023794405162334442,
+      "rewards/tag_count_reward": 0.9427083730697632,
+      "step": 1071
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 515.4791717529297,
+      "epoch": 0.536,
+      "grad_norm": 9.256848299391667,
+      "kl": 1.056640625,
+      "learning_rate": 5.71982396408026e-07,
+      "loss": 0.3638,
+      "reward": 2.5147180557250977,
+      "reward_std": 0.4126999229192734,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.016532148234546185,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1072
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 496.6250305175781,
+      "epoch": 0.5365,
+      "grad_norm": 9.704892056750113,
+      "kl": 0.923828125,
+      "learning_rate": 5.711979028193391e-07,
+      "loss": 0.7874,
+      "reward": 2.534700036048889,
+      "reward_std": 0.4814092218875885,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.022591806016862392,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1073
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 442.4583435058594,
+      "epoch": 0.537,
+      "grad_norm": 4.434868047648692,
+      "kl": 0.734375,
+      "learning_rate": 5.704133446581642e-07,
+      "loss": 0.594,
+      "reward": 2.7428154945373535,
+      "reward_std": 0.5054127871990204,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.014129179995507002,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1074
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 739.2708740234375,
+      "epoch": 0.5375,
+      "grad_norm": 10.643100205082053,
+      "kl": 1.72265625,
+      "learning_rate": 5.696287243144012e-07,
+      "loss": 0.8503,
+      "reward": 2.1527023315429688,
+      "reward_std": 0.8699755072593689,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.8680556118488312,
+      "rewards/repetition_penalty_reward": -0.01222835062071681,
+      "rewards/tag_count_reward": 0.7760416865348816,
+      "step": 1075
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 386.875,
+      "epoch": 0.538,
+      "grad_norm": 5.50232302994543,
+      "kl": 0.69921875,
+      "learning_rate": 5.688440441781398e-07,
+      "loss": 0.3794,
+      "reward": 2.6137609481811523,
+      "reward_std": 0.37844765186309814,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.026864062063395977,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1076
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 348.2083435058594,
+      "epoch": 0.5385,
+      "grad_norm": 4.831237438916453,
+      "kl": 0.5751953125,
+      "learning_rate": 5.680593066396518e-07,
+      "loss": 0.4982,
+      "reward": 2.895869016647339,
+      "reward_std": 0.267215795814991,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.027742099948227406,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1077
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 411.45835876464844,
+      "epoch": 0.539,
+      "grad_norm": 4.089673654340983,
+      "kl": 0.5966796875,
+      "learning_rate": 5.672745140893839e-07,
+      "loss": 0.449,
+      "reward": 2.5935864448547363,
+      "reward_std": 0.31501081585884094,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.022733049467206,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1078
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 500.3541717529297,
+      "epoch": 0.5395,
+      "grad_norm": 3.580216142280184,
+      "kl": 0.658203125,
+      "learning_rate": 5.664896689179504e-07,
+      "loss": 0.6341,
+      "reward": 2.669227361679077,
+      "reward_std": 0.7126676142215729,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.028689204715192318,
+      "rewards/tag_count_reward": 0.9062500298023224,
+      "step": 1079
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 781.6667175292969,
+      "epoch": 0.54,
+      "grad_norm": 4.955650799366805,
+      "kl": 1.166015625,
+      "learning_rate": 5.657047735161255e-07,
+      "loss": 0.8425,
+      "reward": 2.199775218963623,
+      "reward_std": 0.9051741063594818,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.017238642554730177,
+      "rewards/tag_count_reward": 0.7864583730697632,
+      "step": 1080
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 399.72918701171875,
+      "epoch": 0.5405,
+      "grad_norm": 4.90026140945702,
+      "kl": 0.5576171875,
+      "learning_rate": 5.649198302748368e-07,
+      "loss": 0.4486,
+      "reward": 2.7975977659225464,
+      "reward_std": 0.45839452743530273,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.021846801973879337,
+      "rewards/tag_count_reward": 0.9583333730697632,
+      "step": 1081
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 757.9791870117188,
+      "epoch": 0.541,
+      "grad_norm": 8.049018307327215,
+      "kl": 1.220703125,
+      "learning_rate": 5.641348415851577e-07,
+      "loss": 0.7273,
+      "reward": 2.0966323614120483,
+      "reward_std": 0.7242147624492645,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.01968714315444231,
+      "rewards/tag_count_reward": 0.8177083730697632,
+      "step": 1082
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 374.2916717529297,
+      "epoch": 0.5415,
+      "grad_norm": 8.835700341793663,
+      "kl": 0.77734375,
+      "learning_rate": 5.633498098382998e-07,
+      "loss": 0.4341,
+      "reward": 2.6963536739349365,
+      "reward_std": 0.3628556430339813,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.022396229207515717,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1083
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 802.6875,
+      "epoch": 0.542,
+      "grad_norm": 9.101773741644875,
+      "kl": 1.55078125,
+      "learning_rate": 5.625647374256061e-07,
+      "loss": 0.741,
+      "reward": 2.3281177282333374,
+      "reward_std": 0.6684702336788177,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.909722238779068,
+      "rewards/repetition_penalty_reward": -0.013896321877837181,
+      "rewards/tag_count_reward": 0.7656250298023224,
+      "step": 1084
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 559.9166870117188,
+      "epoch": 0.5425,
+      "grad_norm": 3.17667785466319,
+      "kl": 0.849609375,
+      "learning_rate": 5.617796267385429e-07,
+      "loss": 0.5913,
+      "reward": 2.5073158740997314,
+      "reward_std": 0.6131802946329117,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03261485882103443,
+      "rewards/tag_count_reward": 0.8802083730697632,
+      "step": 1085
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 395.54168701171875,
+      "epoch": 0.543,
+      "grad_norm": 5.388043302367293,
+      "kl": 0.642578125,
+      "learning_rate": 5.60994480168694e-07,
+      "loss": 0.369,
+      "reward": 2.5772793292999268,
+      "reward_std": 0.5051662921905518,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.01820683665573597,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1086
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 585.6250305175781,
+      "epoch": 0.5435,
+      "grad_norm": 9.467820919919676,
+      "kl": 0.974609375,
+      "learning_rate": 5.602093001077517e-07,
+      "loss": 0.667,
+      "reward": 2.562462568283081,
+      "reward_std": 0.6790414154529572,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.010454241652041674,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1087
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 357.81251525878906,
+      "epoch": 0.544,
+      "grad_norm": 10.151144754971412,
+      "kl": 0.7001953125,
+      "learning_rate": 5.594240889475106e-07,
+      "loss": 0.615,
+      "reward": 2.8339508771896362,
+      "reward_std": 0.4007309675216675,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.023687981069087982,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1088
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 653.4583435058594,
+      "epoch": 0.5445,
+      "grad_norm": 5.455861003165772,
+      "kl": 0.9765625,
+      "learning_rate": 5.586388490798604e-07,
+      "loss": 0.6132,
+      "reward": 2.212650716304779,
+      "reward_std": 0.6732420325279236,
+      "rewards/accuracy_reward": 0.4375000223517418,
+      "rewards/reasoning_steps_reward": 0.92361119389534,
+      "rewards/repetition_penalty_reward": -0.018252158537507057,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1089
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 648.2291870117188,
+      "epoch": 0.545,
+      "grad_norm": 7.571198091437875,
+      "kl": 1.234375,
+      "learning_rate": 5.578535828967777e-07,
+      "loss": 0.8813,
+      "reward": 2.416181802749634,
+      "reward_std": 0.9853273630142212,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.875,
+      "rewards/repetition_penalty_reward": -0.01090173190459609,
+      "rewards/tag_count_reward": 0.8229166865348816,
+      "step": 1090
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 752.1458435058594,
+      "epoch": 0.5455,
+      "grad_norm": 6.561919666240053,
+      "kl": 1.169921875,
+      "learning_rate": 5.570682927903193e-07,
+      "loss": 0.4967,
+      "reward": 2.3344783782958984,
+      "reward_std": 0.5779477655887604,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.01621603313833475,
+      "rewards/tag_count_reward": 0.8229166865348816,
+      "step": 1091
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 620.2708435058594,
+      "epoch": 0.546,
+      "grad_norm": 8.795113051800255,
+      "kl": 1.126953125,
+      "learning_rate": 5.562829811526154e-07,
+      "loss": 0.6867,
+      "reward": 2.5719066858291626,
+      "reward_std": 0.6096862554550171,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.02184334397315979,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1092
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 448.1458435058594,
+      "epoch": 0.5465,
+      "grad_norm": 5.044503857903457,
+      "kl": 0.6328125,
+      "learning_rate": 5.554976503758612e-07,
+      "loss": 0.7076,
+      "reward": 2.7934606075286865,
+      "reward_std": 0.3946908265352249,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.017303182743489742,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1093
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 521.8541870117188,
+      "epoch": 0.547,
+      "grad_norm": 11.91339929157061,
+      "kl": 0.794921875,
+      "learning_rate": 5.547123028523106e-07,
+      "loss": 0.4694,
+      "reward": 2.475094199180603,
+      "reward_std": 0.4313882440328598,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.028378095477819443,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1094
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 608.2291717529297,
+      "epoch": 0.5475,
+      "grad_norm": 21.561567034341035,
+      "kl": 0.744140625,
+      "learning_rate": 5.539269409742683e-07,
+      "loss": 0.8278,
+      "reward": 2.4341979026794434,
+      "reward_std": 0.6945553719997406,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.024135553278028965,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1095
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 534.8750305175781,
+      "epoch": 0.548,
+      "grad_norm": 25.643096112271426,
+      "kl": 1.119140625,
+      "learning_rate": 5.531415671340826e-07,
+      "loss": 0.5556,
+      "reward": 2.2709327936172485,
+      "reward_std": 0.5148254334926605,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.02594222454354167,
+      "rewards/tag_count_reward": 0.8593750298023224,
+      "step": 1096
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 337.12501525878906,
+      "epoch": 0.5485,
+      "grad_norm": 15.402452878812943,
+      "kl": 0.4892578125,
+      "learning_rate": 5.523561837241387e-07,
+      "loss": 0.2395,
+      "reward": 2.4139556884765625,
+      "reward_std": 0.25268905609846115,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.018336026929318905,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1097
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 396.93751525878906,
+      "epoch": 0.549,
+      "grad_norm": 20.68700015444551,
+      "kl": 0.609375,
+      "learning_rate": 5.515707931368507e-07,
+      "loss": 0.3136,
+      "reward": 2.6157283782958984,
+      "reward_std": 0.503062829375267,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.01968833664432168,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1098
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 549.8333435058594,
+      "epoch": 0.5495,
+      "grad_norm": 41.64035805095098,
+      "kl": 0.728515625,
+      "learning_rate": 5.507853977646543e-07,
+      "loss": 0.6883,
+      "reward": 2.3382461071014404,
+      "reward_std": 0.7803103625774384,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.017656567506492138,
+      "rewards/tag_count_reward": 0.8489583432674408,
+      "step": 1099
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 434.16668701171875,
+      "epoch": 0.55,
+      "grad_norm": 42.17393716054865,
+      "kl": 0.666015625,
+      "learning_rate": 5.5e-07,
+      "loss": 0.5129,
+      "reward": 2.6097629070281982,
+      "reward_std": 0.6973889470100403,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.023917713202536106,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1100
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 546.1875152587891,
+      "epoch": 0.5505,
+      "grad_norm": 19.4698112016211,
+      "kl": 0.84375,
+      "learning_rate": 5.492146022353459e-07,
+      "loss": 0.3103,
+      "reward": 2.297290325164795,
+      "reward_std": 0.6591008305549622,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.02041808795183897,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1101
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 500.8958435058594,
+      "epoch": 0.551,
+      "grad_norm": 27.46742900229019,
+      "kl": 0.751953125,
+      "learning_rate": 5.484292068631494e-07,
+      "loss": 0.9024,
+      "reward": 2.3236552476882935,
+      "reward_std": 0.5587631464004517,
+      "rewards/accuracy_reward": 0.4583333358168602,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.01662263460457325,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1102
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 422.1041717529297,
+      "epoch": 0.5515,
+      "grad_norm": 6.678322908726758,
+      "kl": 0.634765625,
+      "learning_rate": 5.476438162758611e-07,
+      "loss": 0.3346,
+      "reward": 2.5052073001861572,
+      "reward_std": 0.44907619804143906,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.01909833773970604,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1103
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 395.1458435058594,
+      "epoch": 0.552,
+      "grad_norm": 7.97609070470728,
+      "kl": 0.7421875,
+      "learning_rate": 5.468584328659172e-07,
+      "loss": 0.3472,
+      "reward": 2.4966301918029785,
+      "reward_std": 0.3519609868526459,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.018995044752955437,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1104
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 498.5000305175781,
+      "epoch": 0.5525,
+      "grad_norm": 8.65361833333178,
+      "kl": 0.7255859375,
+      "learning_rate": 5.460730590257317e-07,
+      "loss": 0.4795,
+      "reward": 2.138282537460327,
+      "reward_std": 0.5641656816005707,
+      "rewards/accuracy_reward": 0.2916666716337204,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.017967475578188896,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1105
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 457.9583435058594,
+      "epoch": 0.553,
+      "grad_norm": 12.3661318307847,
+      "kl": 0.677734375,
+      "learning_rate": 5.452876971476896e-07,
+      "loss": 0.7296,
+      "reward": 2.612514019012451,
+      "reward_std": 0.6336883902549744,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.021166563034057617,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1106
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 515.6250305175781,
+      "epoch": 0.5535,
+      "grad_norm": 10.944643674304801,
+      "kl": 0.74609375,
+      "learning_rate": 5.445023496241388e-07,
+      "loss": 0.8027,
+      "reward": 2.3295055627822876,
+      "reward_std": 0.5938288271427155,
+      "rewards/accuracy_reward": 0.5000000111758709,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.01598059432581067,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1107
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 544.2708435058594,
+      "epoch": 0.554,
+      "grad_norm": 6.272584778246144,
+      "kl": 1.17578125,
+      "learning_rate": 5.437170188473847e-07,
+      "loss": 0.5575,
+      "reward": 2.0059564113616943,
+      "reward_std": 0.6833735406398773,
+      "rewards/accuracy_reward": 0.2708333432674408,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.01661309227347374,
+      "rewards/tag_count_reward": 0.8489583432674408,
+      "step": 1108
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 388.2916717529297,
+      "epoch": 0.5545,
+      "grad_norm": 6.425896412413649,
+      "kl": 0.81640625,
+      "learning_rate": 5.429317072096807e-07,
+      "loss": 0.3538,
+      "reward": 2.409471869468689,
+      "reward_std": 0.5708480477333069,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.017611466348171234,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1109
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 555.4375,
+      "epoch": 0.555,
+      "grad_norm": 8.904743072445507,
+      "kl": 1.025390625,
+      "learning_rate": 5.421464171032224e-07,
+      "loss": 0.8839,
+      "reward": 2.551916480064392,
+      "reward_std": 0.8339214324951172,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.012319804634898901,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1110
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 421.37501525878906,
+      "epoch": 0.5555,
+      "grad_norm": 9.390143696400399,
+      "kl": 0.7890625,
+      "learning_rate": 5.413611509201396e-07,
+      "loss": 0.6891,
+      "reward": 2.6331262588500977,
+      "reward_std": 0.6708503365516663,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.04395711608231068,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1111
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 566.3333740234375,
+      "epoch": 0.556,
+      "grad_norm": 7.055072585098017,
+      "kl": 1.05078125,
+      "learning_rate": 5.405759110524894e-07,
+      "loss": 0.688,
+      "reward": 2.20969557762146,
+      "reward_std": 0.6133010685443878,
+      "rewards/accuracy_reward": 0.3750000223517418,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.029887686483561993,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1112
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 460.81251525878906,
+      "epoch": 0.5565,
+      "grad_norm": 6.681017604333264,
+      "kl": 0.806640625,
+      "learning_rate": 5.397906998922483e-07,
+      "loss": 0.4867,
+      "reward": 2.5056605339050293,
+      "reward_std": 0.42460909485816956,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.022117381915450096,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1113
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 502.1666717529297,
+      "epoch": 0.557,
+      "grad_norm": 7.9919915253075136,
+      "kl": 0.951171875,
+      "learning_rate": 5.390055198313061e-07,
+      "loss": 0.8675,
+      "reward": 2.357274293899536,
+      "reward_std": 0.6162551641464233,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.015989708248525858,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1114
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 457.6250305175781,
+      "epoch": 0.5575,
+      "grad_norm": 6.025174035853018,
+      "kl": 0.744140625,
+      "learning_rate": 5.382203732614571e-07,
+      "loss": 0.5821,
+      "reward": 2.6463170051574707,
+      "reward_std": 0.5895664393901825,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.015141477808356285,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1115
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 648.9791870117188,
+      "epoch": 0.558,
+      "grad_norm": 4.919326916823641,
+      "kl": 1.15234375,
+      "learning_rate": 5.37435262574394e-07,
+      "loss": 0.6887,
+      "reward": 2.228627920150757,
+      "reward_std": 0.6992431879043579,
+      "rewards/accuracy_reward": 0.4375000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.017899959348142147,
+      "rewards/tag_count_reward": 0.8437500298023224,
+      "step": 1116
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 543.4375,
+      "epoch": 0.5585,
+      "grad_norm": 6.777349325411678,
+      "kl": 0.990234375,
+      "learning_rate": 5.366501901617001e-07,
+      "loss": 0.7701,
+      "reward": 2.5158661603927612,
+      "reward_std": 0.6982110738754272,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.902777761220932,
+      "rewards/repetition_penalty_reward": -0.011911679524928331,
+      "rewards/tag_count_reward": 0.8958333730697632,
+      "step": 1117
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 367.2708435058594,
+      "epoch": 0.559,
+      "grad_norm": 4.625628969961506,
+      "kl": 0.611328125,
+      "learning_rate": 5.358651584148423e-07,
+      "loss": 0.4419,
+      "reward": 2.7811131477355957,
+      "reward_std": 0.3881167322397232,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04006761498749256,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1118
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 481.58335876464844,
+      "epoch": 0.5595,
+      "grad_norm": 5.8630513324216755,
+      "kl": 0.779296875,
+      "learning_rate": 5.350801697251633e-07,
+      "loss": 0.5938,
+      "reward": 2.6719554662704468,
+      "reward_std": 0.5751761198043823,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.03116954304277897,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1119
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 393.0833435058594,
+      "epoch": 0.56,
+      "grad_norm": 5.093420961702017,
+      "kl": 0.5625,
+      "learning_rate": 5.342952264838747e-07,
+      "loss": 0.4989,
+      "reward": 2.689083218574524,
+      "reward_std": 0.388533353805542,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.01230586925521493,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1120
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 365.7083435058594,
+      "epoch": 0.5605,
+      "grad_norm": 5.605297281726827,
+      "kl": 0.5947265625,
+      "learning_rate": 5.335103310820496e-07,
+      "loss": 0.5772,
+      "reward": 2.8223936557769775,
+      "reward_std": 0.38263121247291565,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.017884175293147564,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1121
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 601.6666870117188,
+      "epoch": 0.561,
+      "grad_norm": 6.832349813903946,
+      "kl": 0.763671875,
+      "learning_rate": 5.32725485910616e-07,
+      "loss": 0.8745,
+      "reward": 2.4064905643463135,
+      "reward_std": 0.7045398056507111,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.01885672379285097,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1122
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 450.0833435058594,
+      "epoch": 0.5615,
+      "grad_norm": 4.514039507778649,
+      "kl": 0.666015625,
+      "learning_rate": 5.319406933603482e-07,
+      "loss": 0.5151,
+      "reward": 2.757925033569336,
+      "reward_std": 0.3729214509949088,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.0250612860545516,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1123
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 649.3541870117188,
+      "epoch": 0.562,
+      "grad_norm": 7.659959107724474,
+      "kl": 0.943359375,
+      "learning_rate": 5.311559558218603e-07,
+      "loss": 0.6994,
+      "reward": 2.274022102355957,
+      "reward_std": 0.7414794862270355,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9513888657093048,
+      "rewards/repetition_penalty_reward": -0.03153356537222862,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1124
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 671.4166870117188,
+      "epoch": 0.5625,
+      "grad_norm": 6.923021149158682,
+      "kl": 0.923828125,
+      "learning_rate": 5.303712756855988e-07,
+      "loss": 0.8844,
+      "reward": 2.3472338914871216,
+      "reward_std": 0.7413772344589233,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9305556416511536,
+      "rewards/repetition_penalty_reward": -0.015613417141139507,
+      "rewards/tag_count_reward": 0.828125,
+      "step": 1125
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 338.54168701171875,
+      "epoch": 0.563,
+      "grad_norm": 6.103986634370641,
+      "kl": 0.55078125,
+      "learning_rate": 5.295866553418358e-07,
+      "loss": 0.5005,
+      "reward": 2.7973004579544067,
+      "reward_std": 0.39559805393218994,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.03256077412515879,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1126
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 624.5000152587891,
+      "epoch": 0.5635,
+      "grad_norm": 6.343512742022832,
+      "kl": 0.99609375,
+      "learning_rate": 5.288020971806608e-07,
+      "loss": 0.4152,
+      "reward": 2.39252108335495,
+      "reward_std": 0.40336638130247593,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.015465098433196545,
+      "rewards/tag_count_reward": 0.8385416865348816,
+      "step": 1127
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 343.5833435058594,
+      "epoch": 0.564,
+      "grad_norm": 24.603696739447937,
+      "kl": 0.9208984375,
+      "learning_rate": 5.28017603591974e-07,
+      "loss": 0.3151,
+      "reward": 2.6542476415634155,
+      "reward_std": 0.28428603522479534,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.012418974190950394,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1128
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 271.2291717529297,
+      "epoch": 0.5645,
+      "grad_norm": 5.473040679290024,
+      "kl": 0.490234375,
+      "learning_rate": 5.27233176965479e-07,
+      "loss": 0.0469,
+      "reward": 2.9089274406433105,
+      "reward_std": 0.12796843331307173,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.028572553768754005,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1129
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 626.6458435058594,
+      "epoch": 0.565,
+      "grad_norm": 13.844398470198444,
+      "kl": 0.90625,
+      "learning_rate": 5.264488196906752e-07,
+      "loss": 0.7776,
+      "reward": 2.4067180156707764,
+      "reward_std": 0.6049718260765076,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.020365470554679632,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1130
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 575.9583435058594,
+      "epoch": 0.5655,
+      "grad_norm": 47.74850593269307,
+      "kl": 1.37890625,
+      "learning_rate": 5.256645341568511e-07,
+      "loss": 0.3055,
+      "reward": 2.4448657035827637,
+      "reward_std": 0.2919907867908478,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.015203722519800067,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1131
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 457.9375305175781,
+      "epoch": 0.566,
+      "grad_norm": 16.058299463435173,
+      "kl": 0.744140625,
+      "learning_rate": 5.248803227530763e-07,
+      "loss": 0.7605,
+      "reward": 2.6384165287017822,
+      "reward_std": 0.5586017668247223,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.0195696409791708,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1132
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 282.8333435058594,
+      "epoch": 0.5665,
+      "grad_norm": 4.518324512537823,
+      "kl": 0.4765625,
+      "learning_rate": 5.240961878681947e-07,
+      "loss": 0.0132,
+      "reward": 2.93279492855072,
+      "reward_std": 0.15127281844615936,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.025538288988173008,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1133
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 480.04168701171875,
+      "epoch": 0.567,
+      "grad_norm": 11.667992967616525,
+      "kl": 0.814453125,
+      "learning_rate": 5.233121318908173e-07,
+      "loss": 0.4719,
+      "reward": 2.2194743156433105,
+      "reward_std": 0.4004889130592346,
+      "rewards/accuracy_reward": 0.3750000223517418,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.018372977152466774,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1134
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 400.4583435058594,
+      "epoch": 0.5675,
+      "grad_norm": 6.73610126947889,
+      "kl": 0.7626953125,
+      "learning_rate": 5.225281572093143e-07,
+      "loss": 0.2514,
+      "reward": 2.7708280086517334,
+      "reward_std": 0.23756166687235236,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03299156203866005,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1135
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 791.6875305175781,
+      "epoch": 0.568,
+      "grad_norm": 25.279730322882415,
+      "kl": 1.59765625,
+      "learning_rate": 5.21744266211809e-07,
+      "loss": 0.8005,
+      "reward": 2.3063154816627502,
+      "reward_std": 0.6179798245429993,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.016601404175162315,
+      "rewards/tag_count_reward": 0.78125,
+      "step": 1136
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 466.16668701171875,
+      "epoch": 0.5685,
+      "grad_norm": 18.939649860781724,
+      "kl": 0.869140625,
+      "learning_rate": 5.20960461286169e-07,
+      "loss": 0.6348,
+      "reward": 2.7132285833358765,
+      "reward_std": 0.41872841119766235,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02635475154966116,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1137
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 704.2500305175781,
+      "epoch": 0.569,
+      "grad_norm": 23.295755447219996,
+      "kl": 1.234375,
+      "learning_rate": 5.2017674482e-07,
+      "loss": 0.7717,
+      "reward": 2.3411978483200073,
+      "reward_std": 0.7033064663410187,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.019913168624043465,
+      "rewards/tag_count_reward": 0.8333333432674408,
+      "step": 1138
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 512.9166717529297,
+      "epoch": 0.5695,
+      "grad_norm": 14.21734574883039,
+      "kl": 1.03515625,
+      "learning_rate": 5.193931192006385e-07,
+      "loss": 0.5358,
+      "reward": 2.592092990875244,
+      "reward_std": 0.6145272552967072,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.02249028254300356,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1139
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 394.45835876464844,
+      "epoch": 0.57,
+      "grad_norm": 6.405748026127363,
+      "kl": 0.642578125,
+      "learning_rate": 5.186095868151436e-07,
+      "loss": 0.3287,
+      "reward": 2.8148388862609863,
+      "reward_std": 0.3306480962783098,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03238342609256506,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1140
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 643.0208435058594,
+      "epoch": 0.5705,
+      "grad_norm": 11.8102138105351,
+      "kl": 1.228515625,
+      "learning_rate": 5.178261500502912e-07,
+      "loss": 0.5072,
+      "reward": 2.575770854949951,
+      "reward_std": 0.3936486691236496,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.014507037587463856,
+      "rewards/tag_count_reward": 0.8333333432674408,
+      "step": 1141
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 484.1875,
+      "epoch": 0.571,
+      "grad_norm": 19.62442109664474,
+      "kl": 1.203125,
+      "learning_rate": 5.170428112925659e-07,
+      "loss": 0.6059,
+      "reward": 2.3984756469726562,
+      "reward_std": 0.5576977431774139,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.02860766276717186,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1142
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 339.9583435058594,
+      "epoch": 0.5715,
+      "grad_norm": 15.224539776880427,
+      "kl": 0.6611328125,
+      "learning_rate": 5.162595729281526e-07,
+      "loss": 0.3318,
+      "reward": 2.6164965629577637,
+      "reward_std": 0.2758345529437065,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.018920221365988255,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1143
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 400.4583435058594,
+      "epoch": 0.572,
+      "grad_norm": 16.037796811858144,
+      "kl": 1.94921875,
+      "learning_rate": 5.154764373429315e-07,
+      "loss": 0.5664,
+      "reward": 2.7207865715026855,
+      "reward_std": 0.4415482133626938,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.024005182087421417,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1144
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 488.3750305175781,
+      "epoch": 0.5725,
+      "grad_norm": 89.73106922514725,
+      "kl": 7.046875,
+      "learning_rate": 5.146934069224698e-07,
+      "loss": 0.8408,
+      "reward": 2.387607216835022,
+      "reward_std": 0.6802188009023666,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.016906835604459047,
+      "rewards/tag_count_reward": 0.7864583432674408,
+      "step": 1145
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 334.6041717529297,
+      "epoch": 0.573,
+      "grad_norm": 32.93197336588902,
+      "kl": 2.5234375,
+      "learning_rate": 5.139104840520135e-07,
+      "loss": 0.4021,
+      "reward": 2.706830859184265,
+      "reward_std": 0.45073381066322327,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.017127559520304203,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1146
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.18751525878906,
+      "epoch": 0.5735,
+      "grad_norm": 16.862260635413147,
+      "kl": 1.87109375,
+      "learning_rate": 5.131276711164815e-07,
+      "loss": 0.2396,
+      "reward": 2.566492795944214,
+      "reward_std": 0.4739261567592621,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.028993389569222927,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1147
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.3333435058594,
+      "epoch": 0.574,
+      "grad_norm": 34.79221151751134,
+      "kl": 3.4921875,
+      "learning_rate": 5.123449705004581e-07,
+      "loss": 0.3237,
+      "reward": 2.6664516925811768,
+      "reward_std": 0.6325987875461578,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02625663299113512,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1148
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 253.3541717529297,
+      "epoch": 0.5745,
+      "grad_norm": 12.048802514250214,
+      "kl": 0.947265625,
+      "learning_rate": 5.115623845881847e-07,
+      "loss": 0.0768,
+      "reward": 2.8428802490234375,
+      "reward_std": 0.40082596242427826,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.019967169500887394,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1149
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 302.7708435058594,
+      "epoch": 0.575,
+      "grad_norm": 10.82498445900699,
+      "kl": 0.498046875,
+      "learning_rate": 5.107799157635538e-07,
+      "loss": 0.0567,
+      "reward": 2.6700661182403564,
+      "reward_std": 0.20545261912047863,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02264238614588976,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1150
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 256.1041717529297,
+      "epoch": 0.5755,
+      "grad_norm": 47.5496817156261,
+      "kl": 4.015625,
+      "learning_rate": 5.099975664101014e-07,
+      "loss": 0.1526,
+      "reward": 1.90774405002594,
+      "reward_std": 0.7336077690124512,
+      "rewards/accuracy_reward": 0.2708333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000894069672,
+      "rewards/repetition_penalty_reward": -0.014130960684269667,
+      "rewards/tag_count_reward": 0.7135416865348816,
+      "step": 1151
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 276.0833435058594,
+      "epoch": 0.576,
+      "grad_norm": 11.1836621188091,
+      "kl": 1.27734375,
+      "learning_rate": 5.09215338910999e-07,
+      "loss": 0.062,
+      "reward": 2.778138041496277,
+      "reward_std": 0.30686675012111664,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02915370836853981,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1152
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.47918701171875,
+      "epoch": 0.5765,
+      "grad_norm": 8.241389663503151,
+      "kl": 0.83984375,
+      "learning_rate": 5.084332356490472e-07,
+      "loss": -0.0037,
+      "reward": 2.4805880784988403,
+      "reward_std": 0.40117160230875015,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.043717604130506516,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1153
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 253.1666717529297,
+      "epoch": 0.577,
+      "grad_norm": 65.55942523736746,
+      "kl": 6.75,
+      "learning_rate": 5.076512590066685e-07,
+      "loss": 0.0028,
+      "reward": 2.0602548122406006,
+      "reward_std": 0.6848001033067703,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.01960639003664255,
+      "rewards/tag_count_reward": 0.6979166865348816,
+      "step": 1154
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.9583435058594,
+      "epoch": 0.5775,
+      "grad_norm": 57.13616985785417,
+      "kl": 4.9140625,
+      "learning_rate": 5.068694113658992e-07,
+      "loss": -0.0814,
+      "reward": 2.022095203399658,
+      "reward_std": 0.7005333751440048,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.8680555820465088,
+      "rewards/repetition_penalty_reward": -0.023043738678097725,
+      "rewards/tag_count_reward": 0.6979166865348816,
+      "step": 1155
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 255.75000762939453,
+      "epoch": 0.578,
+      "grad_norm": 5.38110971277495,
+      "kl": 1.228515625,
+      "learning_rate": 5.060876951083828e-07,
+      "loss": -0.0316,
+      "reward": 2.7976086139678955,
+      "reward_std": 0.3995959609746933,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.01662743976339698,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1156
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 295.50001525878906,
+      "epoch": 0.5785,
+      "grad_norm": 50.70936605802299,
+      "kl": 3.1552734375,
+      "learning_rate": 5.053061126153624e-07,
+      "loss": 0.0544,
+      "reward": 2.074872672557831,
+      "reward_std": 0.20399940758943558,
+      "rewards/accuracy_reward": 0.25,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.017141404328867793,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1157
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 294.4791717529297,
+      "epoch": 0.579,
+      "grad_norm": 64.52110252352004,
+      "kl": 2.74609375,
+      "learning_rate": 5.045246662676741e-07,
+      "loss": 0.0535,
+      "reward": 2.5658187866210938,
+      "reward_std": 0.34509243071079254,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.036611984483897686,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1158
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 258.00000762939453,
+      "epoch": 0.5795,
+      "grad_norm": 15.605596683642418,
+      "kl": 1.49609375,
+      "learning_rate": 5.037433584457389e-07,
+      "loss": -0.0383,
+      "reward": 2.314841389656067,
+      "reward_std": 0.7232391238212585,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.015019847080111504,
+      "rewards/tag_count_reward": 0.78125,
+      "step": 1159
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 256.02083587646484,
+      "epoch": 0.58,
+      "grad_norm": 12.22918579768311,
+      "kl": 0.7314453125,
+      "learning_rate": 5.02962191529556e-07,
+      "loss": -0.0099,
+      "reward": 2.601512908935547,
+      "reward_std": 0.5186595022678375,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.013070614542812109,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1160
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.2708435058594,
+      "epoch": 0.5805,
+      "grad_norm": 21.869574669814327,
+      "kl": 2.5107421875,
+      "learning_rate": 5.021811678986951e-07,
+      "loss": 0.0476,
+      "reward": 2.564257025718689,
+      "reward_std": 0.4466948760673404,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9444444179534912,
+      "rewards/repetition_penalty_reward": -0.015604046639055014,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1161
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.12501525878906,
+      "epoch": 0.581,
+      "grad_norm": 8.780241833303409,
+      "kl": 1.43359375,
+      "learning_rate": 5.014002899322896e-07,
+      "loss": -0.0048,
+      "reward": 2.296339511871338,
+      "reward_std": 0.45754382014274597,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.026577199809253216,
+      "rewards/tag_count_reward": 0.8020833432674408,
+      "step": 1162
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 300.7708435058594,
+      "epoch": 0.5815,
+      "grad_norm": 21.9309323286652,
+      "kl": 2.3203125,
+      "learning_rate": 5.006195600090296e-07,
+      "loss": -0.0216,
+      "reward": 2.414031744003296,
+      "reward_std": 0.7338520288467407,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.02520434372127056,
+      "rewards/tag_count_reward": 0.828125,
+      "step": 1163
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 228.08333587646484,
+      "epoch": 0.582,
+      "grad_norm": 28.587099800302813,
+      "kl": 3.21875,
+      "learning_rate": 4.998389805071536e-07,
+      "loss": 0.0103,
+      "reward": 2.427722692489624,
+      "reward_std": 0.7513796091079712,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.006305098533630371,
+      "rewards/tag_count_reward": 0.7812500298023224,
+      "step": 1164
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 266.7708435058594,
+      "epoch": 0.5825,
+      "grad_norm": 19.590917470188238,
+      "kl": 2.234375,
+      "learning_rate": 4.990585538044419e-07,
+      "loss": 0.0747,
+      "reward": 2.6480218172073364,
+      "reward_std": 0.6686672568321228,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.013436626642942429,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1165
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 253.2916717529297,
+      "epoch": 0.583,
+      "grad_norm": 20.553956833130847,
+      "kl": 2.9140625,
+      "learning_rate": 4.982782822782101e-07,
+      "loss": 0.0401,
+      "reward": 1.9601789712905884,
+      "reward_std": 0.4777047038078308,
+      "rewards/accuracy_reward": 0.375,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.013779422268271446,
+      "rewards/tag_count_reward": 0.6614583432674408,
+      "step": 1166
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.50001525878906,
+      "epoch": 0.5835,
+      "grad_norm": 89.06303811158742,
+      "kl": 3.669921875,
+      "learning_rate": 4.974981683053001e-07,
+      "loss": 0.1299,
+      "reward": 2.885384678840637,
+      "reward_std": 0.34253598749637604,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.015657078940421343,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1167
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.3958435058594,
+      "epoch": 0.584,
+      "grad_norm": 13.515172514998415,
+      "kl": 0.8046875,
+      "learning_rate": 4.967182142620745e-07,
+      "loss": 0.0289,
+      "reward": 2.4830256700515747,
+      "reward_std": 0.39654435217380524,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.022182647138834,
+      "rewards/tag_count_reward": 0.859375,
+      "step": 1168
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 267.4583435058594,
+      "epoch": 0.5845,
+      "grad_norm": 16.061739704037873,
+      "kl": 1.3359375,
+      "learning_rate": 4.959384225244087e-07,
+      "loss": 0.0793,
+      "reward": 1.9616607427597046,
+      "reward_std": 0.44488397240638733,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.017505998257547617,
+      "rewards/tag_count_reward": 0.6458333432674408,
+      "step": 1169
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 269.1666717529297,
+      "epoch": 0.585,
+      "grad_norm": 7.397678949657141,
+      "kl": 0.7978515625,
+      "learning_rate": 4.951587954676837e-07,
+      "loss": 0.0111,
+      "reward": 2.502622127532959,
+      "reward_std": 0.2668640099000186,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.01126689650118351,
+      "rewards/tag_count_reward": 0.8333333432674408,
+      "step": 1170
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 226.93750762939453,
+      "epoch": 0.5855,
+      "grad_norm": 16.086138152776787,
+      "kl": 1.734375,
+      "learning_rate": 4.943793354667783e-07,
+      "loss": 0.0168,
+      "reward": 2.2989531755447388,
+      "reward_std": 0.43994690477848053,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9236112236976624,
+      "rewards/repetition_penalty_reward": -0.015283002983778715,
+      "rewards/tag_count_reward": 0.7864583432674408,
+      "step": 1171
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 282.8333435058594,
+      "epoch": 0.586,
+      "grad_norm": 11.885572960330673,
+      "kl": 1.953125,
+      "learning_rate": 4.93600044896063e-07,
+      "loss": 0.0119,
+      "reward": 2.477262020111084,
+      "reward_std": 0.5187919661402702,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.022738128900527954,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1172
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.91668701171875,
+      "epoch": 0.5865,
+      "grad_norm": 188.62778027268504,
+      "kl": 9.859375,
+      "learning_rate": 4.928209261293923e-07,
+      "loss": 0.3053,
+      "reward": 2.367971897125244,
+      "reward_std": 0.6133864223957062,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.013972645625472069,
+      "rewards/tag_count_reward": 0.7916666865348816,
+      "step": 1173
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 248.85417938232422,
+      "epoch": 0.587,
+      "grad_norm": 541.9716559112582,
+      "kl": 18.40625,
+      "learning_rate": 4.920419815400968e-07,
+      "loss": 0.6135,
+      "reward": 2.119057536125183,
+      "reward_std": 0.6997096538543701,
+      "rewards/accuracy_reward": 0.4375000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.01983139617368579,
+      "rewards/tag_count_reward": 0.7291666865348816,
+      "step": 1174
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 232.4166717529297,
+      "epoch": 0.5875,
+      "grad_norm": 47.221216815875486,
+      "kl": 4.671875,
+      "learning_rate": 4.912632135009769e-07,
+      "loss": 0.0897,
+      "reward": 1.7002249956130981,
+      "reward_std": 0.7154510319232941,
+      "rewards/accuracy_reward": 0.2708333432674408,
+      "rewards/reasoning_steps_reward": 0.8958334028720856,
+      "rewards/repetition_penalty_reward": -0.008108395617455244,
+      "rewards/tag_count_reward": 0.5416666865348816,
+      "step": 1175
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 248.64583587646484,
+      "epoch": 0.588,
+      "grad_norm": 620.752818719277,
+      "kl": 5.140625,
+      "learning_rate": 4.904846243842949e-07,
+      "loss": 0.109,
+      "reward": 1.9387726187705994,
+      "reward_std": 0.8643776774406433,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.017824689392000437,
+      "rewards/tag_count_reward": 0.6093750149011612,
+      "step": 1176
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.43750762939453,
+      "epoch": 0.5885,
+      "grad_norm": 12.885096090616047,
+      "kl": 1.40234375,
+      "learning_rate": 4.897062165617686e-07,
+      "loss": -0.0543,
+      "reward": 1.8845123052597046,
+      "reward_std": 0.6029966920614243,
+      "rewards/accuracy_reward": 0.33333333395421505,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.016529476270079613,
+      "rewards/tag_count_reward": 0.609375,
+      "step": 1177
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 239.1666717529297,
+      "epoch": 0.589,
+      "grad_norm": 10.235870138091864,
+      "kl": 1.046875,
+      "learning_rate": 4.88927992404563e-07,
+      "loss": -0.055,
+      "reward": 2.0582520961761475,
+      "reward_std": 0.688510000705719,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.011192373465746641,
+      "rewards/tag_count_reward": 0.6875000298023224,
+      "step": 1178
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 277.9166717529297,
+      "epoch": 0.5895,
+      "grad_norm": 15.676254056984863,
+      "kl": 1.47265625,
+      "learning_rate": 4.881499542832841e-07,
+      "loss": -0.0441,
+      "reward": 1.829641878604889,
+      "reward_std": 0.8753339052200317,
+      "rewards/accuracy_reward": 0.3541666865348816,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.01237204298377037,
+      "rewards/tag_count_reward": 0.5781250298023224,
+      "step": 1179
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 254.35417938232422,
+      "epoch": 0.59,
+      "grad_norm": 15.872841179430333,
+      "kl": 1.21484375,
+      "learning_rate": 4.873721045679706e-07,
+      "loss": -0.0526,
+      "reward": 1.8742225170135498,
+      "reward_std": 0.628957211971283,
+      "rewards/accuracy_reward": 0.2500000111758709,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.009458072949200869,
+      "rewards/tag_count_reward": 0.7447916865348816,
+      "step": 1180
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 272.5208435058594,
+      "epoch": 0.5905,
+      "grad_norm": 17.36437993847142,
+      "kl": 2.4609375,
+      "learning_rate": 4.865944456280878e-07,
+      "loss": 0.0072,
+      "reward": 1.6619837880134583,
+      "reward_std": 0.573122188448906,
+      "rewards/accuracy_reward": 0.0833333358168602,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.00989127904176712,
+      "rewards/tag_count_reward": 0.6510416865348816,
+      "step": 1181
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.0625,
+      "epoch": 0.591,
+      "grad_norm": 18.143037794685533,
+      "kl": 3.7890625,
+      "learning_rate": 4.858169798325198e-07,
+      "loss": 0.0575,
+      "reward": 1.7578097581863403,
+      "reward_std": 0.9271320402622223,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.75,
+      "rewards/repetition_penalty_reward": -0.013023747596889734,
+      "rewards/tag_count_reward": 0.6250000149011612,
+      "step": 1182
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.2916717529297,
+      "epoch": 0.5915,
+      "grad_norm": 12.26718270823858,
+      "kl": 1.2421875,
+      "learning_rate": 4.850397095495621e-07,
+      "loss": -0.0167,
+      "reward": 2.673969864845276,
+      "reward_std": 0.5497699528932571,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.025683030486106873,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1183
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.2291717529297,
+      "epoch": 0.592,
+      "grad_norm": 10.94115924929318,
+      "kl": 2.15234375,
+      "learning_rate": 4.842626371469149e-07,
+      "loss": -0.0423,
+      "reward": 2.5248043537139893,
+      "reward_std": 0.6554135978221893,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.02033464703708887,
+      "rewards/tag_count_reward": 0.8645833730697632,
+      "step": 1184
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.81251525878906,
+      "epoch": 0.5925,
+      "grad_norm": 19.120132180785248,
+      "kl": 2.685546875,
+      "learning_rate": 4.834857649916752e-07,
+      "loss": 0.0412,
+      "reward": 2.3415642976760864,
+      "reward_std": 0.4712478220462799,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.033435771241784096,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1185
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.60418701171875,
+      "epoch": 0.593,
+      "grad_norm": 29.326687780779196,
+      "kl": 2.50390625,
+      "learning_rate": 4.827090954503308e-07,
+      "loss": -0.0046,
+      "reward": 2.505213499069214,
+      "reward_std": 0.5307941734790802,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02777272555977106,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1186
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.37500762939453,
+      "epoch": 0.5935,
+      "grad_norm": 89.56476050571975,
+      "kl": 8.875,
+      "learning_rate": 4.819326308887513e-07,
+      "loss": 0.2113,
+      "reward": 1.9479205012321472,
+      "reward_std": 0.5512201189994812,
+      "rewards/accuracy_reward": 0.2916666716337204,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.01735728792846203,
+      "rewards/tag_count_reward": 0.7500000298023224,
+      "step": 1187
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 282.0833435058594,
+      "epoch": 0.594,
+      "grad_norm": 15.951627791356907,
+      "kl": 3.03125,
+      "learning_rate": 4.811563736721829e-07,
+      "loss": -0.0638,
+      "reward": 2.4049184322357178,
+      "reward_std": 0.6997295022010803,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.018692869693040848,
+      "rewards/tag_count_reward": 0.8333333432674408,
+      "step": 1188
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 266.9583435058594,
+      "epoch": 0.5945,
+      "grad_norm": 20.534655927942946,
+      "kl": 1.666015625,
+      "learning_rate": 4.803803261652395e-07,
+      "loss": 0.0088,
+      "reward": 2.855802297592163,
+      "reward_std": 0.3537350296974182,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.043503398075699806,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1189
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.6041717529297,
+      "epoch": 0.595,
+      "grad_norm": 6.707455759169878,
+      "kl": 0.607421875,
+      "learning_rate": 4.79604490731896e-07,
+      "loss": 0.0076,
+      "reward": 2.3484212160110474,
+      "reward_std": 0.3202339052222669,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.012690091505646706,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1190
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.3541717529297,
+      "epoch": 0.5955,
+      "grad_norm": 9.733669029070196,
+      "kl": 1.18359375,
+      "learning_rate": 4.788288697354824e-07,
+      "loss": -0.0888,
+      "reward": 2.4767041206359863,
+      "reward_std": 0.5791968405246735,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.026768106035888195,
+      "rewards/tag_count_reward": 0.8437500298023224,
+      "step": 1191
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.3125,
+      "epoch": 0.596,
+      "grad_norm": 5.518435395583406,
+      "kl": 0.6328125,
+      "learning_rate": 4.780534655386743e-07,
+      "loss": 0.0373,
+      "reward": 2.48011314868927,
+      "reward_std": 0.3792032450437546,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.030303513631224632,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1192
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 275.3333435058594,
+      "epoch": 0.5965,
+      "grad_norm": 10.222004650555824,
+      "kl": 0.916015625,
+      "learning_rate": 4.772782805034876e-07,
+      "loss": -0.0505,
+      "reward": 2.466397762298584,
+      "reward_std": 0.637960284948349,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03533835709095001,
+      "rewards/tag_count_reward": 0.8489583432674408,
+      "step": 1193
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 246.20834350585938,
+      "epoch": 0.597,
+      "grad_norm": 83.67864904271147,
+      "kl": 4.06640625,
+      "learning_rate": 4.7650331699127013e-07,
+      "loss": 0.021,
+      "reward": 2.561871290206909,
+      "reward_std": 0.4868089556694031,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.0197259820997715,
+      "rewards/tag_count_reward": 0.8385416865348816,
+      "step": 1194
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.2708435058594,
+      "epoch": 0.5975,
+      "grad_norm": 1778.754909364379,
+      "kl": 56.125,
+      "learning_rate": 4.75728577362695e-07,
+      "loss": 1.2038,
+      "reward": 2.206901729106903,
+      "reward_std": 0.6084506213665009,
+      "rewards/accuracy_reward": 0.5208333358168602,
+      "rewards/reasoning_steps_reward": 0.9791667461395264,
+      "rewards/repetition_penalty_reward": -0.02747342176735401,
+      "rewards/tag_count_reward": 0.734375,
+      "step": 1195
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.1458435058594,
+      "epoch": 0.598,
+      "grad_norm": 43.303276292552525,
+      "kl": 3.14453125,
+      "learning_rate": 4.749540639777539e-07,
+      "loss": 0.2523,
+      "reward": 2.6388306617736816,
+      "reward_std": 0.47564953565597534,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.013947171624749899,
+      "rewards/tag_count_reward": 0.8958333730697632,
+      "step": 1196
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 297.4375,
+      "epoch": 0.5985,
+      "grad_norm": 6.55423415450962,
+      "kl": 0.69921875,
+      "learning_rate": 4.741797791957489e-07,
+      "loss": -0.0043,
+      "reward": 2.5631258487701416,
+      "reward_std": 0.4409189820289612,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0254158116877079,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1197
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 286.5,
+      "epoch": 0.599,
+      "grad_norm": 79.15687028509065,
+      "kl": 5.171875,
+      "learning_rate": 4.7340572537528547e-07,
+      "loss": 0.1315,
+      "reward": 2.3954086303710938,
+      "reward_std": 0.5703263282775879,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.022994363214820623,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1198
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 274.0208435058594,
+      "epoch": 0.5995,
+      "grad_norm": 24.952831112368063,
+      "kl": 3.373046875,
+      "learning_rate": 4.7263190487426563e-07,
+      "loss": -0.0475,
+      "reward": 2.256582021713257,
+      "reward_std": 0.5435906499624252,
+      "rewards/accuracy_reward": 0.5000000223517418,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.028140274807810783,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1199
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 316.5208435058594,
+      "epoch": 0.6,
+      "grad_norm": 5.061254957850967,
+      "kl": 0.705078125,
+      "learning_rate": 4.7185832004988133e-07,
+      "loss": -0.0041,
+      "reward": 2.658898711204529,
+      "reward_std": 0.46046267449855804,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.03207360953092575,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1200
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 261.5416717529297,
+      "epoch": 0.6005,
+      "grad_norm": 35.65864208267965,
+      "kl": 4.21484375,
+      "learning_rate": 4.710849732586059e-07,
+      "loss": 0.0959,
+      "reward": 2.372064232826233,
+      "reward_std": 0.5162666738033295,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9444444179534912,
+      "rewards/repetition_penalty_reward": -0.015088722575455904,
+      "rewards/tag_count_reward": 0.8177083432674408,
+      "step": 1201
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 300.2708435058594,
+      "epoch": 0.601,
+      "grad_norm": 14.890784763265065,
+      "kl": 1.40234375,
+      "learning_rate": 4.703118668561875e-07,
+      "loss": 0.0295,
+      "reward": 2.602800130844116,
+      "reward_std": 0.5039637088775635,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03088043723255396,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1202
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 275.93751525878906,
+      "epoch": 0.6015,
+      "grad_norm": 24.01630825109177,
+      "kl": 2.609375,
+      "learning_rate": 4.6953900319764274e-07,
+      "loss": 0.0056,
+      "reward": 2.8198323249816895,
+      "reward_std": 0.4068307876586914,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.025654025375843048,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1203
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 311.0416717529297,
+      "epoch": 0.602,
+      "grad_norm": 48.98626731878835,
+      "kl": 4.4609375,
+      "learning_rate": 4.68766384637248e-07,
+      "loss": 0.0732,
+      "reward": 1.9985000491142273,
+      "reward_std": 0.7038697004318237,
+      "rewards/accuracy_reward": 0.354166679084301,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.022333373315632343,
+      "rewards/tag_count_reward": 0.75,
+      "step": 1204
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 308.41668701171875,
+      "epoch": 0.6025,
+      "grad_norm": 3.501918254202678,
+      "kl": 0.4033203125,
+      "learning_rate": 4.679940135285336e-07,
+      "loss": 0.0275,
+      "reward": 2.9462958574295044,
+      "reward_std": 0.0838532904163003,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03287102282047272,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1205
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 304.06251525878906,
+      "epoch": 0.603,
+      "grad_norm": 11.808559352326427,
+      "kl": 0.861328125,
+      "learning_rate": 4.672218922242759e-07,
+      "loss": 0.0187,
+      "reward": 2.636156678199768,
+      "reward_std": 0.512239083647728,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.011412853142246604,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1206
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 324.18751525878906,
+      "epoch": 0.6035,
+      "grad_norm": 9.023520837002595,
+      "kl": 0.73046875,
+      "learning_rate": 4.664500230764903e-07,
+      "loss": -0.1363,
+      "reward": 2.22309547662735,
+      "reward_std": 0.5313438028097153,
+      "rewards/accuracy_reward": 0.4166666679084301,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.026904682628810406,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1207
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 307.2916717529297,
+      "epoch": 0.604,
+      "grad_norm": 10.414640951029634,
+      "kl": 0.888671875,
+      "learning_rate": 4.656784084364238e-07,
+      "loss": -0.0566,
+      "reward": 2.740461230278015,
+      "reward_std": 0.5690730065107346,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.02516377530992031,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1208
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 336.6458435058594,
+      "epoch": 0.6045,
+      "grad_norm": 5.915397925387995,
+      "kl": 1.1640625,
+      "learning_rate": 4.6490705065454883e-07,
+      "loss": -0.099,
+      "reward": 2.537236452102661,
+      "reward_std": 0.5170433670282364,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.02179141901433468,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1209
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 318.18751525878906,
+      "epoch": 0.605,
+      "grad_norm": 18.20904524508889,
+      "kl": 2.44140625,
+      "learning_rate": 4.641359520805548e-07,
+      "loss": 0.0451,
+      "reward": 2.4042060375213623,
+      "reward_std": 0.44842204451560974,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.041974639520049095,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1210
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 377.29168701171875,
+      "epoch": 0.6055,
+      "grad_norm": 43.49583663170755,
+      "kl": 4.55859375,
+      "learning_rate": 4.6336511506334177e-07,
+      "loss": 0.0108,
+      "reward": 2.3035465478897095,
+      "reward_std": 0.5203736424446106,
+      "rewards/accuracy_reward": 0.4583333358168602,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.038467422127723694,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1211
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.9375,
+      "epoch": 0.606,
+      "grad_norm": 42.363711059687056,
+      "kl": 4.953125,
+      "learning_rate": 4.6259454195101267e-07,
+      "loss": 0.1363,
+      "reward": 2.4941481351852417,
+      "reward_std": 0.5663126707077026,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.023213034495711327,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1212
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.8541717529297,
+      "epoch": 0.6065,
+      "grad_norm": 53.23769285154482,
+      "kl": 6.1328125,
+      "learning_rate": 4.61824235090867e-07,
+      "loss": 0.2492,
+      "reward": 2.665413498878479,
+      "reward_std": 0.4333008825778961,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.018614262342453003,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1213
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.75001525878906,
+      "epoch": 0.607,
+      "grad_norm": 23.490826994445097,
+      "kl": 3.400390625,
+      "learning_rate": 4.6105419682939316e-07,
+      "loss": 0.0216,
+      "reward": 2.4767391681671143,
+      "reward_std": 0.47361528873443604,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.030205383896827698,
+      "rewards/tag_count_reward": 0.8958333730697632,
+      "step": 1214
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.75001525878906,
+      "epoch": 0.6075,
+      "grad_norm": 7.699965204481391,
+      "kl": 0.775390625,
+      "learning_rate": 4.602844295122613e-07,
+      "loss": 0.0431,
+      "reward": 2.93735134601593,
+      "reward_std": 0.11865681782364845,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03139885421842337,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1215
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 299.6666717529297,
+      "epoch": 0.608,
+      "grad_norm": 8.096084671606365,
+      "kl": 1.009765625,
+      "learning_rate": 4.59514935484316e-07,
+      "loss": 0.049,
+      "reward": 2.79358172416687,
+      "reward_std": 0.4377841055393219,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.024126553907990456,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1216
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 329.0416717529297,
+      "epoch": 0.6085,
+      "grad_norm": 14.138163667060162,
+      "kl": 1.8046875,
+      "learning_rate": 4.5874571708956953e-07,
+      "loss": -0.0943,
+      "reward": 2.6621501445770264,
+      "reward_std": 0.6885767579078674,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.037502871826291084,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1217
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 317.6041717529297,
+      "epoch": 0.609,
+      "grad_norm": 10.580306326635657,
+      "kl": 1.7734375,
+      "learning_rate": 4.579767766711944e-07,
+      "loss": 0.085,
+      "reward": 2.5517923831939697,
+      "reward_std": 0.4928157329559326,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.019388345535844564,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1218
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 301.25001525878906,
+      "epoch": 0.6095,
+      "grad_norm": 23.76275938679832,
+      "kl": 4.3046875,
+      "learning_rate": 4.572081165715167e-07,
+      "loss": 0.0931,
+      "reward": 2.224815607070923,
+      "reward_std": 0.5812746435403824,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.030392706394195557,
+      "rewards/tag_count_reward": 0.796875,
+      "step": 1219
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 304.5416717529297,
+      "epoch": 0.61,
+      "grad_norm": 16.029210941264648,
+      "kl": 3.0,
+      "learning_rate": 4.5643973913200837e-07,
+      "loss": 0.0726,
+      "reward": 2.567992091178894,
+      "reward_std": 0.6492457389831543,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03791070729494095,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1220
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 286.0,
+      "epoch": 0.6105,
+      "grad_norm": 14.40075724881573,
+      "kl": 2.84375,
+      "learning_rate": 4.556716466932803e-07,
+      "loss": 0.0603,
+      "reward": 2.1511141061782837,
+      "reward_std": 0.6159583330154419,
+      "rewards/accuracy_reward": 0.3333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.017288665752857924,
+      "rewards/tag_count_reward": 0.8489583730697632,
+      "step": 1221
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 322.87501525878906,
+      "epoch": 0.611,
+      "grad_norm": 7.252852057442951,
+      "kl": 0.806640625,
+      "learning_rate": 4.549038415950751e-07,
+      "loss": -0.048,
+      "reward": 2.8554985523223877,
+      "reward_std": 0.38053126633167267,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03339047962799668,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1222
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 299.3333435058594,
+      "epoch": 0.6115,
+      "grad_norm": 9.575941886870105,
+      "kl": 2.7734375,
+      "learning_rate": 4.5413632617626054e-07,
+      "loss": -0.1175,
+      "reward": 2.2987377047538757,
+      "reward_std": 0.7605935335159302,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556416511536,
+      "rewards/repetition_penalty_reward": -0.02765120565891266,
+      "rewards/tag_count_reward": 0.8125000298023224,
+      "step": 1223
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 319.8333435058594,
+      "epoch": 0.612,
+      "grad_norm": 7.684972316087842,
+      "kl": 1.193359375,
+      "learning_rate": 4.5336910277482155e-07,
+      "loss": -0.0379,
+      "reward": 2.6804885864257812,
+      "reward_std": 0.4239862561225891,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02437271736562252,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1224
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 263.9166717529297,
+      "epoch": 0.6125,
+      "grad_norm": 9.7169462842633,
+      "kl": 2.3359375,
+      "learning_rate": 4.526021737278537e-07,
+      "loss": -0.0673,
+      "reward": 2.2160778045654297,
+      "reward_std": 0.7273769080638885,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.8819445371627808,
+      "rewards/repetition_penalty_reward": -0.02524164505302906,
+      "rewards/tag_count_reward": 0.8177083432674408,
+      "step": 1225
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 327.37501525878906,
+      "epoch": 0.613,
+      "grad_norm": 23.176995418112103,
+      "kl": 3.3359375,
+      "learning_rate": 4.51835541371556e-07,
+      "loss": 0.0103,
+      "reward": 2.078250527381897,
+      "reward_std": 0.5842549949884415,
+      "rewards/accuracy_reward": 0.31250002048909664,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.020707917399704456,
+      "rewards/tag_count_reward": 0.8281250298023224,
+      "step": 1226
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.18750762939453,
+      "epoch": 0.6135,
+      "grad_norm": 105.22916350823351,
+      "kl": 10.765625,
+      "learning_rate": 4.5106920804122304e-07,
+      "loss": 0.1645,
+      "reward": 2.2080377340316772,
+      "reward_std": 0.8133403062820435,
+      "rewards/accuracy_reward": 0.5833333730697632,
+      "rewards/reasoning_steps_reward": 0.8749999701976776,
+      "rewards/repetition_penalty_reward": -0.015920499339699745,
+      "rewards/tag_count_reward": 0.7656250298023224,
+      "step": 1227
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 343.1875,
+      "epoch": 0.614,
+      "grad_norm": 34.43707466221288,
+      "kl": 2.984375,
+      "learning_rate": 4.503031760712397e-07,
+      "loss": 0.0916,
+      "reward": 2.6236408948898315,
+      "reward_std": 0.5556021928787231,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02219252847135067,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1228
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.6041717529297,
+      "epoch": 0.6145,
+      "grad_norm": 47.071050622736806,
+      "kl": 7.921875,
+      "learning_rate": 4.4953744779507197e-07,
+      "loss": 0.0256,
+      "reward": 2.315521240234375,
+      "reward_std": 0.8524576425552368,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.021284347400069237,
+      "rewards/tag_count_reward": 0.8229166865348816,
+      "step": 1229
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 285.0,
+      "epoch": 0.615,
+      "grad_norm": 10.5407270563147,
+      "kl": 2.294921875,
+      "learning_rate": 4.4877202554526084e-07,
+      "loss": -0.0323,
+      "reward": 2.4727187156677246,
+      "reward_std": 0.6452691853046417,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.03422584570944309,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1230
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 294.8333435058594,
+      "epoch": 0.6155,
+      "grad_norm": 8.990594800844864,
+      "kl": 0.833984375,
+      "learning_rate": 4.480069116534151e-07,
+      "loss": -0.0071,
+      "reward": 2.5792577266693115,
+      "reward_std": 0.3115484416484833,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03011729847639799,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1231
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.8333435058594,
+      "epoch": 0.616,
+      "grad_norm": 8.186829347268167,
+      "kl": 1.681640625,
+      "learning_rate": 4.4724210845020494e-07,
+      "loss": -0.0651,
+      "reward": 2.419158935546875,
+      "reward_std": 0.5638627856969833,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.033966176211833954,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1232
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.5416717529297,
+      "epoch": 0.6165,
+      "grad_norm": 16.90957738932618,
+      "kl": 2.40625,
+      "learning_rate": 4.4647761826535303e-07,
+      "loss": -0.0256,
+      "reward": 2.2396740913391113,
+      "reward_std": 0.492450088262558,
+      "rewards/accuracy_reward": 0.458333358168602,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.03810371086001396,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1233
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.3541717529297,
+      "epoch": 0.617,
+      "grad_norm": 11.105722560883672,
+      "kl": 1.4140625,
+      "learning_rate": 4.457134434276293e-07,
+      "loss": -0.1259,
+      "reward": 2.2069406509399414,
+      "reward_std": 0.5749087035655975,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.03437890112400055,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1234
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.7083435058594,
+      "epoch": 0.6175,
+      "grad_norm": 9.032805027425631,
+      "kl": 1.0859375,
+      "learning_rate": 4.449495862648427e-07,
+      "loss": 0.0574,
+      "reward": 2.7222955226898193,
+      "reward_std": 0.23863910883665085,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.022496161051094532,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1235
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.2083435058594,
+      "epoch": 0.618,
+      "grad_norm": 17.78102072531694,
+      "kl": 2.375,
+      "learning_rate": 4.441860491038345e-07,
+      "loss": 0.0578,
+      "reward": 2.586129903793335,
+      "reward_std": 0.6346015930175781,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.02324526757001877,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1236
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.7708435058594,
+      "epoch": 0.6185,
+      "grad_norm": 9.03437362621798,
+      "kl": 1.70703125,
+      "learning_rate": 4.4342283427047164e-07,
+      "loss": -0.0525,
+      "reward": 2.5365703105926514,
+      "reward_std": 0.5162703096866608,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.024193717166781425,
+      "rewards/tag_count_reward": 0.9427083730697632,
+      "step": 1237
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 312.4375,
+      "epoch": 0.619,
+      "grad_norm": 15.517046311400755,
+      "kl": 2.40234375,
+      "learning_rate": 4.4265994408963867e-07,
+      "loss": 0.063,
+      "reward": 2.5961966514587402,
+      "reward_std": 0.6436916887760162,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02880349662154913,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1238
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.25001525878906,
+      "epoch": 0.6195,
+      "grad_norm": 24.87465794144074,
+      "kl": 3.734375,
+      "learning_rate": 4.418973808852313e-07,
+      "loss": 0.0546,
+      "reward": 2.469591498374939,
+      "reward_std": 0.5368989706039429,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.018255963921546936,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1239
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 342.12501525878906,
+      "epoch": 0.62,
+      "grad_norm": 12.84411480237786,
+      "kl": 2.25390625,
+      "learning_rate": 4.4113514698014953e-07,
+      "loss": 0.024,
+      "reward": 2.184650182723999,
+      "reward_std": 0.47897736728191376,
+      "rewards/accuracy_reward": 0.3541666679084301,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.023683225736021996,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1240
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 296.5208435058594,
+      "epoch": 0.6205,
+      "grad_norm": 8.658225737289808,
+      "kl": 1.1953125,
+      "learning_rate": 4.403732446962899e-07,
+      "loss": -0.0936,
+      "reward": 2.3716037273406982,
+      "reward_std": 0.558637946844101,
+      "rewards/accuracy_reward": 0.5208333358168602,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.01554908649995923,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1241
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 312.60418701171875,
+      "epoch": 0.621,
+      "grad_norm": 7.919211211849244,
+      "kl": 1.11328125,
+      "learning_rate": 4.3961167635453876e-07,
+      "loss": 0.0181,
+      "reward": 2.5375410318374634,
+      "reward_std": 0.37939758598804474,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02148678805679083,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1242
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.50001525878906,
+      "epoch": 0.6215,
+      "grad_norm": 5.99197945202883,
+      "kl": 0.673828125,
+      "learning_rate": 4.388504442747657e-07,
+      "loss": -0.0048,
+      "reward": 2.649431347846985,
+      "reward_std": 0.3595951795578003,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.034596722573041916,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1243
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 325.56251525878906,
+      "epoch": 0.622,
+      "grad_norm": 7.039763497814729,
+      "kl": 1.1123046875,
+      "learning_rate": 4.3808955077581546e-07,
+      "loss": -0.0806,
+      "reward": 2.560014247894287,
+      "reward_std": 0.45148639380931854,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.025055398233234882,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1244
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 334.18751525878906,
+      "epoch": 0.6225,
+      "grad_norm": 9.138240023491278,
+      "kl": 0.78515625,
+      "learning_rate": 4.373289981755013e-07,
+      "loss": -0.0215,
+      "reward": 2.5294055938720703,
+      "reward_std": 0.3349990248680115,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.050455573946237564,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1245
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.5,
+      "epoch": 0.623,
+      "grad_norm": 8.492208011646788,
+      "kl": 1.853515625,
+      "learning_rate": 4.365687887905988e-07,
+      "loss": -0.0616,
+      "reward": 2.7429254055023193,
+      "reward_std": 0.5102400928735733,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.01575501961633563,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1246
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 297.93751525878906,
+      "epoch": 0.6235,
+      "grad_norm": 57.99567662946909,
+      "kl": 7.73828125,
+      "learning_rate": 4.358089249368375e-07,
+      "loss": 0.086,
+      "reward": 2.5771158933639526,
+      "reward_std": 0.5258086919784546,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.021842443384230137,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1247
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 276.8333435058594,
+      "epoch": 0.624,
+      "grad_norm": 341.16700521948934,
+      "kl": 29.625,
+      "learning_rate": 4.350494089288943e-07,
+      "loss": 0.676,
+      "reward": 2.2864397764205933,
+      "reward_std": 0.4129794090986252,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.027796387672424316,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1248
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.68751525878906,
+      "epoch": 0.6245,
+      "grad_norm": 181.81088614690876,
+      "kl": 15.53125,
+      "learning_rate": 4.3429024308038686e-07,
+      "loss": 0.2868,
+      "reward": 2.834537148475647,
+      "reward_std": 0.33811767399311066,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.03178227413445711,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1249
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 297.0625,
+      "epoch": 0.625,
+      "grad_norm": 11.08135265703862,
+      "kl": 1.1259765625,
+      "learning_rate": 4.3353142970386557e-07,
+      "loss": 0.1305,
+      "reward": 2.7022109031677246,
+      "reward_std": 0.10300289653241634,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.01653942931443453,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1250
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.9583435058594,
+      "epoch": 0.6255,
+      "grad_norm": 124.39669515969251,
+      "kl": 14.46875,
+      "learning_rate": 4.327729711108082e-07,
+      "loss": 0.2354,
+      "reward": 2.619844675064087,
+      "reward_std": 0.6426686346530914,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.019044365733861923,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1251
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 314.0208435058594,
+      "epoch": 0.626,
+      "grad_norm": 4.466923459475626,
+      "kl": 0.685546875,
+      "learning_rate": 4.3201486961161093e-07,
+      "loss": -0.0116,
+      "reward": 2.894577383995056,
+      "reward_std": 0.2078223153948784,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.029034032486379147,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1252
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 294.9791717529297,
+      "epoch": 0.6265,
+      "grad_norm": 26.605555594559796,
+      "kl": 1.48828125,
+      "learning_rate": 4.312571275155823e-07,
+      "loss": -0.038,
+      "reward": 2.5493518114089966,
+      "reward_std": 0.5423057973384857,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.040926164016127586,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1253
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 328.18751525878906,
+      "epoch": 0.627,
+      "grad_norm": 8.302364671310045,
+      "kl": 0.7890625,
+      "learning_rate": 4.304997471309361e-07,
+      "loss": -0.0157,
+      "reward": 2.6195199489593506,
+      "reward_std": 0.5493273586034775,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.045410582795739174,
+      "rewards/tag_count_reward": 0.9427083730697632,
+      "step": 1254
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 304.4166717529297,
+      "epoch": 0.6275,
+      "grad_norm": 13.946595632340095,
+      "kl": 1.984375,
+      "learning_rate": 4.297427307647844e-07,
+      "loss": -0.0215,
+      "reward": 2.4039559364318848,
+      "reward_std": 0.4896374046802521,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02312744501978159,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1255
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.3958435058594,
+      "epoch": 0.628,
+      "grad_norm": 11.365303476859737,
+      "kl": 0.7919921875,
+      "learning_rate": 4.2898608072313045e-07,
+      "loss": -0.0273,
+      "reward": 2.542173385620117,
+      "reward_std": 0.4263547882437706,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.027271111495792866,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1256
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 301.06251525878906,
+      "epoch": 0.6285,
+      "grad_norm": 6.488072817521581,
+      "kl": 1.6943359375,
+      "learning_rate": 4.2822979931086144e-07,
+      "loss": 0.064,
+      "reward": 2.6413581371307373,
+      "reward_std": 0.32615266740322113,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.025308695621788502,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1257
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.4166717529297,
+      "epoch": 0.629,
+      "grad_norm": 5.571259374449054,
+      "kl": 0.7109375,
+      "learning_rate": 4.2747388883174154e-07,
+      "loss": -0.0249,
+      "reward": 2.5305432081222534,
+      "reward_std": 0.34117555618286133,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.03716518171131611,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1258
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 319.0208435058594,
+      "epoch": 0.6295,
+      "grad_norm": 10.818868109831325,
+      "kl": 1.11328125,
+      "learning_rate": 4.267183515884054e-07,
+      "loss": -0.0971,
+      "reward": 2.6202439069747925,
+      "reward_std": 0.6800469756126404,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.02732554590329528,
+      "rewards/tag_count_reward": 0.9114583730697632,
+      "step": 1259
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.8333435058594,
+      "epoch": 0.63,
+      "grad_norm": 5.109825873578351,
+      "kl": 1.12890625,
+      "learning_rate": 4.2596318988235037e-07,
+      "loss": -0.0602,
+      "reward": 2.7829357385635376,
+      "reward_std": 0.4666026383638382,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.024355984292924404,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1260
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.4166717529297,
+      "epoch": 0.6305,
+      "grad_norm": 113.69520754686019,
+      "kl": 6.984375,
+      "learning_rate": 4.2520840601392996e-07,
+      "loss": 0.0285,
+      "reward": 2.4763646125793457,
+      "reward_std": 0.5876338928937912,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02189927827566862,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1261
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 277.20833587646484,
+      "epoch": 0.631,
+      "grad_norm": 4.878770353368535,
+      "kl": 0.8046875,
+      "learning_rate": 4.2445400228234687e-07,
+      "loss": -0.0089,
+      "reward": 2.2267041206359863,
+      "reward_std": 0.21407258417457342,
+      "rewards/accuracy_reward": 0.3125,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.023295993451029062,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1262
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 248.00000762939453,
+      "epoch": 0.6315,
+      "grad_norm": 91.61805536571175,
+      "kl": 14.0,
+      "learning_rate": 4.2369998098564554e-07,
+      "loss": 0.1059,
+      "reward": 2.389165997505188,
+      "reward_std": 0.8639044761657715,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.8819444179534912,
+      "rewards/repetition_penalty_reward": -0.024028603918850422,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1263
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.4791793823242,
+      "epoch": 0.632,
+      "grad_norm": 61.3227518293924,
+      "kl": 7.921875,
+      "learning_rate": 4.2294634442070553e-07,
+      "loss": 0.0403,
+      "reward": 2.4555550813674927,
+      "reward_std": 0.7910381853580475,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.02361162193119526,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1264
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 298.97918701171875,
+      "epoch": 0.6325,
+      "grad_norm": 11.219358419697596,
+      "kl": 2.75390625,
+      "learning_rate": 4.2219309488323487e-07,
+      "loss": 0.0209,
+      "reward": 2.7085201740264893,
+      "reward_std": 0.5442517399787903,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.029327161610126495,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1265
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 275.7708435058594,
+      "epoch": 0.633,
+      "grad_norm": 398.00956099056515,
+      "kl": 23.125,
+      "learning_rate": 4.214402346677619e-07,
+      "loss": 0.2082,
+      "reward": 2.259987235069275,
+      "reward_std": 0.9177981615066528,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.8263889253139496,
+      "rewards/repetition_penalty_reward": -0.024735040962696075,
+      "rewards/tag_count_reward": 0.8333333432674408,
+      "step": 1266
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.9583435058594,
+      "epoch": 0.6335,
+      "grad_norm": 24.983686575360647,
+      "kl": 3.240234375,
+      "learning_rate": 4.206877660676297e-07,
+      "loss": 0.0538,
+      "reward": 2.548527956008911,
+      "reward_std": 0.6168785095214844,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.01570827839896083,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1267
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.6875,
+      "epoch": 0.634,
+      "grad_norm": 29.77346744252865,
+      "kl": 3.23046875,
+      "learning_rate": 4.1993569137498776e-07,
+      "loss": 0.0976,
+      "reward": 2.5990532636642456,
+      "reward_std": 0.3927089273929596,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.01726605836302042,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1268
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.7708435058594,
+      "epoch": 0.6345,
+      "grad_norm": 6.809776019842994,
+      "kl": 1.5078125,
+      "learning_rate": 4.1918401288078633e-07,
+      "loss": -0.0669,
+      "reward": 2.4295482635498047,
+      "reward_std": 0.6562110781669617,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.014896340668201447,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1269
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 242.54167938232422,
+      "epoch": 0.635,
+      "grad_norm": 13.637945833280545,
+      "kl": 2.11328125,
+      "learning_rate": 4.1843273287476854e-07,
+      "loss": -0.2244,
+      "reward": 2.2396379709243774,
+      "reward_std": 0.7876444458961487,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.8472222685813904,
+      "rewards/repetition_penalty_reward": -0.024251021444797516,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1270
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 224.18750762939453,
+      "epoch": 0.6355,
+      "grad_norm": 9.401287077014372,
+      "kl": 2.37109375,
+      "learning_rate": 4.1768185364546326e-07,
+      "loss": -0.2551,
+      "reward": 1.9714569449424744,
+      "reward_std": 0.8513112664222717,
+      "rewards/accuracy_reward": 0.5000000223517418,
+      "rewards/reasoning_steps_reward": 0.7500000596046448,
+      "rewards/repetition_penalty_reward": -0.018126386683434248,
+      "rewards/tag_count_reward": 0.7395833432674408,
+      "step": 1271
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 319.7083435058594,
+      "epoch": 0.636,
+      "grad_norm": 5.439379439237209,
+      "kl": 0.857421875,
+      "learning_rate": 4.1693137748017915e-07,
+      "loss": -0.0795,
+      "reward": 2.6746604442596436,
+      "reward_std": 0.536898672580719,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.02672865055501461,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1272
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 255.56250762939453,
+      "epoch": 0.6365,
+      "grad_norm": 6.202209412890006,
+      "kl": 0.8515625,
+      "learning_rate": 4.161813066649963e-07,
+      "loss": -0.0982,
+      "reward": 2.706597089767456,
+      "reward_std": 0.5873250961303711,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.013889186084270477,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1273
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 253.87500762939453,
+      "epoch": 0.637,
+      "grad_norm": 8.311119282622155,
+      "kl": 1.412109375,
+      "learning_rate": 4.15431643484761e-07,
+      "loss": -0.0722,
+      "reward": 2.4473708868026733,
+      "reward_std": 0.5833633840084076,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.02832351950928569,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1274
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.5416717529297,
+      "epoch": 0.6375,
+      "grad_norm": 5.198591187578019,
+      "kl": 1.04296875,
+      "learning_rate": 4.146823902230772e-07,
+      "loss": -0.0484,
+      "reward": 2.8356668949127197,
+      "reward_std": 0.44375722110271454,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.020235823933035135,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1275
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 230.58334350585938,
+      "epoch": 0.638,
+      "grad_norm": 6.926096135008214,
+      "kl": 1.96484375,
+      "learning_rate": 4.1393354916230005e-07,
+      "loss": -0.0609,
+      "reward": 2.677065134048462,
+      "reward_std": 0.594237208366394,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.01043500192463398,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1276
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 238.68750762939453,
+      "epoch": 0.6385,
+      "grad_norm": 48.40360467461647,
+      "kl": 3.640625,
+      "learning_rate": 4.1318512258352936e-07,
+      "loss": 0.0027,
+      "reward": 2.614649772644043,
+      "reward_std": 0.4902600198984146,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.008614218328148127,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1277
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 256.50001525878906,
+      "epoch": 0.639,
+      "grad_norm": 17.05514052018502,
+      "kl": 3.43359375,
+      "learning_rate": 4.124371127666024e-07,
+      "loss": -0.045,
+      "reward": 2.4537200927734375,
+      "reward_std": 0.6568253189325333,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.013293889816850424,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1278
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.0833435058594,
+      "epoch": 0.6395,
+      "grad_norm": 12.772767939709558,
+      "kl": 3.58984375,
+      "learning_rate": 4.1168952199008677e-07,
+      "loss": -0.074,
+      "reward": 2.3198667764663696,
+      "reward_std": 0.6525312066078186,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.02561924420297146,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1279
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.2916717529297,
+      "epoch": 0.64,
+      "grad_norm": 13.074456042005389,
+      "kl": 2.43359375,
+      "learning_rate": 4.1094235253127374e-07,
+      "loss": -0.0559,
+      "reward": 2.621940493583679,
+      "reward_std": 0.6388831436634064,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03430960513651371,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1280
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.3541717529297,
+      "epoch": 0.6405,
+      "grad_norm": 20.83070325460604,
+      "kl": 3.6875,
+      "learning_rate": 4.101956066661708e-07,
+      "loss": -0.0974,
+      "reward": 2.285021424293518,
+      "reward_std": 0.5357790589332581,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.018798216711729765,
+      "rewards/tag_count_reward": 0.8802083730697632,
+      "step": 1281
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 274.04168701171875,
+      "epoch": 0.641,
+      "grad_norm": 12.971542858607965,
+      "kl": 2.666015625,
+      "learning_rate": 4.0944928666949527e-07,
+      "loss": -0.0943,
+      "reward": 2.2288068532943726,
+      "reward_std": 0.6288715898990631,
+      "rewards/accuracy_reward": 0.416666679084301,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.014248816296458244,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1282
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.25,
+      "epoch": 0.6415,
+      "grad_norm": 11.653407474012347,
+      "kl": 2.9140625,
+      "learning_rate": 4.0870339481466774e-07,
+      "loss": -0.0501,
+      "reward": 1.9373607635498047,
+      "reward_std": 0.5385548174381256,
+      "rewards/accuracy_reward": 0.3125000111758709,
+      "rewards/reasoning_steps_reward": 0.8472222685813904,
+      "rewards/repetition_penalty_reward": -0.02444479614496231,
+      "rewards/tag_count_reward": 0.8020833432674408,
+      "step": 1283
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.56250762939453,
+      "epoch": 0.642,
+      "grad_norm": 8.453230476877637,
+      "kl": 0.779296875,
+      "learning_rate": 4.079579333738039e-07,
+      "loss": 0.0324,
+      "reward": 2.838531732559204,
+      "reward_std": 0.28950972855091095,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.036468397825956345,
+      "rewards/tag_count_reward": 0.9583333730697632,
+      "step": 1284
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 274.7916717529297,
+      "epoch": 0.6425,
+      "grad_norm": 8.516876725525995,
+      "kl": 0.984375,
+      "learning_rate": 4.0721290461770863e-07,
+      "loss": -0.0514,
+      "reward": 2.528114676475525,
+      "reward_std": 0.6328976899385452,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.018760375678539276,
+      "rewards/tag_count_reward": 0.8802083730697632,
+      "step": 1285
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 254.66667938232422,
+      "epoch": 0.643,
+      "grad_norm": 11.245200868491317,
+      "kl": 1.1171875,
+      "learning_rate": 4.064683108158685e-07,
+      "loss": -0.1401,
+      "reward": 2.3839638233184814,
+      "reward_std": 0.6318697333335876,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.022286144085228443,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1286
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 277.9583435058594,
+      "epoch": 0.6435,
+      "grad_norm": 6.37277111754768,
+      "kl": 1.4609375,
+      "learning_rate": 4.057241542364457e-07,
+      "loss": -0.1055,
+      "reward": 2.32645583152771,
+      "reward_std": 0.5588487088680267,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.03465544432401657,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1287
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.1666717529297,
+      "epoch": 0.644,
+      "grad_norm": 16.043998757477272,
+      "kl": 4.4296875,
+      "learning_rate": 4.0498043714627006e-07,
+      "loss": -0.1519,
+      "reward": 2.1210697889328003,
+      "reward_std": 0.9449804127216339,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.8333333730697632,
+      "rewards/repetition_penalty_reward": -0.024763552471995354,
+      "rewards/tag_count_reward": 0.7916666865348816,
+      "step": 1288
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 302.4791717529297,
+      "epoch": 0.6445,
+      "grad_norm": 8.66946655884018,
+      "kl": 1.9375,
+      "learning_rate": 4.042371618108329e-07,
+      "loss": -0.0738,
+      "reward": 2.5021121501922607,
+      "reward_std": 0.7234326303005219,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.029137907549738884,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1289
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 258.39583587646484,
+      "epoch": 0.645,
+      "grad_norm": 6.507177763175018,
+      "kl": 2.046875,
+      "learning_rate": 4.034943304942796e-07,
+      "loss": -0.1071,
+      "reward": 2.659383535385132,
+      "reward_std": 0.7204856872558594,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.021172089502215385,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1290
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 269.87500762939453,
+      "epoch": 0.6455,
+      "grad_norm": 8.84789787566044,
+      "kl": 1.96484375,
+      "learning_rate": 4.027519454594033e-07,
+      "loss": -0.0137,
+      "reward": 2.724217653274536,
+      "reward_std": 0.6305935382843018,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.022310104221105576,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1291
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 228.14584350585938,
+      "epoch": 0.646,
+      "grad_norm": 54.434135625320714,
+      "kl": 7.625,
+      "learning_rate": 4.020100089676376e-07,
+      "loss": 0.0351,
+      "reward": 2.350663661956787,
+      "reward_std": 0.5377750098705292,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.015656011179089546,
+      "rewards/tag_count_reward": 0.8385416865348816,
+      "step": 1292
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 246.37500762939453,
+      "epoch": 0.6465,
+      "grad_norm": 36.29934720854378,
+      "kl": 9.203125,
+      "learning_rate": 4.012685232790497e-07,
+      "loss": -0.1163,
+      "reward": 2.269049644470215,
+      "reward_std": 0.7809967696666718,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.017408739775419235,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1293
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 326.56251525878906,
+      "epoch": 0.647,
+      "grad_norm": 13.353341754245465,
+      "kl": 1.373046875,
+      "learning_rate": 4.005274906523336e-07,
+      "loss": -0.0158,
+      "reward": 2.862304449081421,
+      "reward_std": 0.2681258460506797,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03700113669037819,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1294
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 234.875,
+      "epoch": 0.6475,
+      "grad_norm": 17.584019029906294,
+      "kl": 4.6484375,
+      "learning_rate": 3.9978691334480306e-07,
+      "loss": -0.026,
+      "reward": 2.621469259262085,
+      "reward_std": 0.6262362897396088,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.022628027945756912,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1295
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.25000762939453,
+      "epoch": 0.648,
+      "grad_norm": 20.539771262644624,
+      "kl": 2.46484375,
+      "learning_rate": 3.9904679361238526e-07,
+      "loss": -0.0646,
+      "reward": 2.5812262296676636,
+      "reward_std": 0.40000835061073303,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.019468236714601517,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1296
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.3125,
+      "epoch": 0.6485,
+      "grad_norm": 12.020081562329604,
+      "kl": 1.96875,
+      "learning_rate": 3.9830713370961313e-07,
+      "loss": 0.0126,
+      "reward": 2.7614296674728394,
+      "reward_std": 0.3552953898906708,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.01287601888179779,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1297
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 250.25001525878906,
+      "epoch": 0.649,
+      "grad_norm": 7.055794372513674,
+      "kl": 1.11328125,
+      "learning_rate": 3.975679358896189e-07,
+      "loss": -0.0177,
+      "reward": 2.7628384828567505,
+      "reward_std": 0.35927814757451415,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.027092115953564644,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1298
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 224.70834350585938,
+      "epoch": 0.6495,
+      "grad_norm": 10.528185417792633,
+      "kl": 1.76953125,
+      "learning_rate": 3.968292024041275e-07,
+      "loss": -0.1546,
+      "reward": 2.3756613731384277,
+      "reward_std": 0.7093529403209686,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.8472222089767456,
+      "rewards/repetition_penalty_reward": -0.018436014652252197,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1299
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.25,
+      "epoch": 0.65,
+      "grad_norm": 13.557597077977892,
+      "kl": 2.1328125,
+      "learning_rate": 3.9609093550344907e-07,
+      "loss": -0.0656,
+      "reward": 2.1020314693450928,
+      "reward_std": 0.7527401447296143,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.8680556118488312,
+      "rewards/repetition_penalty_reward": -0.02644097339361906,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1300
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.0208435058594,
+      "epoch": 0.6505,
+      "grad_norm": 5.083335577912237,
+      "kl": 0.4794921875,
+      "learning_rate": 3.953531374364728e-07,
+      "loss": 0.0102,
+      "reward": 2.6702791452407837,
+      "reward_std": 0.36818480491638184,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03284592926502228,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1301
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.6875,
+      "epoch": 0.651,
+      "grad_norm": 7.37275238994372,
+      "kl": 1.07421875,
+      "learning_rate": 3.946158104506594e-07,
+      "loss": -0.0573,
+      "reward": 2.6404601335525513,
+      "reward_std": 0.5752403140068054,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.029678759165108204,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1302
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 263.7291717529297,
+      "epoch": 0.6515,
+      "grad_norm": 6.525040125531069,
+      "kl": 2.5,
+      "learning_rate": 3.938789567920349e-07,
+      "loss": -0.0352,
+      "reward": 2.368511199951172,
+      "reward_std": 0.5849394798278809,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.02211400307714939,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1303
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 238.2916717529297,
+      "epoch": 0.652,
+      "grad_norm": 13.454959697482677,
+      "kl": 2.1875,
+      "learning_rate": 3.931425787051832e-07,
+      "loss": -0.0314,
+      "reward": 2.6546378135681152,
+      "reward_std": 0.6842877864837646,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.012028906028717756,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1304
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.8541793823242,
+      "epoch": 0.6525,
+      "grad_norm": 28.026859752832458,
+      "kl": 5.671875,
+      "learning_rate": 3.924066784332396e-07,
+      "loss": -0.1091,
+      "reward": 2.2767953872680664,
+      "reward_std": 0.6093401312828064,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.8750000298023224,
+      "rewards/repetition_penalty_reward": -0.025287946220487356,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1305
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.56250762939453,
+      "epoch": 0.653,
+      "grad_norm": 18.361877129436802,
+      "kl": 2.220703125,
+      "learning_rate": 3.9167125821788416e-07,
+      "loss": -0.0351,
+      "reward": 2.4435291290283203,
+      "reward_std": 0.3841914087533951,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.011332074645906687,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1306
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 274.87501525878906,
+      "epoch": 0.6535,
+      "grad_norm": 9.18937969115552,
+      "kl": 2.7890625,
+      "learning_rate": 3.909363202993343e-07,
+      "loss": -0.053,
+      "reward": 2.573843240737915,
+      "reward_std": 0.5460654944181442,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.02685134019702673,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1307
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 239.64584350585938,
+      "epoch": 0.654,
+      "grad_norm": 9.701372830225981,
+      "kl": 2.421875,
+      "learning_rate": 3.902018669163384e-07,
+      "loss": -0.0271,
+      "reward": 2.419216573238373,
+      "reward_std": 0.44797763228416443,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.007866791682317853,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1308
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.66668701171875,
+      "epoch": 0.6545,
+      "grad_norm": 6.341395491310268,
+      "kl": 1.55078125,
+      "learning_rate": 3.894679003061686e-07,
+      "loss": -0.0434,
+      "reward": 2.7170095443725586,
+      "reward_std": 0.5187918990850449,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.027782058343291283,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1309
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.10418701171875,
+      "epoch": 0.655,
+      "grad_norm": 4.622903893254928,
+      "kl": 0.580078125,
+      "learning_rate": 3.8873442270461485e-07,
+      "loss": -0.0032,
+      "reward": 2.8831721544265747,
+      "reward_std": 0.2194080576300621,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.01613345229998231,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1310
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 234.56250762939453,
+      "epoch": 0.6555,
+      "grad_norm": 6.403895926939174,
+      "kl": 1.96875,
+      "learning_rate": 3.88001436345977e-07,
+      "loss": -0.0914,
+      "reward": 2.64253830909729,
+      "reward_std": 0.5459905564785004,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.015447880607098341,
+      "rewards/tag_count_reward": 0.9427083730697632,
+      "step": 1311
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 254.02083587646484,
+      "epoch": 0.656,
+      "grad_norm": 4.624188058528878,
+      "kl": 1.1171875,
+      "learning_rate": 3.872689434630585e-07,
+      "loss": -0.0328,
+      "reward": 2.666394829750061,
+      "reward_std": 0.4922281354665756,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.02110534254461527,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1312
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 252.8541717529297,
+      "epoch": 0.6565,
+      "grad_norm": 10.421845265889829,
+      "kl": 0.916015625,
+      "learning_rate": 3.8653694628715984e-07,
+      "loss": -0.0879,
+      "reward": 2.8016003370285034,
+      "reward_std": 0.5981872081756592,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.010899642948061228,
+      "rewards/tag_count_reward": 0.9583333730697632,
+      "step": 1313
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 229.02084350585938,
+      "epoch": 0.657,
+      "grad_norm": 10.523509449540256,
+      "kl": 2.55859375,
+      "learning_rate": 3.8580544704807117e-07,
+      "loss": -0.0991,
+      "reward": 2.440296769142151,
+      "reward_std": 0.6269799470901489,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.012828361243009567,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1314
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 251.89583587646484,
+      "epoch": 0.6575,
+      "grad_norm": 21.647956275310747,
+      "kl": 3.4375,
+      "learning_rate": 3.850744479740663e-07,
+      "loss": -0.0654,
+      "reward": 2.577293634414673,
+      "reward_std": 0.5918062180280685,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.018192541785538197,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1315
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 195.00000762939453,
+      "epoch": 0.658,
+      "grad_norm": 63.82832688411775,
+      "kl": 7.328125,
+      "learning_rate": 3.843439512918949e-07,
+      "loss": 0.0124,
+      "reward": 1.5498919486999512,
+      "reward_std": 0.5158031135797501,
+      "rewards/accuracy_reward": 0.0416666679084301,
+      "rewards/reasoning_steps_reward": 0.7638889253139496,
+      "rewards/repetition_penalty_reward": -0.016080408822745085,
+      "rewards/tag_count_reward": 0.7604166865348816,
+      "step": 1316
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 252.0625,
+      "epoch": 0.6585,
+      "grad_norm": 6.870792569824665,
+      "kl": 2.2265625,
+      "learning_rate": 3.8361395922677687e-07,
+      "loss": -0.1222,
+      "reward": 2.7788249254226685,
+      "reward_std": 0.597758024930954,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.012841662392020226,
+      "rewards/tag_count_reward": 0.9583333730697632,
+      "step": 1317
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.00001525878906,
+      "epoch": 0.659,
+      "grad_norm": 12.883962513964198,
+      "kl": 2.453125,
+      "learning_rate": 3.8288447400239443e-07,
+      "loss": -0.0232,
+      "reward": 2.208429753780365,
+      "reward_std": 0.39593251049518585,
+      "rewards/accuracy_reward": 0.3541666679084301,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.025945200584828854,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1318
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 271.18751525878906,
+      "epoch": 0.6595,
+      "grad_norm": 11.437028269960171,
+      "kl": 2.12109375,
+      "learning_rate": 3.82155497840886e-07,
+      "loss": -0.0909,
+      "reward": 2.5074195861816406,
+      "reward_std": 0.697433352470398,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.015149833634495735,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1319
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 244.72917938232422,
+      "epoch": 0.66,
+      "grad_norm": 15.92814404052159,
+      "kl": 3.4296875,
+      "learning_rate": 3.8142703296283953e-07,
+      "loss": -0.0835,
+      "reward": 2.2199000120162964,
+      "reward_std": 0.7194198369979858,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.02836394216865301,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1320
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 234.8541717529297,
+      "epoch": 0.6605,
+      "grad_norm": 21.139528290510924,
+      "kl": 4.59375,
+      "learning_rate": 3.806990815872855e-07,
+      "loss": -0.0918,
+      "reward": 2.3190531730651855,
+      "reward_std": 0.38901272416114807,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.010808073915541172,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1321
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 237.95834350585938,
+      "epoch": 0.661,
+      "grad_norm": 8.548692602029842,
+      "kl": 2.4921875,
+      "learning_rate": 3.7997164593168983e-07,
+      "loss": -0.062,
+      "reward": 2.5779298543930054,
+      "reward_std": 0.7175216972827911,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.022764576133340597,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1322
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 238.06250762939453,
+      "epoch": 0.6615,
+      "grad_norm": 8.538300938371009,
+      "kl": 1.42578125,
+      "learning_rate": 3.7924472821194765e-07,
+      "loss": -0.1421,
+      "reward": 2.6461488008499146,
+      "reward_std": 0.7617769837379456,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.018781788181513548,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1323
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 248.62500762939453,
+      "epoch": 0.662,
+      "grad_norm": 10.10993358277435,
+      "kl": 1.47265625,
+      "learning_rate": 3.785183306423767e-07,
+      "loss": 0.0362,
+      "reward": 2.4081956148147583,
+      "reward_std": 0.4368290901184082,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.010207333602011204,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1324
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 243.58334350585938,
+      "epoch": 0.6625,
+      "grad_norm": 6.92812761375714,
+      "kl": 1.0556640625,
+      "learning_rate": 3.777924554357096e-07,
+      "loss": -0.0024,
+      "reward": 2.8137227296829224,
+      "reward_std": 0.41092177480459213,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.02308299019932747,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1325
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.8333435058594,
+      "epoch": 0.663,
+      "grad_norm": 7.315502843369173,
+      "kl": 2.02734375,
+      "learning_rate": 3.7706710480308835e-07,
+      "loss": -0.1371,
+      "reward": 2.6259608268737793,
+      "reward_std": 0.6581228971481323,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.03549747634679079,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1326
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.2291717529297,
+      "epoch": 0.6635,
+      "grad_norm": 6.63436769667506,
+      "kl": 1.81640625,
+      "learning_rate": 3.7634228095405673e-07,
+      "loss": -0.0385,
+      "reward": 2.8240283727645874,
+      "reward_std": 0.34141383320093155,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.01972164958715439,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1327
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.5208435058594,
+      "epoch": 0.664,
+      "grad_norm": 20.757623767950136,
+      "kl": 3.8828125,
+      "learning_rate": 3.7561798609655373e-07,
+      "loss": -0.0051,
+      "reward": 2.598018765449524,
+      "reward_std": 0.724356472492218,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9444444179534912,
+      "rewards/repetition_penalty_reward": -0.018300842493772507,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1328
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 222.7291717529297,
+      "epoch": 0.6645,
+      "grad_norm": 21.24530948252979,
+      "kl": 4.796875,
+      "learning_rate": 3.748942224369073e-07,
+      "loss": -0.0478,
+      "reward": 2.4989744424819946,
+      "reward_std": 0.6526926159858704,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.016650736331939697,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1329
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 251.75001525878906,
+      "epoch": 0.665,
+      "grad_norm": 35.64687506708033,
+      "kl": 6.21875,
+      "learning_rate": 3.7417099217982686e-07,
+      "loss": -0.0819,
+      "reward": 2.379560112953186,
+      "reward_std": 0.7591554820537567,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.881944477558136,
+      "rewards/repetition_penalty_reward": -0.02321775909513235,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1330
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.9583435058594,
+      "epoch": 0.6655,
+      "grad_norm": 5.687701052224813,
+      "kl": 2.15234375,
+      "learning_rate": 3.734482975283975e-07,
+      "loss": -0.0797,
+      "reward": 2.6574783325195312,
+      "reward_std": 0.4438095688819885,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.026549477130174637,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1331
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 312.4791717529297,
+      "epoch": 0.666,
+      "grad_norm": 5.722398391516816,
+      "kl": 1.5,
+      "learning_rate": 3.72726140684072e-07,
+      "loss": -0.0211,
+      "reward": 2.485495686531067,
+      "reward_std": 0.32544514536857605,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.03186549246311188,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1332
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.3333435058594,
+      "epoch": 0.6665,
+      "grad_norm": 4.960361026761474,
+      "kl": 0.828125,
+      "learning_rate": 3.720045238466658e-07,
+      "loss": -0.0164,
+      "reward": 2.5775978565216064,
+      "reward_std": 0.36035288870334625,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.021360510494560003,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1333
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 246.4166717529297,
+      "epoch": 0.667,
+      "grad_norm": 6.897070121071065,
+      "kl": 2.5703125,
+      "learning_rate": 3.712834492143487e-07,
+      "loss": -0.088,
+      "reward": 2.4616823196411133,
+      "reward_std": 0.5930516719818115,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.022692805156111717,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1334
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 248.16667938232422,
+      "epoch": 0.6675,
+      "grad_norm": 4.566188182751088,
+      "kl": 1.150390625,
+      "learning_rate": 3.7056291898363925e-07,
+      "loss": -0.0822,
+      "reward": 2.779423713684082,
+      "reward_std": 0.4814029037952423,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.013978923205286264,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1335
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.1458435058594,
+      "epoch": 0.668,
+      "grad_norm": 8.6793215256614,
+      "kl": 1.9140625,
+      "learning_rate": 3.6984293534939737e-07,
+      "loss": -0.1695,
+      "reward": 2.448217272758484,
+      "reward_std": 0.7062622308731079,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.02053278312087059,
+      "rewards/tag_count_reward": 0.9270833730697632,
+      "step": 1336
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 272.2708435058594,
+      "epoch": 0.6685,
+      "grad_norm": 6.644806226411769,
+      "kl": 1.1875,
+      "learning_rate": 3.69123500504818e-07,
+      "loss": -0.0284,
+      "reward": 2.466250419616699,
+      "reward_std": 0.5085738003253937,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.028541546314954758,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1337
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.5416717529297,
+      "epoch": 0.669,
+      "grad_norm": 8.436837507583917,
+      "kl": 1.466796875,
+      "learning_rate": 3.6840461664142444e-07,
+      "loss": -0.0728,
+      "reward": 2.5604257583618164,
+      "reward_std": 0.6860225200653076,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.029852177016437054,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1338
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 281.06251525878906,
+      "epoch": 0.6695,
+      "grad_norm": 10.964332836550879,
+      "kl": 1.158203125,
+      "learning_rate": 3.6768628594906193e-07,
+      "loss": -0.0204,
+      "reward": 2.8616209030151367,
+      "reward_std": 0.35454705357551575,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.023795696906745434,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1339
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.00000762939453,
+      "epoch": 0.67,
+      "grad_norm": 6.098760743160745,
+      "kl": 1.83984375,
+      "learning_rate": 3.6696851061588994e-07,
+      "loss": -0.044,
+      "reward": 2.4193036556243896,
+      "reward_std": 0.4224669486284256,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.014724194537848234,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1340
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 263.6458435058594,
+      "epoch": 0.6705,
+      "grad_norm": 14.11730319397243,
+      "kl": 3.6484375,
+      "learning_rate": 3.6625129282837685e-07,
+      "loss": -0.0756,
+      "reward": 2.5297341346740723,
+      "reward_std": 0.5198619514703751,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.031029692851006985,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1341
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 302.2708435058594,
+      "epoch": 0.671,
+      "grad_norm": 5.630265152909872,
+      "kl": 1.7578125,
+      "learning_rate": 3.655346347712922e-07,
+      "loss": -0.0712,
+      "reward": 2.448781371116638,
+      "reward_std": 0.4181585758924484,
+      "rewards/accuracy_reward": 0.5625000223517418,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.025177019648253918,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1342
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.9166717529297,
+      "epoch": 0.6715,
+      "grad_norm": 13.132812973478561,
+      "kl": 4.578125,
+      "learning_rate": 3.6481853862770107e-07,
+      "loss": -0.1407,
+      "reward": 2.560381054878235,
+      "reward_std": 0.7369968295097351,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.028160739690065384,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1343
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.12501525878906,
+      "epoch": 0.672,
+      "grad_norm": 16.153270842061996,
+      "kl": 3.3125,
+      "learning_rate": 3.641030065789562e-07,
+      "loss": -0.173,
+      "reward": 2.4971349239349365,
+      "reward_std": 0.7496102154254913,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.01675412245094776,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1344
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 254.72917938232422,
+      "epoch": 0.6725,
+      "grad_norm": 13.478278088046087,
+      "kl": 2.83203125,
+      "learning_rate": 3.6338804080469253e-07,
+      "loss": -0.1632,
+      "reward": 2.4378998279571533,
+      "reward_std": 0.7053618729114532,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.013489224947988987,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1345
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.9375,
+      "epoch": 0.673,
+      "grad_norm": 11.718823360410598,
+      "kl": 0.984375,
+      "learning_rate": 3.6267364348281946e-07,
+      "loss": -0.0235,
+      "reward": 2.578608751296997,
+      "reward_std": 0.4408767372369766,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.02208577375859022,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1346
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.33333587646484,
+      "epoch": 0.6735,
+      "grad_norm": 6.659825601151128,
+      "kl": 1.31640625,
+      "learning_rate": 3.6195981678951535e-07,
+      "loss": -0.0251,
+      "reward": 2.802467465400696,
+      "reward_std": 0.32479756511747837,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.02739358227699995,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1347
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.2916793823242,
+      "epoch": 0.674,
+      "grad_norm": 5.94749137911606,
+      "kl": 1.763671875,
+      "learning_rate": 3.612465628992203e-07,
+      "loss": -0.1275,
+      "reward": 2.4290517568588257,
+      "reward_std": 0.6968129873275757,
+      "rewards/accuracy_reward": 0.5833333730697632,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.020601162686944008,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1348
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.43751525878906,
+      "epoch": 0.6745,
+      "grad_norm": 5.403140530841979,
+      "kl": 1.240234375,
+      "learning_rate": 3.60533883984629e-07,
+      "loss": -0.0099,
+      "reward": 2.505295157432556,
+      "reward_std": 0.556377038359642,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.020746564492583275,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1349
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.25,
+      "epoch": 0.675,
+      "grad_norm": 24.75284339330913,
+      "kl": 0.8349609375,
+      "learning_rate": 3.5982178221668533e-07,
+      "loss": -0.0314,
+      "reward": 2.6801000833511353,
+      "reward_std": 0.3536926209926605,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02476109704002738,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1350
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.06251525878906,
+      "epoch": 0.6755,
+      "grad_norm": 7.877722164133154,
+      "kl": 0.8818359375,
+      "learning_rate": 3.591102597645743e-07,
+      "loss": 0.0142,
+      "reward": 2.4543099403381348,
+      "reward_std": 0.3884861320257187,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.014439986553043127,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1351
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.60418701171875,
+      "epoch": 0.676,
+      "grad_norm": 9.392382547183344,
+      "kl": 0.9609375,
+      "learning_rate": 3.5839931879571725e-07,
+      "loss": 0.0275,
+      "reward": 2.4401715993881226,
+      "reward_std": 0.32545991241931915,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.0337868882343173,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1352
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.87501525878906,
+      "epoch": 0.6765,
+      "grad_norm": 7.560942470949891,
+      "kl": 1.6328125,
+      "learning_rate": 3.5768896147576344e-07,
+      "loss": -0.0042,
+      "reward": 2.3383556604385376,
+      "reward_std": 0.6181593537330627,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.02796376869082451,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1353
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.2708435058594,
+      "epoch": 0.677,
+      "grad_norm": 6.137807981682768,
+      "kl": 1.51171875,
+      "learning_rate": 3.5697918996858443e-07,
+      "loss": 0.0373,
+      "reward": 2.459334135055542,
+      "reward_std": 0.5140793472528458,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02156880311667919,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1354
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.35418701171875,
+      "epoch": 0.6775,
+      "grad_norm": 18.4925775719971,
+      "kl": 1.98046875,
+      "learning_rate": 3.5627000643626704e-07,
+      "loss": 0.0317,
+      "reward": 2.6040374040603638,
+      "reward_std": 0.40690936520695686,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02096256148070097,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1355
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.25,
+      "epoch": 0.678,
+      "grad_norm": 32.05231705546735,
+      "kl": 2.2890625,
+      "learning_rate": 3.555614130391079e-07,
+      "loss": -0.1021,
+      "reward": 2.4450795650482178,
+      "reward_std": 0.5791552066802979,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.02887882199138403,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1356
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.81251525878906,
+      "epoch": 0.6785,
+      "grad_norm": 11.77786900698309,
+      "kl": 1.3203125,
+      "learning_rate": 3.5485341193560503e-07,
+      "loss": -0.0302,
+      "reward": 2.523303985595703,
+      "reward_std": 0.5757817924022675,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.04093226231634617,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1357
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.8541717529297,
+      "epoch": 0.679,
+      "grad_norm": 5.649915735065612,
+      "kl": 0.642578125,
+      "learning_rate": 3.5414600528245266e-07,
+      "loss": 0.0445,
+      "reward": 2.7024723291397095,
+      "reward_std": 0.05420066323131323,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03711113706231117,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1358
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 269.7916717529297,
+      "epoch": 0.6795,
+      "grad_norm": 5.052148500574189,
+      "kl": 1.94140625,
+      "learning_rate": 3.534391952345341e-07,
+      "loss": -0.2133,
+      "reward": 2.543895125389099,
+      "reward_std": 0.7587614357471466,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9166667461395264,
+      "rewards/repetition_penalty_reward": -0.023813419975340366,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1359
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.64584350585938,
+      "epoch": 0.68,
+      "grad_norm": 19.199009856910283,
+      "kl": 3.46875,
+      "learning_rate": 3.5273298394491515e-07,
+      "loss": -0.0754,
+      "reward": 1.7231401801109314,
+      "reward_std": 0.5748586058616638,
+      "rewards/accuracy_reward": 0.125,
+      "rewards/reasoning_steps_reward": 0.8611111342906952,
+      "rewards/repetition_penalty_reward": -0.023387585766613483,
+      "rewards/tag_count_reward": 0.7604166865348816,
+      "step": 1360
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.62500762939453,
+      "epoch": 0.6805,
+      "grad_norm": 6.3532586709420995,
+      "kl": 2.1875,
+      "learning_rate": 3.5202737356483816e-07,
+      "loss": -0.1683,
+      "reward": 2.469906449317932,
+      "reward_std": 0.6009674370288849,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.02314911223948002,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1361
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.1666717529297,
+      "epoch": 0.681,
+      "grad_norm": 6.5401180752318036,
+      "kl": 2.05078125,
+      "learning_rate": 3.513223662437147e-07,
+      "loss": -0.1042,
+      "reward": 2.417715549468994,
+      "reward_std": 0.5224019438028336,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.028464973904192448,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1362
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 263.5833435058594,
+      "epoch": 0.6815,
+      "grad_norm": 6.66636418634202,
+      "kl": 1.2109375,
+      "learning_rate": 3.5061796412911913e-07,
+      "loss": -0.0175,
+      "reward": 2.8467352390289307,
+      "reward_std": 0.4228966236114502,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.016112081240862608,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1363
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 288.7083435058594,
+      "epoch": 0.682,
+      "grad_norm": 5.548234698164087,
+      "kl": 0.7255859375,
+      "learning_rate": 3.4991416936678276e-07,
+      "loss": 0.0183,
+      "reward": 2.5735777616500854,
+      "reward_std": 0.2825750932097435,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.0288529209792614,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1364
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 302.5208435058594,
+      "epoch": 0.6825,
+      "grad_norm": 7.01523336912401,
+      "kl": 1.494140625,
+      "learning_rate": 3.49210984100586e-07,
+      "loss": -0.0757,
+      "reward": 2.2636520862579346,
+      "reward_std": 0.5402050614356995,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.02627858892083168,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1365
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.4375,
+      "epoch": 0.683,
+      "grad_norm": 10.223491193197818,
+      "kl": 2.58203125,
+      "learning_rate": 3.4850841047255364e-07,
+      "loss": 0.0061,
+      "reward": 2.299806833267212,
+      "reward_std": 0.5171876847743988,
+      "rewards/accuracy_reward": 0.4166666716337204,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.01963762380182743,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1366
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 306.25,
+      "epoch": 0.6835,
+      "grad_norm": 3.9777954578862067,
+      "kl": 0.681640625,
+      "learning_rate": 3.4780645062284665e-07,
+      "loss": 0.049,
+      "reward": 2.467653274536133,
+      "reward_std": 0.14734814129769802,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.025402450934052467,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1367
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 299.4791717529297,
+      "epoch": 0.684,
+      "grad_norm": 7.507696165222269,
+      "kl": 1.6875,
+      "learning_rate": 3.471051066897562e-07,
+      "loss": -0.0996,
+      "reward": 2.167901575565338,
+      "reward_std": 0.4075617045164108,
+      "rewards/accuracy_reward": 0.2916666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.024806811474263668,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1368
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.7083435058594,
+      "epoch": 0.6845,
+      "grad_norm": 6.142733918004118,
+      "kl": 2.0546875,
+      "learning_rate": 3.4640438080969773e-07,
+      "loss": -0.1285,
+      "reward": 2.334149956703186,
+      "reward_std": 0.4152573347091675,
+      "rewards/accuracy_reward": 0.4375000149011612,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.02175288973376155,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1369
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 281.2708435058594,
+      "epoch": 0.685,
+      "grad_norm": 19.072228574321905,
+      "kl": 2.3271484375,
+      "learning_rate": 3.45704275117204e-07,
+      "loss": 0.064,
+      "reward": 2.5171972513198853,
+      "reward_std": 0.44150757789611816,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.03141411580145359,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1370
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.72918701171875,
+      "epoch": 0.6855,
+      "grad_norm": 3.7189442550634975,
+      "kl": 0.521484375,
+      "learning_rate": 3.450047917449181e-07,
+      "loss": 0.0285,
+      "reward": 2.536484122276306,
+      "reward_std": 0.20072168111801147,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02601606771349907,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1371
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 248.20833587646484,
+      "epoch": 0.686,
+      "grad_norm": 7.372964659791776,
+      "kl": 1.83984375,
+      "learning_rate": 3.4430593282358777e-07,
+      "loss": -0.0345,
+      "reward": 2.7801530361175537,
+      "reward_std": 0.5550233423709869,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.02540257666260004,
+      "rewards/tag_count_reward": 0.9583333730697632,
+      "step": 1372
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 272.66668701171875,
+      "epoch": 0.6865,
+      "grad_norm": 11.180474899161313,
+      "kl": 1.767578125,
+      "learning_rate": 3.4360770048205843e-07,
+      "loss": 0.0396,
+      "reward": 2.6890029907226562,
+      "reward_std": 0.49107812345027924,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.024538762867450714,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1373
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.4791717529297,
+      "epoch": 0.687,
+      "grad_norm": 7.7693979337518275,
+      "kl": 0.7421875,
+      "learning_rate": 3.429100968472668e-07,
+      "loss": 0.0575,
+      "reward": 2.973837971687317,
+      "reward_std": 0.017366621643304825,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02616224344819784,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1374
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.4166717529297,
+      "epoch": 0.6875,
+      "grad_norm": 3.5163971736947826,
+      "kl": 1.734375,
+      "learning_rate": 3.4221312404423486e-07,
+      "loss": -0.0705,
+      "reward": 2.387254238128662,
+      "reward_std": 0.3396844193339348,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.025940215215086937,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1375
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.5208435058594,
+      "epoch": 0.688,
+      "grad_norm": 4.361183296818996,
+      "kl": 0.716796875,
+      "learning_rate": 3.4151678419606233e-07,
+      "loss": 0.007,
+      "reward": 2.502146601676941,
+      "reward_std": 0.3024323433637619,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.02215902181342244,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1376
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 254.18750762939453,
+      "epoch": 0.6885,
+      "grad_norm": 5.83748552522261,
+      "kl": 2.4140625,
+      "learning_rate": 3.4082107942392136e-07,
+      "loss": -0.0837,
+      "reward": 2.4153069257736206,
+      "reward_std": 0.5053312480449677,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.03955421969294548,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1377
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 285.0416717529297,
+      "epoch": 0.689,
+      "grad_norm": 5.928807362280472,
+      "kl": 0.73046875,
+      "learning_rate": 3.4012601184704904e-07,
+      "loss": 0.0042,
+      "reward": 2.7761462926864624,
+      "reward_std": 0.2563507854938507,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.020728619303554296,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1378
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 255.58334350585938,
+      "epoch": 0.6895,
+      "grad_norm": 6.1951381664884595,
+      "kl": 1.62890625,
+      "learning_rate": 3.3943158358274203e-07,
+      "loss": -0.1059,
+      "reward": 2.466705560684204,
+      "reward_std": 0.49967700242996216,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.029822363518178463,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1379
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 300.8333435058594,
+      "epoch": 0.69,
+      "grad_norm": 6.849165033673041,
+      "kl": 0.73828125,
+      "learning_rate": 3.387377967463493e-07,
+      "loss": 0.0257,
+      "reward": 2.7919795513153076,
+      "reward_std": 0.2739688716828823,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.030937272123992443,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1380
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 253.33334350585938,
+      "epoch": 0.6905,
+      "grad_norm": 4.414135015976259,
+      "kl": 0.666015625,
+      "learning_rate": 3.3804465345126545e-07,
+      "loss": 0.0039,
+      "reward": 2.940259575843811,
+      "reward_std": 0.16437675757333636,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.014601573813706636,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1381
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.1458435058594,
+      "epoch": 0.691,
+      "grad_norm": 4.724734711788012,
+      "kl": 0.767578125,
+      "learning_rate": 3.3735215580892575e-07,
+      "loss": -0.1102,
+      "reward": 2.497297525405884,
+      "reward_std": 0.42764711380004883,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.02179985586553812,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1382
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 272.125,
+      "epoch": 0.6915,
+      "grad_norm": 6.326351779698143,
+      "kl": 0.69921875,
+      "learning_rate": 3.366603059287977e-07,
+      "loss": -0.0165,
+      "reward": 2.3360326290130615,
+      "reward_std": 0.4005787819623947,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02334247063845396,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1383
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.3541793823242,
+      "epoch": 0.692,
+      "grad_norm": 7.84619977481907,
+      "kl": 1.3828125,
+      "learning_rate": 3.359691059183761e-07,
+      "loss": -0.0385,
+      "reward": 2.366807222366333,
+      "reward_std": 0.606653481721878,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.025554182939231396,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1384
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.5625,
+      "epoch": 0.6925,
+      "grad_norm": 6.807191768545,
+      "kl": 0.876953125,
+      "learning_rate": 3.3527855788317614e-07,
+      "loss": 0.0225,
+      "reward": 2.5943243503570557,
+      "reward_std": 0.487922340631485,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.0341478455811739,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1385
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.04168701171875,
+      "epoch": 0.693,
+      "grad_norm": 9.428973328086897,
+      "kl": 0.693359375,
+      "learning_rate": 3.3458866392672694e-07,
+      "loss": -0.0436,
+      "reward": 2.774644374847412,
+      "reward_std": 0.39781980216503143,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.01875837054103613,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1386
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 286.7708435058594,
+      "epoch": 0.6935,
+      "grad_norm": 6.33506224802435,
+      "kl": 0.98046875,
+      "learning_rate": 3.338994261505649e-07,
+      "loss": -0.0637,
+      "reward": 2.1281405091285706,
+      "reward_std": 0.37489962577819824,
+      "rewards/accuracy_reward": 0.2291666716337204,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.026373423635959625,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1387
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 301.1666793823242,
+      "epoch": 0.694,
+      "grad_norm": 3.780779728016683,
+      "kl": 0.5712890625,
+      "learning_rate": 3.3321084665422803e-07,
+      "loss": -0.0363,
+      "reward": 2.5015201568603516,
+      "reward_std": 0.32130593061447144,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.026257574558258057,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1388
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.0416717529297,
+      "epoch": 0.6945,
+      "grad_norm": 4.691021479308191,
+      "kl": 1.3203125,
+      "learning_rate": 3.325229275352489e-07,
+      "loss": -0.0273,
+      "reward": 2.5804378986358643,
+      "reward_std": 0.4816815108060837,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.0237288074567914,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1389
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.50001525878906,
+      "epoch": 0.695,
+      "grad_norm": 5.62934560927229,
+      "kl": 0.6484375,
+      "learning_rate": 3.3183567088914833e-07,
+      "loss": 0.0106,
+      "reward": 2.7332091331481934,
+      "reward_std": 0.08886073343455791,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02720770798623562,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1390
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.4375,
+      "epoch": 0.6955,
+      "grad_norm": 6.2652786345135985,
+      "kl": 0.77734375,
+      "learning_rate": 3.3114907880942933e-07,
+      "loss": 0.0052,
+      "reward": 2.7369773387908936,
+      "reward_std": 0.24390956666320562,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.023439462296664715,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1391
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.5208435058594,
+      "epoch": 0.696,
+      "grad_norm": 4.763664401378893,
+      "kl": 0.587890625,
+      "learning_rate": 3.3046315338757026e-07,
+      "loss": 0.0138,
+      "reward": 2.908858060836792,
+      "reward_std": 0.17314723134040833,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.028641941025853157,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1392
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.62500762939453,
+      "epoch": 0.6965,
+      "grad_norm": 32.37138858122824,
+      "kl": 3.26953125,
+      "learning_rate": 3.297778967130191e-07,
+      "loss": 0.0044,
+      "reward": 2.3406665325164795,
+      "reward_std": 0.43019232153892517,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.0187085154466331,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1393
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 285.6458435058594,
+      "epoch": 0.697,
+      "grad_norm": 3.700595698109454,
+      "kl": 0.462890625,
+      "learning_rate": 3.290933108731866e-07,
+      "loss": 0.03,
+      "reward": 2.709470510482788,
+      "reward_std": 0.254656121134758,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.04052966833114624,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1394
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 269.93750762939453,
+      "epoch": 0.6975,
+      "grad_norm": 17.617668970781107,
+      "kl": 3.083984375,
+      "learning_rate": 3.2840939795343987e-07,
+      "loss": -0.0307,
+      "reward": 2.570359468460083,
+      "reward_std": 0.3866235390305519,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.025126777589321136,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1395
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.7916717529297,
+      "epoch": 0.698,
+      "grad_norm": 5.286876953487483,
+      "kl": 1.009765625,
+      "learning_rate": 3.2772616003709616e-07,
+      "loss": -0.0073,
+      "reward": 2.618911862373352,
+      "reward_std": 0.3726077973842621,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.016504944302141666,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1396
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 245.6875,
+      "epoch": 0.6985,
+      "grad_norm": 11.84313036080623,
+      "kl": 2.0234375,
+      "learning_rate": 3.270435992054166e-07,
+      "loss": -0.0171,
+      "reward": 2.6750162839889526,
+      "reward_std": 0.4495895653963089,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.009011534042656422,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1397
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 266.5208435058594,
+      "epoch": 0.699,
+      "grad_norm": 4.6172662372737525,
+      "kl": 0.720703125,
+      "learning_rate": 3.263617175376001e-07,
+      "loss": -0.0352,
+      "reward": 2.491300940513611,
+      "reward_std": 0.3568393215537071,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.019115855917334557,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1398
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 312.0208435058594,
+      "epoch": 0.6995,
+      "grad_norm": 6.661016856595855,
+      "kl": 1.9296875,
+      "learning_rate": 3.2568051711077636e-07,
+      "loss": 0.0014,
+      "reward": 2.4803801774978638,
+      "reward_std": 0.2543382793664932,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.024828164838254452,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1399
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 245.64584350585938,
+      "epoch": 0.7,
+      "grad_norm": 7.280485227211617,
+      "kl": 1.5078125,
+      "learning_rate": 3.250000000000001e-07,
+      "loss": 0.0254,
+      "reward": 2.686321496963501,
+      "reward_std": 0.4712058752775192,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.011595276184380054,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1400
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 247.22917938232422,
+      "epoch": 0.7005,
+      "grad_norm": 5.073476939105119,
+      "kl": 0.724609375,
+      "learning_rate": 3.2432016827824414e-07,
+      "loss": -0.0513,
+      "reward": 2.620236396789551,
+      "reward_std": 0.33665385842323303,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02038868237286806,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1401
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.4791717529297,
+      "epoch": 0.701,
+      "grad_norm": 14.591352187407846,
+      "kl": 1.109375,
+      "learning_rate": 3.2364102401639423e-07,
+      "loss": 0.0143,
+      "reward": 2.708950996398926,
+      "reward_std": 0.34444018453359604,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.01674342295154929,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1402
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 250.79167938232422,
+      "epoch": 0.7015,
+      "grad_norm": 4.900616214888936,
+      "kl": 1.1083984375,
+      "learning_rate": 3.229625692832414e-07,
+      "loss": -0.008,
+      "reward": 2.674125075340271,
+      "reward_std": 0.4905224144458771,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.020319399423897266,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1403
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 281.2708435058594,
+      "epoch": 0.702,
+      "grad_norm": 3.8548797240419503,
+      "kl": 0.87109375,
+      "learning_rate": 3.222848061454764e-07,
+      "loss": -0.0311,
+      "reward": 2.7337580919265747,
+      "reward_std": 0.4215731620788574,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.017978041898459196,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1404
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 258.8333435058594,
+      "epoch": 0.7025,
+      "grad_norm": 7.245780935291608,
+      "kl": 0.900390625,
+      "learning_rate": 3.216077366676833e-07,
+      "loss": -0.0148,
+      "reward": 2.862648129463196,
+      "reward_std": 0.3005357086658478,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0331853311508894,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1405
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 277.7083435058594,
+      "epoch": 0.703,
+      "grad_norm": 8.179242595228185,
+      "kl": 1.078125,
+      "learning_rate": 3.209313629123329e-07,
+      "loss": 0.0065,
+      "reward": 2.753788113594055,
+      "reward_std": 0.3614218980073929,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0378786064684391,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1406
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 301.5208435058594,
+      "epoch": 0.7035,
+      "grad_norm": 8.893778669993276,
+      "kl": 2.21875,
+      "learning_rate": 3.2025568693977745e-07,
+      "loss": -0.0732,
+      "reward": 2.5992143154144287,
+      "reward_std": 0.45963357388973236,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.03446643240749836,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1407
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.4791717529297,
+      "epoch": 0.704,
+      "grad_norm": 19.76315212110327,
+      "kl": 4.63671875,
+      "learning_rate": 3.195807108082429e-07,
+      "loss": -0.1105,
+      "reward": 2.3212958574295044,
+      "reward_std": 0.6959238648414612,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.018981934990733862,
+      "rewards/tag_count_reward": 0.8958333730697632,
+      "step": 1408
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.7708435058594,
+      "epoch": 0.7045,
+      "grad_norm": 7.783421626644704,
+      "kl": 2.55859375,
+      "learning_rate": 3.1890643657382356e-07,
+      "loss": -0.1032,
+      "reward": 2.7160589694976807,
+      "reward_std": 0.5336126536130905,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.028732833918184042,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1409
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 212.7291717529297,
+      "epoch": 0.705,
+      "grad_norm": 13.071825269979621,
+      "kl": 3.9296875,
+      "learning_rate": 3.182328662904756e-07,
+      "loss": -0.0981,
+      "reward": 2.5850327014923096,
+      "reward_std": 0.6792955249547958,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.02087017334997654,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1410
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 227.12500762939453,
+      "epoch": 0.7055,
+      "grad_norm": 16.647908802505288,
+      "kl": 3.201171875,
+      "learning_rate": 3.175600020100112e-07,
+      "loss": -0.0142,
+      "reward": 2.422129988670349,
+      "reward_std": 0.3781207883730531,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.011897902470082045,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1411
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 251.8541717529297,
+      "epoch": 0.706,
+      "grad_norm": 4.246659669908856,
+      "kl": 2.57421875,
+      "learning_rate": 3.168878457820915e-07,
+      "loss": -0.193,
+      "reward": 2.498375415802002,
+      "reward_std": 0.7053702175617218,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.902777761220932,
+      "rewards/repetition_penalty_reward": -0.024194156285375357,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1412
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.00001525878906,
+      "epoch": 0.7065,
+      "grad_norm": 4.186288561909542,
+      "kl": 0.861328125,
+      "learning_rate": 3.162163996542209e-07,
+      "loss": -0.1177,
+      "reward": 2.7583051919937134,
+      "reward_std": 0.5227586776018143,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04204211104661226,
+      "rewards/tag_count_reward": 0.9739583730697632,
+      "step": 1413
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 295.8541717529297,
+      "epoch": 0.707,
+      "grad_norm": 4.18530097815396,
+      "kl": 0.830078125,
+      "learning_rate": 3.155456656717408e-07,
+      "loss": -0.0055,
+      "reward": 2.917997360229492,
+      "reward_std": 0.1906261881813407,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.029919438064098358,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1414
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 229.52083587646484,
+      "epoch": 0.7075,
+      "grad_norm": 7.3471907992432515,
+      "kl": 1.296875,
+      "learning_rate": 3.14875645877823e-07,
+      "loss": -0.1006,
+      "reward": 2.7960588932037354,
+      "reward_std": 0.5089210569858551,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.028593890368938446,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1415
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.8333435058594,
+      "epoch": 0.708,
+      "grad_norm": 5.66147418702201,
+      "kl": 1.16015625,
+      "learning_rate": 3.142063423134644e-07,
+      "loss": -0.0736,
+      "reward": 2.4612691402435303,
+      "reward_std": 0.4342493712902069,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.026578163262456656,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1416
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 277.2708435058594,
+      "epoch": 0.7085,
+      "grad_norm": 4.258559874095392,
+      "kl": 0.826171875,
+      "learning_rate": 3.135377570174796e-07,
+      "loss": 0.0009,
+      "reward": 2.634892225265503,
+      "reward_std": 0.32757639279589057,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.02830233983695507,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1417
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 258.3333435058594,
+      "epoch": 0.709,
+      "grad_norm": 5.069921662523948,
+      "kl": 2.03515625,
+      "learning_rate": 3.1286989202649503e-07,
+      "loss": -0.0317,
+      "reward": 2.490046262741089,
+      "reward_std": 0.43885305523872375,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.027314997278153896,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1418
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.12500762939453,
+      "epoch": 0.7095,
+      "grad_norm": 6.797930142073019,
+      "kl": 3.453125,
+      "learning_rate": 3.122027493749438e-07,
+      "loss": -0.1656,
+      "reward": 1.9841939210891724,
+      "reward_std": 0.6511839628219604,
+      "rewards/accuracy_reward": 0.2708333432674408,
+      "rewards/reasoning_steps_reward": 0.8611111044883728,
+      "rewards/repetition_penalty_reward": -0.02275061421096325,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1419
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 256.3125,
+      "epoch": 0.71,
+      "grad_norm": 7.202911096329466,
+      "kl": 3.12109375,
+      "learning_rate": 3.115363310950578e-07,
+      "loss": -0.1218,
+      "reward": 2.4370113611221313,
+      "reward_std": 0.6423384845256805,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.8750000298023224,
+      "rewards/repetition_penalty_reward": -0.016113675199449062,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1420
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 248.43750762939453,
+      "epoch": 0.7105,
+      "grad_norm": 7.687028342637116,
+      "kl": 2.15625,
+      "learning_rate": 3.1087063921686263e-07,
+      "loss": -0.0065,
+      "reward": 2.5198620557785034,
+      "reward_std": 0.6163989156484604,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.023540794849395752,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1421
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.33333587646484,
+      "epoch": 0.711,
+      "grad_norm": 6.503537489968444,
+      "kl": 1.65625,
+      "learning_rate": 3.102056757681715e-07,
+      "loss": -0.0039,
+      "reward": 2.3564590215682983,
+      "reward_std": 0.3768700957298279,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03243005648255348,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1422
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.125,
+      "epoch": 0.7115,
+      "grad_norm": 13.267307709264951,
+      "kl": 2.6328125,
+      "learning_rate": 3.0954144277457817e-07,
+      "loss": -0.0765,
+      "reward": 2.4606250524520874,
+      "reward_std": 0.48642072081565857,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.009861073223873973,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1423
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.93750762939453,
+      "epoch": 0.712,
+      "grad_norm": 7.851695017414299,
+      "kl": 1.34765625,
+      "learning_rate": 3.0887794225945143e-07,
+      "loss": -0.0492,
+      "reward": 2.350769519805908,
+      "reward_std": 0.5201582908630371,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.036383312195539474,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1424
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 234.89584350585938,
+      "epoch": 0.7125,
+      "grad_norm": 8.683593317223446,
+      "kl": 2.890625,
+      "learning_rate": 3.0821517624392925e-07,
+      "loss": -0.069,
+      "reward": 2.6100372076034546,
+      "reward_std": 0.6106880903244019,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.030587902292609215,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1425
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 239.77084350585938,
+      "epoch": 0.713,
+      "grad_norm": 16.77464036064384,
+      "kl": 3.6015625,
+      "learning_rate": 3.075531467469116e-07,
+      "loss": -0.0936,
+      "reward": 2.576740264892578,
+      "reward_std": 0.7488699555397034,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.013537466991692781,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1426
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 327.06251525878906,
+      "epoch": 0.7135,
+      "grad_norm": 7.518467200296574,
+      "kl": 1.0859375,
+      "learning_rate": 3.0689185578505525e-07,
+      "loss": -0.0554,
+      "reward": 2.6223039627075195,
+      "reward_std": 0.5526255667209625,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.027001574635505676,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1427
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 276.62501525878906,
+      "epoch": 0.714,
+      "grad_norm": 8.336081909558045,
+      "kl": 1.48828125,
+      "learning_rate": 3.062313053727671e-07,
+      "loss": -0.0128,
+      "reward": 2.6898101568222046,
+      "reward_std": 0.5765750408172607,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.016787134110927582,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1428
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 251.14583587646484,
+      "epoch": 0.7145,
+      "grad_norm": 13.511892818726327,
+      "kl": 2.55859375,
+      "learning_rate": 3.055714975221981e-07,
+      "loss": 0.0218,
+      "reward": 2.5956475734710693,
+      "reward_std": 0.34704746305942535,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.022408071905374527,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1429
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.8541793823242,
+      "epoch": 0.715,
+      "grad_norm": 9.586287180266988,
+      "kl": 1.05078125,
+      "learning_rate": 3.0491243424323783e-07,
+      "loss": 0.0298,
+      "reward": 2.318286895751953,
+      "reward_std": 0.4003664702177048,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03414386324584484,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1430
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 336.4583435058594,
+      "epoch": 0.7155,
+      "grad_norm": 4.201243088427591,
+      "kl": 1.953125,
+      "learning_rate": 3.0425411754350694e-07,
+      "loss": -0.0551,
+      "reward": 2.253325581550598,
+      "reward_std": 0.4799940884113312,
+      "rewards/accuracy_reward": 0.39583333395421505,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.03660505823791027,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1431
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 261.1458435058594,
+      "epoch": 0.716,
+      "grad_norm": 11.754368643179218,
+      "kl": 2.6796875,
+      "learning_rate": 3.0359654942835247e-07,
+      "loss": 0.0551,
+      "reward": 2.5765037536621094,
+      "reward_std": 0.49608848989009857,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.02766304276883602,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1432
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.6458435058594,
+      "epoch": 0.7165,
+      "grad_norm": 4.101040359968373,
+      "kl": 0.55078125,
+      "learning_rate": 3.029397319008407e-07,
+      "loss": -0.0074,
+      "reward": 2.9076608419418335,
+      "reward_std": 0.1762450411915779,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.029839315451681614,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1433
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.2083435058594,
+      "epoch": 0.717,
+      "grad_norm": 5.6015729375673065,
+      "kl": 0.94140625,
+      "learning_rate": 3.02283666961752e-07,
+      "loss": -0.0133,
+      "reward": 2.6073429584503174,
+      "reward_std": 0.20822132378816605,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.028073765337467194,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1434
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.7708435058594,
+      "epoch": 0.7175,
+      "grad_norm": 13.758583049621224,
+      "kl": 3.21875,
+      "learning_rate": 3.016283566095739e-07,
+      "loss": -0.0096,
+      "reward": 2.5433467626571655,
+      "reward_std": 0.5455273687839508,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.024361703544855118,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1435
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.7708435058594,
+      "epoch": 0.718,
+      "grad_norm": 6.637973784958379,
+      "kl": 1.078125,
+      "learning_rate": 3.0097380284049523e-07,
+      "loss": -0.0344,
+      "reward": 2.5512558221817017,
+      "reward_std": 0.4321069046854973,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.021661010570824146,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1436
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 256.9166793823242,
+      "epoch": 0.7185,
+      "grad_norm": 9.5020255723838,
+      "kl": 2.4609375,
+      "learning_rate": 3.003200076484004e-07,
+      "loss": -0.0598,
+      "reward": 2.710210919380188,
+      "reward_std": 0.5970990657806396,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.8958334028720856,
+      "rewards/repetition_penalty_reward": -0.018955985084176064,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1437
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.14583587646484,
+      "epoch": 0.719,
+      "grad_norm": 6.729758300453091,
+      "kl": 1.2109375,
+      "learning_rate": 2.996669730248628e-07,
+      "loss": 0.0368,
+      "reward": 2.680380702018738,
+      "reward_std": 0.4900428205728531,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.022744507528841496,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1438
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 297.37501525878906,
+      "epoch": 0.7195,
+      "grad_norm": 6.100933401861528,
+      "kl": 1.5703125,
+      "learning_rate": 2.9901470095913943e-07,
+      "loss": -0.139,
+      "reward": 2.433695912361145,
+      "reward_std": 0.47993409633636475,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.033318111672997475,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1439
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 255.10417938232422,
+      "epoch": 0.72,
+      "grad_norm": 17.639971258796862,
+      "kl": 4.703125,
+      "learning_rate": 2.9836319343816397e-07,
+      "loss": -0.0614,
+      "reward": 2.4460188150405884,
+      "reward_std": 0.6759764552116394,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.01925905141979456,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1440
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 379.8541717529297,
+      "epoch": 0.7205,
+      "grad_norm": 6.145673583189092,
+      "kl": 1.65625,
+      "learning_rate": 2.977124524465413e-07,
+      "loss": -0.0369,
+      "reward": 2.5621336698532104,
+      "reward_std": 0.4305399991571903,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.029880317859351635,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1441
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.8333435058594,
+      "epoch": 0.721,
+      "grad_norm": 22.556717516314336,
+      "kl": 3.953125,
+      "learning_rate": 2.9706247996654134e-07,
+      "loss": 0.046,
+      "reward": 2.0625728368759155,
+      "reward_std": 0.7029457688331604,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.8750000298023224,
+      "rewards/repetition_penalty_reward": -0.02076055482029915,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1442
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 326.5833435058594,
+      "epoch": 0.7215,
+      "grad_norm": 9.337104428836039,
+      "kl": 1.0546875,
+      "learning_rate": 2.964132779780929e-07,
+      "loss": 0.0703,
+      "reward": 2.6360541582107544,
+      "reward_std": 0.4235610216856003,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.0358209777623415,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1443
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.29168701171875,
+      "epoch": 0.722,
+      "grad_norm": 14.451017089953675,
+      "kl": 1.880859375,
+      "learning_rate": 2.9576484845877793e-07,
+      "loss": -0.0315,
+      "reward": 2.36261785030365,
+      "reward_std": 0.4776066839694977,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.029743347316980362,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1444
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.62501525878906,
+      "epoch": 0.7225,
+      "grad_norm": 9.176318494305416,
+      "kl": 1.43359375,
+      "learning_rate": 2.9511719338382535e-07,
+      "loss": -0.0615,
+      "reward": 2.627773642539978,
+      "reward_std": 0.6298104226589203,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.019795984961092472,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1445
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.5416717529297,
+      "epoch": 0.723,
+      "grad_norm": 10.455365880612927,
+      "kl": 2.8125,
+      "learning_rate": 2.944703147261046e-07,
+      "loss": -0.1501,
+      "reward": 2.3291326761245728,
+      "reward_std": 0.7611142992973328,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.902777910232544,
+      "rewards/repetition_penalty_reward": -0.03197855316102505,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1446
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 261.89583587646484,
+      "epoch": 0.7235,
+      "grad_norm": 5.3123582713501944,
+      "kl": 1.1015625,
+      "learning_rate": 2.938242144561201e-07,
+      "loss": 0.055,
+      "reward": 2.4350517988204956,
+      "reward_std": 0.3026663661003113,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.016337126959115267,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1447
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.6041717529297,
+      "epoch": 0.724,
+      "grad_norm": 10.893245730461343,
+      "kl": 1.38671875,
+      "learning_rate": 2.931788945420058e-07,
+      "loss": -0.043,
+      "reward": 2.7075858116149902,
+      "reward_std": 0.4429877921938896,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.021580896340310574,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1448
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.8958435058594,
+      "epoch": 0.7245,
+      "grad_norm": 7.744283708478767,
+      "kl": 1.791015625,
+      "learning_rate": 2.925343569495178e-07,
+      "loss": -0.053,
+      "reward": 2.6358524560928345,
+      "reward_std": 0.5802061557769775,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03602285124361515,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1449
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.2291717529297,
+      "epoch": 0.725,
+      "grad_norm": 6.348495492774551,
+      "kl": 1.56640625,
+      "learning_rate": 2.918906036420294e-07,
+      "loss": -0.0218,
+      "reward": 2.6777302026748657,
+      "reward_std": 0.554046094417572,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.023658874444663525,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1450
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.7291717529297,
+      "epoch": 0.7255,
+      "grad_norm": 12.67301139932153,
+      "kl": 2.2890625,
+      "learning_rate": 2.9124763658052474e-07,
+      "loss": 0.0824,
+      "reward": 2.542311429977417,
+      "reward_std": 0.4715491533279419,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.03234157059341669,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1451
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.2708435058594,
+      "epoch": 0.726,
+      "grad_norm": 6.540507187939509,
+      "kl": 1.77734375,
+      "learning_rate": 2.9060545772359305e-07,
+      "loss": -0.0138,
+      "reward": 2.2951096296310425,
+      "reward_std": 0.7114408314228058,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02259882539510727,
+      "rewards/tag_count_reward": 0.8802083730697632,
+      "step": 1452
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 316.4583435058594,
+      "epoch": 0.7265,
+      "grad_norm": 3.9719963516737162,
+      "kl": 1.443359375,
+      "learning_rate": 2.8996406902742267e-07,
+      "loss": 0.0444,
+      "reward": 2.827430486679077,
+      "reward_std": 0.2230637795291841,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.018055669963359833,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1453
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.6666717529297,
+      "epoch": 0.727,
+      "grad_norm": 7.2887085683958945,
+      "kl": 1.142578125,
+      "learning_rate": 2.893234724457946e-07,
+      "loss": 0.0405,
+      "reward": 2.776036858558655,
+      "reward_std": 0.2234293557703495,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.024310494773089886,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1454
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 297.56251525878906,
+      "epoch": 0.7275,
+      "grad_norm": 7.258259475742343,
+      "kl": 1.33203125,
+      "learning_rate": 2.886836699300771e-07,
+      "loss": 0.0238,
+      "reward": 2.5538902282714844,
+      "reward_std": 0.48146432638168335,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666567325592,
+      "rewards/repetition_penalty_reward": -0.029443158768117428,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1455
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 288.2083435058594,
+      "epoch": 0.728,
+      "grad_norm": 4.296720231595159,
+      "kl": 0.537109375,
+      "learning_rate": 2.8804466342921987e-07,
+      "loss": 0.0545,
+      "reward": 2.8775835037231445,
+      "reward_std": 0.20204564975574613,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.018249690299853683,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1456
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 316.5625,
+      "epoch": 0.7285,
+      "grad_norm": 14.229710817664285,
+      "kl": 1.59375,
+      "learning_rate": 2.874064548897472e-07,
+      "loss": 0.0002,
+      "reward": 2.417072057723999,
+      "reward_std": 0.6614536046981812,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.027372430078685284,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1457
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 286.0833435058594,
+      "epoch": 0.729,
+      "grad_norm": 6.574513295769042,
+      "kl": 3.15625,
+      "learning_rate": 2.86769046255753e-07,
+      "loss": -0.0074,
+      "reward": 2.3303749561309814,
+      "reward_std": 0.8391402065753937,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.030736176297068596,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1458
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 303.5833435058594,
+      "epoch": 0.7295,
+      "grad_norm": 5.801082971170413,
+      "kl": 0.763671875,
+      "learning_rate": 2.8613243946889477e-07,
+      "loss": 0.0801,
+      "reward": 2.7063515186309814,
+      "reward_std": 0.49810338020324707,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.034967850893735886,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1459
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.06250762939453,
+      "epoch": 0.73,
+      "grad_norm": 6.729899483661514,
+      "kl": 1.302734375,
+      "learning_rate": 2.854966364683872e-07,
+      "loss": -0.0084,
+      "reward": 2.4949501752853394,
+      "reward_std": 0.4440324157476425,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.015466715674847364,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1460
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.18750762939453,
+      "epoch": 0.7305,
+      "grad_norm": 25.447643773212334,
+      "kl": 4.2109375,
+      "learning_rate": 2.848616391909959e-07,
+      "loss": 0.0692,
+      "reward": 2.4095431566238403,
+      "reward_std": 0.4361160099506378,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.024484614841639996,
+      "rewards/tag_count_reward": 0.9270833730697632,
+      "step": 1461
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.1666793823242,
+      "epoch": 0.731,
+      "grad_norm": 20.974946344266833,
+      "kl": 3.1328125,
+      "learning_rate": 2.842274495710335e-07,
+      "loss": 0.1305,
+      "reward": 2.630603075027466,
+      "reward_std": 0.5745793282985687,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.03953584283590317,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1462
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 301.4791717529297,
+      "epoch": 0.7315,
+      "grad_norm": 50.08951516994972,
+      "kl": 3.041015625,
+      "learning_rate": 2.835940695403512e-07,
+      "loss": 0.1238,
+      "reward": 2.7686582803726196,
+      "reward_std": 0.3026948422193527,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.023008490912616253,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1463
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.18751525878906,
+      "epoch": 0.732,
+      "grad_norm": 9.460276151267854,
+      "kl": 2.908203125,
+      "learning_rate": 2.829615010283344e-07,
+      "loss": 0.0303,
+      "reward": 2.6455318927764893,
+      "reward_std": 0.5820550620555878,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.03502368927001953,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1464
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 275.1041717529297,
+      "epoch": 0.7325,
+      "grad_norm": 10.796804080704874,
+      "kl": 2.96875,
+      "learning_rate": 2.8232974596189653e-07,
+      "loss": 0.0181,
+      "reward": 2.4227362871170044,
+      "reward_std": 0.4896738827228546,
+      "rewards/accuracy_reward": 0.5416666716337204,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.019972197711467743,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1465
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.0416793823242,
+      "epoch": 0.733,
+      "grad_norm": 18.4213130469492,
+      "kl": 3.0,
+      "learning_rate": 2.8169880626547283e-07,
+      "loss": 0.1161,
+      "reward": 2.2570921182632446,
+      "reward_std": 0.6915050446987152,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.022421800531446934,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1466
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 299.1875,
+      "epoch": 0.7335,
+      "grad_norm": 7.219616999713985,
+      "kl": 1.9921875,
+      "learning_rate": 2.8106868386101545e-07,
+      "loss": -0.0582,
+      "reward": 2.4741777181625366,
+      "reward_std": 0.46304862946271896,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.025822297669947147,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1467
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 288.2291717529297,
+      "epoch": 0.734,
+      "grad_norm": 8.310583448002093,
+      "kl": 1.1015625,
+      "learning_rate": 2.8043938066798645e-07,
+      "loss": 0.0088,
+      "reward": 2.6622068881988525,
+      "reward_std": 0.3658689558506012,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.021820951253175735,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1468
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.3333435058594,
+      "epoch": 0.7345,
+      "grad_norm": 10.495545513429683,
+      "kl": 0.80078125,
+      "learning_rate": 2.7981089860335225e-07,
+      "loss": 0.0209,
+      "reward": 2.71765673160553,
+      "reward_std": 0.37030835449695587,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.030607173219323158,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1469
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.3333435058594,
+      "epoch": 0.735,
+      "grad_norm": 3.6146465901165348,
+      "kl": 0.54296875,
+      "learning_rate": 2.791832395815782e-07,
+      "loss": 0.0526,
+      "reward": 2.7022202014923096,
+      "reward_std": 0.08511605486273766,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.026946650817990303,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1470
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.8541717529297,
+      "epoch": 0.7355,
+      "grad_norm": 12.76797627033702,
+      "kl": 0.978515625,
+      "learning_rate": 2.7855640551462287e-07,
+      "loss": 0.0287,
+      "reward": 2.633518099784851,
+      "reward_std": 0.5054084360599518,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.02446805965155363,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1471
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 286.06251525878906,
+      "epoch": 0.736,
+      "grad_norm": 11.77020718227818,
+      "kl": 1.6171875,
+      "learning_rate": 2.7793039831193133e-07,
+      "loss": -0.0021,
+      "reward": 2.617877721786499,
+      "reward_std": 0.5727947354316711,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.04531669802963734,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1472
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.1458435058594,
+      "epoch": 0.7365,
+      "grad_norm": 6.357148681677889,
+      "kl": 0.94140625,
+      "learning_rate": 2.773052198804301e-07,
+      "loss": -0.0548,
+      "reward": 2.5771846771240234,
+      "reward_std": 0.6074622422456741,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.01830137614160776,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1473
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.4166717529297,
+      "epoch": 0.737,
+      "grad_norm": 6.670239361314333,
+      "kl": 1.5390625,
+      "learning_rate": 2.766808721245211e-07,
+      "loss": 0.0276,
+      "reward": 2.690428137779236,
+      "reward_std": 0.5497512817382812,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.023113532923161983,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1474
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 269.18751525878906,
+      "epoch": 0.7375,
+      "grad_norm": 6.355162766928286,
+      "kl": 2.09765625,
+      "learning_rate": 2.760573569460757e-07,
+      "loss": 0.0137,
+      "reward": 2.8231576681137085,
+      "reward_std": 0.3577008843421936,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.041425829753279686,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1475
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.7708435058594,
+      "epoch": 0.738,
+      "grad_norm": 4.4722454967487995,
+      "kl": 1.279296875,
+      "learning_rate": 2.7543467624442956e-07,
+      "loss": 0.0344,
+      "reward": 2.824985980987549,
+      "reward_std": 0.3037413991987705,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.017028134781867266,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 1476
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 252.89583587646484,
+      "epoch": 0.7385,
+      "grad_norm": 16.986839644883492,
+      "kl": 4.03125,
+      "learning_rate": 2.7481283191637605e-07,
+      "loss": 0.0652,
+      "reward": 2.479672074317932,
+      "reward_std": 0.8198486566543579,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.009911454282701015,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1477
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 261.2708435058594,
+      "epoch": 0.739,
+      "grad_norm": 9.368512168345658,
+      "kl": 2.2265625,
+      "learning_rate": 2.741918258561607e-07,
+      "loss": -0.0376,
+      "reward": 2.04381787776947,
+      "reward_std": 0.49914440512657166,
+      "rewards/accuracy_reward": 0.1875000074505806,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.023890579119324684,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1478
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 303.625,
+      "epoch": 0.7395,
+      "grad_norm": 13.277381628251995,
+      "kl": 2.6484375,
+      "learning_rate": 2.7357165995547547e-07,
+      "loss": 0.1547,
+      "reward": 2.4987998008728027,
+      "reward_std": 0.6838244497776031,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.0272420234978199,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1479
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 263.8333435058594,
+      "epoch": 0.74,
+      "grad_norm": 7.347881942486295,
+      "kl": 1.5859375,
+      "learning_rate": 2.729523361034538e-07,
+      "loss": -0.0831,
+      "reward": 2.309889793395996,
+      "reward_std": 0.5420120805501938,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.025179607793688774,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1480
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.7291717529297,
+      "epoch": 0.7405,
+      "grad_norm": 14.604720055826698,
+      "kl": 1.775390625,
+      "learning_rate": 2.7233385618666315e-07,
+      "loss": 0.1195,
+      "reward": 2.609587788581848,
+      "reward_std": 0.3773094117641449,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03103726916015148,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1481
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.4375,
+      "epoch": 0.741,
+      "grad_norm": 34.38249331144065,
+      "kl": 4.140625,
+      "learning_rate": 2.717162220891007e-07,
+      "loss": 0.0792,
+      "reward": 1.9346249103546143,
+      "reward_std": 0.3914206847548485,
+      "rewards/accuracy_reward": 0.12500000558793545,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.01676415279507637,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1482
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 274.0416717529297,
+      "epoch": 0.7415,
+      "grad_norm": 4.498402010991118,
+      "kl": 1.48046875,
+      "learning_rate": 2.7109943569218707e-07,
+      "loss": -0.0545,
+      "reward": 2.460047483444214,
+      "reward_std": 0.543542668223381,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.026063593104481697,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1483
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.5416717529297,
+      "epoch": 0.742,
+      "grad_norm": 9.59444129515133,
+      "kl": 1.5859375,
+      "learning_rate": 2.7048349887476037e-07,
+      "loss": -0.0225,
+      "reward": 2.367347240447998,
+      "reward_std": 0.5063965022563934,
+      "rewards/accuracy_reward": 0.47916667722165585,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.014597164001315832,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1484
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 297.1666717529297,
+      "epoch": 0.7425,
+      "grad_norm": 6.833454160186346,
+      "kl": 1.1015625,
+      "learning_rate": 2.698684135130713e-07,
+      "loss": -0.0404,
+      "reward": 2.429524064064026,
+      "reward_std": 0.6519744396209717,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.04096201993525028,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1485
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.3958435058594,
+      "epoch": 0.743,
+      "grad_norm": 5.515959993557466,
+      "kl": 0.701171875,
+      "learning_rate": 2.692541814807763e-07,
+      "loss": 0.0611,
+      "reward": 2.8888471126556396,
+      "reward_std": 0.18290941882878542,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.0174028305336833,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1486
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 269.5208435058594,
+      "epoch": 0.7435,
+      "grad_norm": 6.265052966360687,
+      "kl": 1.017578125,
+      "learning_rate": 2.686408046489328e-07,
+      "loss": -0.0232,
+      "reward": 2.5907429456710815,
+      "reward_std": 0.3499850779771805,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.020368190482258797,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1487
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 324.0208435058594,
+      "epoch": 0.744,
+      "grad_norm": 6.5782759908928945,
+      "kl": 0.80859375,
+      "learning_rate": 2.6802828488599294e-07,
+      "loss": 0.0469,
+      "reward": 2.6091192960739136,
+      "reward_std": 0.4149770438671112,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02282518707215786,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1488
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.68751525878906,
+      "epoch": 0.7445,
+      "grad_norm": 6.142749229443921,
+      "kl": 0.759765625,
+      "learning_rate": 2.6741662405779796e-07,
+      "loss": 0.0677,
+      "reward": 2.660447597503662,
+      "reward_std": 0.3648398518562317,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.027052627876400948,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1489
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 303.0833435058594,
+      "epoch": 0.745,
+      "grad_norm": 11.211102602501663,
+      "kl": 1.7421875,
+      "learning_rate": 2.6680582402757324e-07,
+      "loss": -0.006,
+      "reward": 2.27160906791687,
+      "reward_std": 0.5131259560585022,
+      "rewards/accuracy_reward": 0.4375000298023224,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.0322106322273612,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1490
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 310.7916717529297,
+      "epoch": 0.7455,
+      "grad_norm": 7.269196197513535,
+      "kl": 1.5234375,
+      "learning_rate": 2.661958866559213e-07,
+      "loss": 0.0242,
+      "reward": 2.570857882499695,
+      "reward_std": 0.2515888065099716,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.035044897347688675,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1491
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 267.81250762939453,
+      "epoch": 0.746,
+      "grad_norm": 16.92104888480676,
+      "kl": 4.099609375,
+      "learning_rate": 2.655868138008171e-07,
+      "loss": -0.074,
+      "reward": 2.5825823545455933,
+      "reward_std": 0.6137717366218567,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.8958334028720856,
+      "rewards/repetition_penalty_reward": -0.032001130282878876,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1492
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 275.7916717529297,
+      "epoch": 0.7465,
+      "grad_norm": 6.220590527633046,
+      "kl": 1.08984375,
+      "learning_rate": 2.649786073176025e-07,
+      "loss": -0.0146,
+      "reward": 2.8706815242767334,
+      "reward_std": 0.18304454255849123,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.019943561404943466,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 1493
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.0416717529297,
+      "epoch": 0.747,
+      "grad_norm": 8.80186866611191,
+      "kl": 1.95703125,
+      "learning_rate": 2.6437126905897967e-07,
+      "loss": 0.0754,
+      "reward": 2.5168925523757935,
+      "reward_std": 0.46842535585165024,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.029982510022819042,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1494
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 250.89584350585938,
+      "epoch": 0.7475,
+      "grad_norm": 11.509437222020193,
+      "kl": 3.453125,
+      "learning_rate": 2.637648008750062e-07,
+      "loss": -0.0703,
+      "reward": 2.5987823009490967,
+      "reward_std": 0.7667766213417053,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.0227454686537385,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1495
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.5208435058594,
+      "epoch": 0.748,
+      "grad_norm": 11.744253799204682,
+      "kl": 2.890625,
+      "learning_rate": 2.631592046130896e-07,
+      "loss": 0.0104,
+      "reward": 2.1920148134231567,
+      "reward_std": 0.5672837495803833,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.02847137115895748,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1496
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.6458435058594,
+      "epoch": 0.7485,
+      "grad_norm": 14.637527744293745,
+      "kl": 3.61328125,
+      "learning_rate": 2.6255448211798103e-07,
+      "loss": -0.0551,
+      "reward": 2.6001073122024536,
+      "reward_std": 0.43944015353918076,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03878166899085045,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1497
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 272.4583435058594,
+      "epoch": 0.749,
+      "grad_norm": 4.858094415340834,
+      "kl": 0.740234375,
+      "learning_rate": 2.6195063523177e-07,
+      "loss": 0.0534,
+      "reward": 2.713899612426758,
+      "reward_std": 0.0610094303265214,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.020475570112466812,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1498
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 269.0833435058594,
+      "epoch": 0.7495,
+      "grad_norm": 5.449268107276577,
+      "kl": 1.134765625,
+      "learning_rate": 2.613476657938789e-07,
+      "loss": 0.0716,
+      "reward": 2.6610913276672363,
+      "reward_std": 0.27378853410482407,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.021200459450483322,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1499
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 272.5416717529297,
+      "epoch": 0.75,
+      "grad_norm": 5.9488407024205685,
+      "kl": 1.52734375,
+      "learning_rate": 2.6074557564105724e-07,
+      "loss": 0.0979,
+      "reward": 2.62813401222229,
+      "reward_std": 0.37979737704154104,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.015963357756845653,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1500
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 267.95833587646484,
+      "epoch": 0.7505,
+      "grad_norm": 4.3632751002774315,
+      "kl": 1.6484375,
+      "learning_rate": 2.6014436660737605e-07,
+      "loss": -0.0364,
+      "reward": 2.6962149143218994,
+      "reward_std": 0.5003593862056732,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.02253533434122801,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1501
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.2708435058594,
+      "epoch": 0.751,
+      "grad_norm": 8.236462804572298,
+      "kl": 1.58984375,
+      "learning_rate": 2.595440405242222e-07,
+      "loss": -0.0203,
+      "reward": 2.7835288047790527,
+      "reward_std": 0.4721106141805649,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.013346214778721333,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1502
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 251.625,
+      "epoch": 0.7515,
+      "grad_norm": 6.981131847692658,
+      "kl": 1.91015625,
+      "learning_rate": 2.589445992202931e-07,
+      "loss": -0.0142,
+      "reward": 2.6148020029067993,
+      "reward_std": 0.5384411364793777,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.020614707842469215,
+      "rewards/tag_count_reward": 0.9062500298023224,
+      "step": 1503
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.31251525878906,
+      "epoch": 0.752,
+      "grad_norm": 6.52821102623756,
+      "kl": 1.8125,
+      "learning_rate": 2.583460445215911e-07,
+      "loss": -0.0262,
+      "reward": 2.4926270246505737,
+      "reward_std": 0.7499818503856659,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.0334148071706295,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1504
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 295.93751525878906,
+      "epoch": 0.7525,
+      "grad_norm": 33.49372224861189,
+      "kl": 1.78125,
+      "learning_rate": 2.5774837825141736e-07,
+      "loss": 0.1068,
+      "reward": 2.8040874004364014,
+      "reward_std": 0.3838850110769272,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.034454355016350746,
+      "rewards/tag_count_reward": 0.9427083730697632,
+      "step": 1505
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 306.62501525878906,
+      "epoch": 0.753,
+      "grad_norm": 4.115301737132047,
+      "kl": 0.611328125,
+      "learning_rate": 2.571516022303671e-07,
+      "loss": 0.0313,
+      "reward": 2.837872266769409,
+      "reward_std": 0.33879856765270233,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02150270715355873,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1506
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 322.7708435058594,
+      "epoch": 0.7535,
+      "grad_norm": 6.925696299115607,
+      "kl": 1.39453125,
+      "learning_rate": 2.565557182763235e-07,
+      "loss": 0.0259,
+      "reward": 2.422218918800354,
+      "reward_std": 0.5452222675085068,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.027433859184384346,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1507
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 298.5,
+      "epoch": 0.754,
+      "grad_norm": 5.78599950453997,
+      "kl": 1.595703125,
+      "learning_rate": 2.5596072820445254e-07,
+      "loss": 0.0677,
+      "reward": 2.514102339744568,
+      "reward_std": 0.51523557305336,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.01714779995381832,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1508
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.6875,
+      "epoch": 0.7545,
+      "grad_norm": 23.247679880499817,
+      "kl": 1.98828125,
+      "learning_rate": 2.5536663382719713e-07,
+      "loss": 0.1553,
+      "reward": 2.4761857986450195,
+      "reward_std": 0.5169162601232529,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.034231009893119335,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1509
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.41668701171875,
+      "epoch": 0.755,
+      "grad_norm": 4.780358683601452,
+      "kl": 1.453125,
+      "learning_rate": 2.547734369542718e-07,
+      "loss": -0.006,
+      "reward": 2.5231523513793945,
+      "reward_std": 0.5849803388118744,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.023722639307379723,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1510
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.54168701171875,
+      "epoch": 0.7555,
+      "grad_norm": 54.12276638774576,
+      "kl": 2.9609375,
+      "learning_rate": 2.5418113939265686e-07,
+      "loss": 0.0877,
+      "reward": 2.6192400455474854,
+      "reward_std": 0.6453405022621155,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.023121179081499577,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1511
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 263.85418701171875,
+      "epoch": 0.756,
+      "grad_norm": 19.449042044921896,
+      "kl": 2.671875,
+      "learning_rate": 2.5358974294659373e-07,
+      "loss": -0.0146,
+      "reward": 2.5455033779144287,
+      "reward_std": 0.7476102709770203,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.020468920469284058,
+      "rewards/tag_count_reward": 0.9270833730697632,
+      "step": 1512
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.97918701171875,
+      "epoch": 0.7565,
+      "grad_norm": 10.92392466120309,
+      "kl": 1.876953125,
+      "learning_rate": 2.5299924941757843e-07,
+      "loss": 0.1007,
+      "reward": 2.4055851697921753,
+      "reward_std": 0.5663338005542755,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.04233153909444809,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1513
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.5416793823242,
+      "epoch": 0.757,
+      "grad_norm": 44.00142117460777,
+      "kl": 4.20703125,
+      "learning_rate": 2.5240966060435674e-07,
+      "loss": -0.0122,
+      "reward": 2.163232743740082,
+      "reward_std": 0.4055490791797638,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.7986111044883728,
+      "rewards/repetition_penalty_reward": -0.020795070566236973,
+      "rewards/tag_count_reward": 0.8229166865348816,
+      "step": 1514
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 275.5625,
+      "epoch": 0.7575,
+      "grad_norm": 6.548140605154211,
+      "kl": 2.34375,
+      "learning_rate": 2.5182097830291824e-07,
+      "loss": -0.1129,
+      "reward": 2.405154585838318,
+      "reward_std": 0.7896367907524109,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.020192706026136875,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1515
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.4791717529297,
+      "epoch": 0.758,
+      "grad_norm": 16.21577117974573,
+      "kl": 1.146484375,
+      "learning_rate": 2.512332043064913e-07,
+      "loss": 0.1167,
+      "reward": 2.509866714477539,
+      "reward_std": 0.5259620547294617,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.03527211956679821,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1516
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.1875,
+      "epoch": 0.7585,
+      "grad_norm": 6.825195538822801,
+      "kl": 1.1484375,
+      "learning_rate": 2.5064634040553767e-07,
+      "loss": 0.0764,
+      "reward": 2.596480369567871,
+      "reward_std": 0.37036192417144775,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.023311403580009937,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1517
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.39583587646484,
+      "epoch": 0.759,
+      "grad_norm": 7.531211105112099,
+      "kl": 2.37890625,
+      "learning_rate": 2.5006038838774647e-07,
+      "loss": 0.0148,
+      "reward": 2.2257198095321655,
+      "reward_std": 0.731947273015976,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.022544228471815586,
+      "rewards/tag_count_reward": 0.8593750298023224,
+      "step": 1518
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.9791717529297,
+      "epoch": 0.7595,
+      "grad_norm": 12.630414631027104,
+      "kl": 1.7578125,
+      "learning_rate": 2.494753500380291e-07,
+      "loss": 0.0332,
+      "reward": 2.614049792289734,
+      "reward_std": 0.6869710385799408,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.05608919635415077,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1519
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 288.54168701171875,
+      "epoch": 0.76,
+      "grad_norm": 5.390351956211983,
+      "kl": 1.11328125,
+      "learning_rate": 2.488912271385139e-07,
+      "loss": 0.0027,
+      "reward": 2.871764063835144,
+      "reward_std": 0.32710327208042145,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.02059698849916458,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1520
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.1041717529297,
+      "epoch": 0.7605,
+      "grad_norm": 15.294086423040786,
+      "kl": 2.3125,
+      "learning_rate": 2.483080214685404e-07,
+      "loss": 0.0703,
+      "reward": 2.6500085592269897,
+      "reward_std": 0.6067156195640564,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.021866421215236187,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1521
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.5833435058594,
+      "epoch": 0.761,
+      "grad_norm": 7.991198731135982,
+      "kl": 1.4140625,
+      "learning_rate": 2.4772573480465445e-07,
+      "loss": 0.0322,
+      "reward": 2.7214081287384033,
+      "reward_std": 0.42735913395881653,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.021647341549396515,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1522
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.81251525878906,
+      "epoch": 0.7615,
+      "grad_norm": 7.133883392763306,
+      "kl": 1.2734375,
+      "learning_rate": 2.471443689206021e-07,
+      "loss": 0.0274,
+      "reward": 2.5906002521514893,
+      "reward_std": 0.45796915888786316,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.030927601736038923,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1523
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.87501525878906,
+      "epoch": 0.762,
+      "grad_norm": 9.686910468630904,
+      "kl": 2.16015625,
+      "learning_rate": 2.465639255873246e-07,
+      "loss": 0.3267,
+      "reward": 2.679943323135376,
+      "reward_std": 0.5434834957122803,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.030126115307211876,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1524
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.2708435058594,
+      "epoch": 0.7625,
+      "grad_norm": 5.1490011118154655,
+      "kl": 2.33203125,
+      "learning_rate": 2.4598440657295286e-07,
+      "loss": 0.0067,
+      "reward": 2.389386534690857,
+      "reward_std": 0.5596683621406555,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.01686352863907814,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1525
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 252.52083587646484,
+      "epoch": 0.763,
+      "grad_norm": 34.19051071296814,
+      "kl": 3.17578125,
+      "learning_rate": 2.454058136428027e-07,
+      "loss": 0.2352,
+      "reward": 2.393932580947876,
+      "reward_std": 0.526284646242857,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.017525773961097002,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1526
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.8125,
+      "epoch": 0.7635,
+      "grad_norm": 17.54640194645924,
+      "kl": 1.44140625,
+      "learning_rate": 2.4482814855936834e-07,
+      "loss": 0.0779,
+      "reward": 2.7223398685455322,
+      "reward_std": 0.46589484065771103,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.018979622051119804,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1527
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 263.0208435058594,
+      "epoch": 0.764,
+      "grad_norm": 9.08850475587058,
+      "kl": 2.18359375,
+      "learning_rate": 2.4425141308231765e-07,
+      "loss": 0.0748,
+      "reward": 2.8009958267211914,
+      "reward_std": 0.44475598144344985,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.009768242482095957,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1528
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 352.81251525878906,
+      "epoch": 0.7645,
+      "grad_norm": 19.330244578947458,
+      "kl": 2.6796875,
+      "learning_rate": 2.43675608968487e-07,
+      "loss": 0.1724,
+      "reward": 1.9503087401390076,
+      "reward_std": 0.43879881501197815,
+      "rewards/accuracy_reward": 0.1666666679084301,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.03580245561897755,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1529
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.8125,
+      "epoch": 0.765,
+      "grad_norm": 10.153421840080243,
+      "kl": 1.87890625,
+      "learning_rate": 2.4310073797187573e-07,
+      "loss": 0.0312,
+      "reward": 2.5163590908050537,
+      "reward_std": 0.5512074381113052,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.0322520462796092,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1530
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 239.02083587646484,
+      "epoch": 0.7655,
+      "grad_norm": 6.7813462328053635,
+      "kl": 1.6484375,
+      "learning_rate": 2.4252680184364045e-07,
+      "loss": -0.043,
+      "reward": 2.6130350828170776,
+      "reward_std": 0.6324526071548462,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.017173412023112178,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1531
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 266.4166717529297,
+      "epoch": 0.766,
+      "grad_norm": 8.265844118560226,
+      "kl": 1.4765625,
+      "learning_rate": 2.4195380233209006e-07,
+      "loss": -0.0266,
+      "reward": 2.7328121662139893,
+      "reward_std": 0.6724408268928528,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.02413230948150158,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1532
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 306.9375,
+      "epoch": 0.7665,
+      "grad_norm": 3.9331602535210513,
+      "kl": 1.0859375,
+      "learning_rate": 2.413817411826807e-07,
+      "loss": -0.1036,
+      "reward": 2.5074063539505005,
+      "reward_std": 0.5442797392606735,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.029052074998617172,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1533
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 240.77084350585938,
+      "epoch": 0.767,
+      "grad_norm": 9.973948039444034,
+      "kl": 2.41796875,
+      "learning_rate": 2.408106201380097e-07,
+      "loss": 0.0,
+      "reward": 2.4841666221618652,
+      "reward_std": 0.5229880660772324,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.021041802130639553,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1534
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 316.625,
+      "epoch": 0.7675,
+      "grad_norm": 11.813541271263112,
+      "kl": 2.3046875,
+      "learning_rate": 2.4024044093781063e-07,
+      "loss": 0.1092,
+      "reward": 2.418421983718872,
+      "reward_std": 0.761476993560791,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.031230845488607883,
+      "rewards/tag_count_reward": 0.8593750298023224,
+      "step": 1535
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.9583435058594,
+      "epoch": 0.768,
+      "grad_norm": 13.907263883986914,
+      "kl": 1.25390625,
+      "learning_rate": 2.3967120531894857e-07,
+      "loss": 0.0569,
+      "reward": 2.579371452331543,
+      "reward_std": 0.3498530462384224,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02132294327020645,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1536
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.4166717529297,
+      "epoch": 0.7685,
+      "grad_norm": 12.302399407816415,
+      "kl": 3.2890625,
+      "learning_rate": 2.391029150154137e-07,
+      "loss": 0.0624,
+      "reward": 2.3881059885025024,
+      "reward_std": 0.6679995059967041,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.014672036748379469,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1537
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 301.5833435058594,
+      "epoch": 0.769,
+      "grad_norm": 12.935252444247103,
+      "kl": 2.51953125,
+      "learning_rate": 2.38535571758317e-07,
+      "loss": 0.0609,
+      "reward": 2.621963858604431,
+      "reward_std": 0.6098971962928772,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.027341697365045547,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1538
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.5833435058594,
+      "epoch": 0.7695,
+      "grad_norm": 6.135311130568607,
+      "kl": 1.037109375,
+      "learning_rate": 2.3796917727588412e-07,
+      "loss": 0.0857,
+      "reward": 2.657423496246338,
+      "reward_std": 0.2529868111014366,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03007663320749998,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1539
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.8333435058594,
+      "epoch": 0.77,
+      "grad_norm": 10.519026184776623,
+      "kl": 2.1875,
+      "learning_rate": 2.374037332934512e-07,
+      "loss": 0.037,
+      "reward": 2.3503577709198,
+      "reward_std": 0.5039149820804596,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.029850583523511887,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1540
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 233.3541717529297,
+      "epoch": 0.7705,
+      "grad_norm": 7.260626069337147,
+      "kl": 1.4296875,
+      "learning_rate": 2.3683924153345854e-07,
+      "loss": 0.0345,
+      "reward": 2.9000203609466553,
+      "reward_std": 0.28021588921546936,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.020118530839681625,
+      "rewards/tag_count_reward": 0.9895833730697632,
+      "step": 1541
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.72918701171875,
+      "epoch": 0.771,
+      "grad_norm": 4.314407100492501,
+      "kl": 1.2578125,
+      "learning_rate": 2.36275703715446e-07,
+      "loss": -0.0121,
+      "reward": 2.534322142601013,
+      "reward_std": 0.3907851278781891,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.022969634272158146,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1542
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 271.81251525878906,
+      "epoch": 0.7715,
+      "grad_norm": 5.670151769824431,
+      "kl": 1.0703125,
+      "learning_rate": 2.357131215560474e-07,
+      "loss": 0.0037,
+      "reward": 2.828339695930481,
+      "reward_std": 0.2872830927371979,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03277149237692356,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1543
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.5,
+      "epoch": 0.772,
+      "grad_norm": 7.6282657037651695,
+      "kl": 2.09375,
+      "learning_rate": 2.3515149676898552e-07,
+      "loss": 0.0448,
+      "reward": 2.5357162952423096,
+      "reward_std": 0.5322327762842178,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.037200456485152245,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1544
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 296.93751525878906,
+      "epoch": 0.7725,
+      "grad_norm": 4.352945225717926,
+      "kl": 0.837890625,
+      "learning_rate": 2.3459083106506712e-07,
+      "loss": 0.221,
+      "reward": 2.649806499481201,
+      "reward_std": 0.2627423256635666,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.027276871260255575,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1545
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 311.8541717529297,
+      "epoch": 0.773,
+      "grad_norm": 5.842490680004781,
+      "kl": 1.6875,
+      "learning_rate": 2.3403112615217693e-07,
+      "loss": 0.1197,
+      "reward": 2.5485793352127075,
+      "reward_std": 0.4063766598701477,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.024337463080883026,
+      "rewards/tag_count_reward": 0.9270833730697632,
+      "step": 1546
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.43751525878906,
+      "epoch": 0.7735,
+      "grad_norm": 6.003285903115502,
+      "kl": 1.828125,
+      "learning_rate": 2.334723837352733e-07,
+      "loss": 0.1055,
+      "reward": 2.6460882425308228,
+      "reward_std": 0.5506278276443481,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.013634048402309418,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1547
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.5833435058594,
+      "epoch": 0.774,
+      "grad_norm": 11.249985053869272,
+      "kl": 3.328125,
+      "learning_rate": 2.3291460551638237e-07,
+      "loss": 0.2142,
+      "reward": 2.2939001321792603,
+      "reward_std": 0.8971810340881348,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.875,
+      "rewards/repetition_penalty_reward": -0.018599930219352245,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1548
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.37501525878906,
+      "epoch": 0.7745,
+      "grad_norm": 11.075594302485063,
+      "kl": 1.55859375,
+      "learning_rate": 2.3235779319459355e-07,
+      "loss": 0.0602,
+      "reward": 2.5785093307495117,
+      "reward_std": 0.29725973308086395,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.037810200825333595,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1549
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 317.93751525878906,
+      "epoch": 0.775,
+      "grad_norm": 11.673776782665986,
+      "kl": 1.67578125,
+      "learning_rate": 2.3180194846605364e-07,
+      "loss": 0.1151,
+      "reward": 2.7235286235809326,
+      "reward_std": 0.585949033498764,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9375000894069672,
+      "rewards/repetition_penalty_reward": -0.026471680030226707,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1550
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.4583435058594,
+      "epoch": 0.7755,
+      "grad_norm": 6.3252296705969115,
+      "kl": 1.35546875,
+      "learning_rate": 2.312470730239621e-07,
+      "loss": -0.0488,
+      "reward": 2.4448201656341553,
+      "reward_std": 0.48708635568618774,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.016985515132546425,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1551
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 294.2083435058594,
+      "epoch": 0.776,
+      "grad_norm": 12.954209363640562,
+      "kl": 2.06640625,
+      "learning_rate": 2.306931685585657e-07,
+      "loss": 0.1214,
+      "reward": 2.228946566581726,
+      "reward_std": 0.35232171416282654,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.026261983439326286,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1552
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.8125,
+      "epoch": 0.7765,
+      "grad_norm": 3.9761424833250136,
+      "kl": 1.26953125,
+      "learning_rate": 2.3014023675715339e-07,
+      "loss": 0.0292,
+      "reward": 2.886070728302002,
+      "reward_std": 0.23379188776016235,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.018443217035382986,
+      "rewards/tag_count_reward": 0.9739583730697632,
+      "step": 1553
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.18751525878906,
+      "epoch": 0.777,
+      "grad_norm": 9.05056417179947,
+      "kl": 1.35546875,
+      "learning_rate": 2.2958827930405162e-07,
+      "loss": 0.1063,
+      "reward": 2.5847833156585693,
+      "reward_std": 0.4528532326221466,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.035008576698601246,
+      "rewards/tag_count_reward": 0.9114583730697632,
+      "step": 1554
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.9791793823242,
+      "epoch": 0.7775,
+      "grad_norm": 104.85246338097696,
+      "kl": 2.3125,
+      "learning_rate": 2.2903729788061834e-07,
+      "loss": 0.1352,
+      "reward": 2.5495121479034424,
+      "reward_std": 0.7037906050682068,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.01993226632475853,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1555
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 299.1458435058594,
+      "epoch": 0.778,
+      "grad_norm": 6.890435145460005,
+      "kl": 1.11328125,
+      "learning_rate": 2.2848729416523859e-07,
+      "loss": 0.0421,
+      "reward": 2.3158843517303467,
+      "reward_std": 0.47459036111831665,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.024393508210778236,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1556
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 302.3541717529297,
+      "epoch": 0.7785,
+      "grad_norm": 9.282545995378012,
+      "kl": 1.044921875,
+      "learning_rate": 2.2793826983331886e-07,
+      "loss": 0.1611,
+      "reward": 2.3017712831497192,
+      "reward_std": 0.3790929764509201,
+      "rewards/accuracy_reward": 0.3958333544433117,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03156219515949488,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1557
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 285.87501525878906,
+      "epoch": 0.779,
+      "grad_norm": 3.6904386519177583,
+      "kl": 0.564453125,
+      "learning_rate": 2.2739022655728277e-07,
+      "loss": 0.0184,
+      "reward": 2.7911198139190674,
+      "reward_std": 0.22485119104385376,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.04915817081928253,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1558
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.875,
+      "epoch": 0.7795,
+      "grad_norm": 9.948120856303264,
+      "kl": 1.69921875,
+      "learning_rate": 2.268431660065651e-07,
+      "loss": 0.1686,
+      "reward": 2.5424914360046387,
+      "reward_std": 0.6018918454647064,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.881944477558136,
+      "rewards/repetition_penalty_reward": -0.021744604222476482,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1559
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.29168701171875,
+      "epoch": 0.78,
+      "grad_norm": 5.310065668788173,
+      "kl": 2.4140625,
+      "learning_rate": 2.2629708984760706e-07,
+      "loss": 0.1289,
+      "reward": 2.24657142162323,
+      "reward_std": 0.9247779548168182,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.8263889253139496,
+      "rewards/repetition_penalty_reward": -0.02252592984586954,
+      "rewards/tag_count_reward": 0.859375,
+      "step": 1560
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.75001525878906,
+      "epoch": 0.7805,
+      "grad_norm": 8.564129818034443,
+      "kl": 1.359375,
+      "learning_rate": 2.2575199974385144e-07,
+      "loss": 0.1631,
+      "reward": 2.4449636936187744,
+      "reward_std": 0.49174782633781433,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.022050250321626663,
+      "rewards/tag_count_reward": 0.9739583730697632,
+      "step": 1561
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 230.7291717529297,
+      "epoch": 0.781,
+      "grad_norm": 5.561119420119422,
+      "kl": 2.189453125,
+      "learning_rate": 2.2520789735573704e-07,
+      "loss": 0.0542,
+      "reward": 2.534723997116089,
+      "reward_std": 0.5092030912637711,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.012151051312685013,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1562
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.93751525878906,
+      "epoch": 0.7815,
+      "grad_norm": 5.117268929695284,
+      "kl": 0.931640625,
+      "learning_rate": 2.2466478434069435e-07,
+      "loss": 0.0107,
+      "reward": 2.3754416704177856,
+      "reward_std": 0.39969532936811447,
+      "rewards/accuracy_reward": 0.4375000111758709,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.025599990040063858,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1563
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 350.6041717529297,
+      "epoch": 0.782,
+      "grad_norm": 9.440710428522506,
+      "kl": 1.81640625,
+      "learning_rate": 2.2412266235313973e-07,
+      "loss": 0.3979,
+      "reward": 2.50728178024292,
+      "reward_std": 0.6064448654651642,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03959329519420862,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1564
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 380.7083435058594,
+      "epoch": 0.7825,
+      "grad_norm": 24.815520443713027,
+      "kl": 2.8515625,
+      "learning_rate": 2.2358153304447066e-07,
+      "loss": 0.4005,
+      "reward": 2.2530555725097656,
+      "reward_std": 0.5099890530109406,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.8958334028720856,
+      "rewards/repetition_penalty_reward": -0.02298619970679283,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1565
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.00001525878906,
+      "epoch": 0.783,
+      "grad_norm": 12.635922668896523,
+      "kl": 2.2734375,
+      "learning_rate": 2.230413980630609e-07,
+      "loss": 0.0445,
+      "reward": 2.488055467605591,
+      "reward_std": 0.5981208235025406,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.032778057269752026,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1566
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 370.56251525878906,
+      "epoch": 0.7835,
+      "grad_norm": 14.898019687710494,
+      "kl": 2.625,
+      "learning_rate": 2.2250225905425532e-07,
+      "loss": 0.2279,
+      "reward": 2.41832172870636,
+      "reward_std": 0.6433684527873993,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.8680556118488312,
+      "rewards/repetition_penalty_reward": -0.0226507056504488,
+      "rewards/tag_count_reward": 0.8437500298023224,
+      "step": 1567
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 332.1458435058594,
+      "epoch": 0.784,
+      "grad_norm": 8.069928988207886,
+      "kl": 2.375,
+      "learning_rate": 2.2196411766036487e-07,
+      "loss": 0.2497,
+      "reward": 2.5703223943710327,
+      "reward_std": 0.686181902885437,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.026900025084614754,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1568
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.8958435058594,
+      "epoch": 0.7845,
+      "grad_norm": 10.489749526563086,
+      "kl": 2.8671875,
+      "learning_rate": 2.2142697552066142e-07,
+      "loss": 0.2871,
+      "reward": 2.377824902534485,
+      "reward_std": 0.7266697287559509,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.019744650460779667,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1569
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.3958435058594,
+      "epoch": 0.785,
+      "grad_norm": 4.268659681870845,
+      "kl": 1.625,
+      "learning_rate": 2.2089083427137329e-07,
+      "loss": -0.0771,
+      "reward": 2.36256742477417,
+      "reward_std": 0.4559687077999115,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.031529780477285385,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1570
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.5416717529297,
+      "epoch": 0.7855,
+      "grad_norm": 8.822752169087817,
+      "kl": 1.3984375,
+      "learning_rate": 2.203556955456796e-07,
+      "loss": -0.0131,
+      "reward": 2.594992756843567,
+      "reward_std": 0.6085972636938095,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.030007286928594112,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1571
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.95833587646484,
+      "epoch": 0.786,
+      "grad_norm": 31.096559251399594,
+      "kl": 1.474609375,
+      "learning_rate": 2.1982156097370557e-07,
+      "loss": 0.2107,
+      "reward": 2.1280053853988647,
+      "reward_std": 0.2961831293068826,
+      "rewards/accuracy_reward": 0.2291666716337204,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.012619667453691363,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1572
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 303.5208435058594,
+      "epoch": 0.7865,
+      "grad_norm": 7.590109325873461,
+      "kl": 2.75390625,
+      "learning_rate": 2.1928843218251803e-07,
+      "loss": 0.0717,
+      "reward": 2.455071449279785,
+      "reward_std": 0.6115701645612717,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.826388955116272,
+      "rewards/repetition_penalty_reward": -0.02756764553487301,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1573
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.1458435058594,
+      "epoch": 0.787,
+      "grad_norm": 10.349493504338458,
+      "kl": 1.09765625,
+      "learning_rate": 2.1875631079611956e-07,
+      "loss": 0.1317,
+      "reward": 2.6623623371124268,
+      "reward_std": 0.5247603058815002,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.026873953640460968,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1574
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 304.43751525878906,
+      "epoch": 0.7875,
+      "grad_norm": 10.846195351334975,
+      "kl": 1.388671875,
+      "learning_rate": 2.1822519843544422e-07,
+      "loss": 0.0464,
+      "reward": 2.253710389137268,
+      "reward_std": 0.6600132286548615,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9097222983837128,
+      "rewards/repetition_penalty_reward": -0.025803642347455025,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1575
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 316.5416717529297,
+      "epoch": 0.788,
+      "grad_norm": 11.7016761594706,
+      "kl": 1.296875,
+      "learning_rate": 2.1769509671835223e-07,
+      "loss": 0.0149,
+      "reward": 2.3044776916503906,
+      "reward_std": 0.6875050067901611,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.881944477558136,
+      "rewards/repetition_penalty_reward": -0.041008614003658295,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1576
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.0,
+      "epoch": 0.7885,
+      "grad_norm": 23.225955314196877,
+      "kl": 1.875,
+      "learning_rate": 2.1716600725962558e-07,
+      "loss": 0.3445,
+      "reward": 2.3711761236190796,
+      "reward_std": 0.7729770541191101,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.02118516620248556,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1577
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 271.04168701171875,
+      "epoch": 0.789,
+      "grad_norm": 9.258843562543557,
+      "kl": 1.7578125,
+      "learning_rate": 2.166379316709625e-07,
+      "loss": -0.0326,
+      "reward": 2.26195228099823,
+      "reward_std": 0.7921708822250366,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.8888888955116272,
+      "rewards/repetition_penalty_reward": -0.02797840256243944,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1578
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 301.75,
+      "epoch": 0.7895,
+      "grad_norm": 10.626888168484715,
+      "kl": 1.875,
+      "learning_rate": 2.1611087156097267e-07,
+      "loss": -0.0186,
+      "reward": 2.7739405632019043,
+      "reward_std": 0.3001384465023875,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.029878957197070122,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1579
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 297.72918701171875,
+      "epoch": 0.79,
+      "grad_norm": 7.09473650785355,
+      "kl": 1.38671875,
+      "learning_rate": 2.1558482853517253e-07,
+      "loss": 0.1454,
+      "reward": 2.6300599575042725,
+      "reward_std": 0.4111658036708832,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.024454043712466955,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1580
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.8333435058594,
+      "epoch": 0.7905,
+      "grad_norm": 5.85391538917019,
+      "kl": 1.34375,
+      "learning_rate": 2.1505980419598063e-07,
+      "loss": 0.0935,
+      "reward": 2.8416703939437866,
+      "reward_std": 0.30897435545921326,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.029857425950467587,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1581
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.6666717529297,
+      "epoch": 0.791,
+      "grad_norm": 5.122296677365158,
+      "kl": 1.146484375,
+      "learning_rate": 2.1453580014271203e-07,
+      "loss": 0.0385,
+      "reward": 2.530571699142456,
+      "reward_std": 0.24665961414575577,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.02324793115258217,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1582
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 271.31251525878906,
+      "epoch": 0.7915,
+      "grad_norm": 19.824116779052915,
+      "kl": 3.8125,
+      "learning_rate": 2.1401281797157395e-07,
+      "loss": 0.1603,
+      "reward": 2.473771572113037,
+      "reward_std": 0.672837495803833,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.8333334028720856,
+      "rewards/repetition_penalty_reward": -0.015811904333531857,
+      "rewards/tag_count_reward": 0.9062500298023224,
+      "step": 1583
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 274.91668701171875,
+      "epoch": 0.792,
+      "grad_norm": 6.585486937770688,
+      "kl": 1.6484375,
+      "learning_rate": 2.134908592756607e-07,
+      "loss": 0.1439,
+      "reward": 2.618052840232849,
+      "reward_std": 0.6666180491447449,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.015627761371433735,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1584
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 340.12501525878906,
+      "epoch": 0.7925,
+      "grad_norm": 18.241947160047147,
+      "kl": 2.865234375,
+      "learning_rate": 2.1296992564494904e-07,
+      "loss": 0.1482,
+      "reward": 2.4161217212677,
+      "reward_std": 0.7258022725582123,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.881944477558136,
+      "rewards/repetition_penalty_reward": -0.028322923928499222,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1585
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.2291793823242,
+      "epoch": 0.793,
+      "grad_norm": 4.017024733526017,
+      "kl": 0.64453125,
+      "learning_rate": 2.124500186662932e-07,
+      "loss": 0.0263,
+      "reward": 2.8568403720855713,
+      "reward_std": 0.22998722037300467,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.01815977320075035,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1586
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 261.6458435058594,
+      "epoch": 0.7935,
+      "grad_norm": 5.17341645646987,
+      "kl": 1.130859375,
+      "learning_rate": 2.1193113992342001e-07,
+      "loss": 0.0359,
+      "reward": 2.874253273010254,
+      "reward_std": 0.3271254301071167,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02505233883857727,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1587
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 326.6041717529297,
+      "epoch": 0.794,
+      "grad_norm": 15.607330298676484,
+      "kl": 2.05078125,
+      "learning_rate": 2.1141329099692406e-07,
+      "loss": 0.2062,
+      "reward": 2.5750577449798584,
+      "reward_std": 0.32292182743549347,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.042997824028134346,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1588
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 339.93751525878906,
+      "epoch": 0.7945,
+      "grad_norm": 10.81146440795989,
+      "kl": 1.345703125,
+      "learning_rate": 2.1089647346426303e-07,
+      "loss": 0.3202,
+      "reward": 2.6132642030715942,
+      "reward_std": 0.2669622115790844,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.022152533754706383,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1589
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 377.9583435058594,
+      "epoch": 0.795,
+      "grad_norm": 5.637731152589917,
+      "kl": 1.8984375,
+      "learning_rate": 2.1038068889975259e-07,
+      "loss": 0.3067,
+      "reward": 2.5194292068481445,
+      "reward_std": 0.6721850037574768,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9027778506278992,
+      "rewards/repetition_penalty_reward": -0.029181976802647114,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1590
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.0833435058594,
+      "epoch": 0.7955,
+      "grad_norm": 12.21283214309097,
+      "kl": 0.978515625,
+      "learning_rate": 2.0986593887456223e-07,
+      "loss": 0.233,
+      "reward": 2.7278926372528076,
+      "reward_std": 0.5767909586429596,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.023843545466661453,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1591
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.22918701171875,
+      "epoch": 0.796,
+      "grad_norm": 5.466559112755202,
+      "kl": 1.630859375,
+      "learning_rate": 2.0935222495670968e-07,
+      "loss": 0.1362,
+      "reward": 2.65623676776886,
+      "reward_std": 0.6166912019252777,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9444444179534912,
+      "rewards/repetition_penalty_reward": -0.02779117412865162,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1592
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 297.2916717529297,
+      "epoch": 0.7965,
+      "grad_norm": 6.094615509828546,
+      "kl": 1.22265625,
+      "learning_rate": 2.088395487110566e-07,
+      "loss": 0.078,
+      "reward": 2.733526110649109,
+      "reward_std": 0.44847723841667175,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.02515467908233404,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1593
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 282.56251525878906,
+      "epoch": 0.797,
+      "grad_norm": 4.218497044118198,
+      "kl": 1.359375,
+      "learning_rate": 2.0832791169930363e-07,
+      "loss": 0.2248,
+      "reward": 2.6115695238113403,
+      "reward_std": 0.34022286534309387,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.018638767302036285,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1594
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 353.7916717529297,
+      "epoch": 0.7975,
+      "grad_norm": 4.25463358775561,
+      "kl": 0.748046875,
+      "learning_rate": 2.078173154799861e-07,
+      "loss": 0.1418,
+      "reward": 2.7314101457595825,
+      "reward_std": 0.33814239501953125,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.03768732026219368,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1595
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 311.1666717529297,
+      "epoch": 0.798,
+      "grad_norm": 7.987189771627374,
+      "kl": 1.658203125,
+      "learning_rate": 2.0730776160846853e-07,
+      "loss": 0.1429,
+      "reward": 2.6225337982177734,
+      "reward_std": 0.3525502234697342,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.023299389518797398,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1596
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 350.0833435058594,
+      "epoch": 0.7985,
+      "grad_norm": 4.859278250656306,
+      "kl": 1.57421875,
+      "learning_rate": 2.0679925163694033e-07,
+      "loss": 0.3676,
+      "reward": 2.552080273628235,
+      "reward_std": 0.8139557242393494,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.031253148801624775,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1597
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.68751525878906,
+      "epoch": 0.799,
+      "grad_norm": 7.356529004359635,
+      "kl": 0.640625,
+      "learning_rate": 2.0629178711441115e-07,
+      "loss": 0.0221,
+      "reward": 2.4863110780715942,
+      "reward_std": 0.23228544741868973,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.013688962906599045,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1598
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.2083435058594,
+      "epoch": 0.7995,
+      "grad_norm": 3.7867120654273236,
+      "kl": 0.576171875,
+      "learning_rate": 2.0578536958670574e-07,
+      "loss": 0.0753,
+      "reward": 2.936859369277954,
+      "reward_std": 0.15612836927175522,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.021473943255841732,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1599
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 296.41668701171875,
+      "epoch": 0.8,
+      "grad_norm": 6.667486304605809,
+      "kl": 0.890625,
+      "learning_rate": 2.0528000059645995e-07,
+      "loss": 0.0294,
+      "reward": 2.498704433441162,
+      "reward_std": 0.3463284894824028,
+      "rewards/accuracy_reward": 0.5416666716337204,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.023865243420004845,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 1600
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.0208435058594,
+      "epoch": 0.8005,
+      "grad_norm": 5.392405663450308,
+      "kl": 1.0859375,
+      "learning_rate": 2.0477568168311525e-07,
+      "loss": 0.1415,
+      "reward": 2.3850942850112915,
+      "reward_std": 0.3655809760093689,
+      "rewards/accuracy_reward": 0.4583333358168602,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.015947438776493073,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1601
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 281.0,
+      "epoch": 0.801,
+      "grad_norm": 10.683966440840944,
+      "kl": 0.97265625,
+      "learning_rate": 2.042724143829146e-07,
+      "loss": 0.1412,
+      "reward": 2.4589797258377075,
+      "reward_std": 0.5436213612556458,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.016714807134121656,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1602
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.875,
+      "epoch": 0.8015,
+      "grad_norm": 3.118222730823267,
+      "kl": 0.625,
+      "learning_rate": 2.037702002288973e-07,
+      "loss": 0.0022,
+      "reward": 2.7855957746505737,
+      "reward_std": 0.193808451294899,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.026904198341071606,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1603
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 339.50001525878906,
+      "epoch": 0.802,
+      "grad_norm": 4.604018186494893,
+      "kl": 0.7734375,
+      "learning_rate": 2.032690407508949e-07,
+      "loss": 0.1366,
+      "reward": 2.8414154052734375,
+      "reward_std": 0.21933773159980774,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.04400132969021797,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1604
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 306.93751525878906,
+      "epoch": 0.8025,
+      "grad_norm": 21.004216450205973,
+      "kl": 1.06640625,
+      "learning_rate": 2.027689374755261e-07,
+      "loss": 0.0028,
+      "reward": 2.575451135635376,
+      "reward_std": 0.37137471139431,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.028715766966342926,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1605
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.37501525878906,
+      "epoch": 0.803,
+      "grad_norm": 6.2354379721667526,
+      "kl": 0.53125,
+      "learning_rate": 2.0226989192619204e-07,
+      "loss": 0.0291,
+      "reward": 2.5753469467163086,
+      "reward_std": 0.19728727638721466,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02881983108818531,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1606
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.0833435058594,
+      "epoch": 0.8035,
+      "grad_norm": 8.229728899859898,
+      "kl": 0.8642578125,
+      "learning_rate": 2.0177190562307224e-07,
+      "loss": 0.0185,
+      "reward": 2.852303981781006,
+      "reward_std": 0.3710853382945061,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.024431976955384016,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1607
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 309.3125,
+      "epoch": 0.804,
+      "grad_norm": 10.300857336577062,
+      "kl": 1.041015625,
+      "learning_rate": 2.0127498008311922e-07,
+      "loss": 0.2301,
+      "reward": 2.592196464538574,
+      "reward_std": 0.41349270567297935,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.018914744723588228,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1608
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 316.75,
+      "epoch": 0.8045,
+      "grad_norm": 7.563778647893601,
+      "kl": 1.94921875,
+      "learning_rate": 2.0077911682005428e-07,
+      "loss": 0.1949,
+      "reward": 2.3974047899246216,
+      "reward_std": 0.6470229029655457,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.03141475468873978,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1609
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.9166717529297,
+      "epoch": 0.805,
+      "grad_norm": 7.811338730077243,
+      "kl": 1.5078125,
+      "learning_rate": 2.0028431734436308e-07,
+      "loss": 0.2798,
+      "reward": 2.4768539667129517,
+      "reward_std": 0.4787053167819977,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.028354503214359283,
+      "rewards/tag_count_reward": 0.9427083730697632,
+      "step": 1610
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.25001525878906,
+      "epoch": 0.8055,
+      "grad_norm": 8.594070426030529,
+      "kl": 2.0,
+      "learning_rate": 1.9979058316329055e-07,
+      "loss": 0.3485,
+      "reward": 2.784187912940979,
+      "reward_std": 0.45950163900852203,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.026576073840260506,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1611
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.1875,
+      "epoch": 0.806,
+      "grad_norm": 4.2199364202164,
+      "kl": 0.599609375,
+      "learning_rate": 1.9929791578083655e-07,
+      "loss": 0.0667,
+      "reward": 2.9232022762298584,
+      "reward_std": 0.15263231098651886,
+      "rewards/accuracy_reward": 0.9583333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03513127192854881,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1612
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 308.875,
+      "epoch": 0.8065,
+      "grad_norm": 8.990639470337845,
+      "kl": 0.84765625,
+      "learning_rate": 1.9880631669775162e-07,
+      "loss": 0.2945,
+      "reward": 2.668937563896179,
+      "reward_std": 0.4244493693113327,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666567325592,
+      "rewards/repetition_penalty_reward": -0.023770801723003387,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1613
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.00001525878906,
+      "epoch": 0.807,
+      "grad_norm": 8.09121179675446,
+      "kl": 0.935546875,
+      "learning_rate": 1.9831578741153155e-07,
+      "loss": 0.0507,
+      "reward": 2.8520257472991943,
+      "reward_std": 0.32762178778648376,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.03512696735560894,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1614
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 296.1458435058594,
+      "epoch": 0.8075,
+      "grad_norm": 9.454399169013804,
+      "kl": 0.595703125,
+      "learning_rate": 1.9782632941641375e-07,
+      "loss": 0.0606,
+      "reward": 2.6736570596694946,
+      "reward_std": 0.13973642978817225,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.019051472656428814,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1615
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 251.06250762939453,
+      "epoch": 0.808,
+      "grad_norm": 4.329960507797107,
+      "kl": 0.650390625,
+      "learning_rate": 1.9733794420337213e-07,
+      "loss": -0.0227,
+      "reward": 2.630091071128845,
+      "reward_std": 0.29190193116664886,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.015742299146950245,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1616
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 336.7291717529297,
+      "epoch": 0.8085,
+      "grad_norm": 6.82490142880819,
+      "kl": 1.55078125,
+      "learning_rate": 1.9685063326011263e-07,
+      "loss": 0.3148,
+      "reward": 2.5862770080566406,
+      "reward_std": 0.7006584405899048,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.04219539649784565,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1617
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.9166717529297,
+      "epoch": 0.809,
+      "grad_norm": 7.151058516859466,
+      "kl": 1.71484375,
+      "learning_rate": 1.9636439807106912e-07,
+      "loss": 0.2832,
+      "reward": 2.503158688545227,
+      "reward_std": 0.4477211833000183,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.01941095246002078,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1618
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 245.93750762939453,
+      "epoch": 0.8095,
+      "grad_norm": 4.443081763622589,
+      "kl": 0.625,
+      "learning_rate": 1.9587924011739826e-07,
+      "loss": 0.0627,
+      "reward": 2.9678823947906494,
+      "reward_std": 0.01614804659038782,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03211780823767185,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1619
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 323.87500762939453,
+      "epoch": 0.81,
+      "grad_norm": 9.980771833809246,
+      "kl": 1.87109375,
+      "learning_rate": 1.9539516087697517e-07,
+      "loss": 0.2491,
+      "reward": 2.570975422859192,
+      "reward_std": 0.36428187415003777,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.022774542681872845,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1620
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 337.4375,
+      "epoch": 0.8105,
+      "grad_norm": 5.299585054431229,
+      "kl": 1.01171875,
+      "learning_rate": 1.9491216182438926e-07,
+      "loss": 0.0517,
+      "reward": 2.593162417411804,
+      "reward_std": 0.40712933242321014,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.04399042949080467,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1621
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.9791717529297,
+      "epoch": 0.811,
+      "grad_norm": 6.996029293693946,
+      "kl": 0.671875,
+      "learning_rate": 1.944302444309393e-07,
+      "loss": 0.0078,
+      "reward": 2.6557544469833374,
+      "reward_std": 0.2550486624240875,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.024801287800073624,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1622
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 295.66668701171875,
+      "epoch": 0.8115,
+      "grad_norm": 3.696905547483718,
+      "kl": 1.087890625,
+      "learning_rate": 1.9394941016462947e-07,
+      "loss": 0.0938,
+      "reward": 2.5640480518341064,
+      "reward_std": 0.3755566477775574,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.036646610125899315,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1623
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 314.22918701171875,
+      "epoch": 0.812,
+      "grad_norm": 10.073124148755838,
+      "kl": 1.7578125,
+      "learning_rate": 1.934696604901642e-07,
+      "loss": 0.2146,
+      "reward": 2.8040868043899536,
+      "reward_std": 0.5022929012775421,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.03271871618926525,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1624
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 281.5,
+      "epoch": 0.8125,
+      "grad_norm": 4.120406566625248,
+      "kl": 0.9375,
+      "learning_rate": 1.929909968689442e-07,
+      "loss": 0.057,
+      "reward": 2.469716787338257,
+      "reward_std": 0.394562803208828,
+      "rewards/accuracy_reward": 0.5416666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.035491637885570526,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1625
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.5625,
+      "epoch": 0.813,
+      "grad_norm": 6.662465778476507,
+      "kl": 1.498046875,
+      "learning_rate": 1.9251342075906179e-07,
+      "loss": 0.1782,
+      "reward": 2.5808849334716797,
+      "reward_std": 0.3251846134662628,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.033698457293212414,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1626
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 344.1458435058594,
+      "epoch": 0.8135,
+      "grad_norm": 18.710995596257547,
+      "kl": 2.18359375,
+      "learning_rate": 1.9203693361529687e-07,
+      "loss": 0.2733,
+      "reward": 2.6579350233078003,
+      "reward_std": 0.6114742159843445,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.01914846431463957,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1627
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 269.29168701171875,
+      "epoch": 0.814,
+      "grad_norm": 4.508553972215575,
+      "kl": 0.888671875,
+      "learning_rate": 1.915615368891117e-07,
+      "loss": 0.0428,
+      "reward": 2.79397451877594,
+      "reward_std": 0.3388333395123482,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.023733829148113728,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1628
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 267.5416717529297,
+      "epoch": 0.8145,
+      "grad_norm": 5.521089193451119,
+      "kl": 0.595703125,
+      "learning_rate": 1.9108723202864723e-07,
+      "loss": 0.0554,
+      "reward": 2.9251835346221924,
+      "reward_std": 0.1714300513267517,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.012316623236984015,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1629
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 371.39583587646484,
+      "epoch": 0.815,
+      "grad_norm": 7.336042303384628,
+      "kl": 1.06640625,
+      "learning_rate": 1.9061402047871833e-07,
+      "loss": 0.2051,
+      "reward": 2.7813527584075928,
+      "reward_std": 0.33912873174995184,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.02420270536094904,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1630
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 431.54168701171875,
+      "epoch": 0.8155,
+      "grad_norm": 9.604124750722942,
+      "kl": 1.875,
+      "learning_rate": 1.9014190368080924e-07,
+      "loss": 0.5587,
+      "reward": 2.37904691696167,
+      "reward_std": 0.7319284677505493,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.023730785585939884,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1631
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 319.25,
+      "epoch": 0.816,
+      "grad_norm": 9.391423961724705,
+      "kl": 1.513671875,
+      "learning_rate": 1.8967088307307e-07,
+      "loss": 0.2808,
+      "reward": 2.566171407699585,
+      "reward_std": 0.5563443601131439,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.029314829036593437,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1632
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 324.5416717529297,
+      "epoch": 0.8165,
+      "grad_norm": 6.896452210575316,
+      "kl": 1.6875,
+      "learning_rate": 1.8920096009031072e-07,
+      "loss": 0.2943,
+      "reward": 2.5484179258346558,
+      "reward_std": 0.5188319385051727,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.02449881285429001,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1633
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.00001525878906,
+      "epoch": 0.817,
+      "grad_norm": 6.919405126450731,
+      "kl": 0.716796875,
+      "learning_rate": 1.887321361639985e-07,
+      "loss": 0.0629,
+      "reward": 2.8598101139068604,
+      "reward_std": 0.2988024652004242,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.03428714908659458,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1634
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 295.60418701171875,
+      "epoch": 0.8175,
+      "grad_norm": 10.231296261630002,
+      "kl": 1.177734375,
+      "learning_rate": 1.8826441272225225e-07,
+      "loss": 0.2059,
+      "reward": 2.720053195953369,
+      "reward_std": 0.5625387728214264,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.03168301936239004,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1635
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 336.7083435058594,
+      "epoch": 0.818,
+      "grad_norm": 10.632802931373286,
+      "kl": 1.51171875,
+      "learning_rate": 1.8779779118983867e-07,
+      "loss": 0.5578,
+      "reward": 2.639328956604004,
+      "reward_std": 0.6874139904975891,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.01865724567323923,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1636
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 299.6458435058594,
+      "epoch": 0.8185,
+      "grad_norm": 10.048893605843471,
+      "kl": 1.33203125,
+      "learning_rate": 1.8733227298816794e-07,
+      "loss": 0.3385,
+      "reward": 2.6121253967285156,
+      "reward_std": 0.5836938470602036,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.025027431547641754,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1637
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 314.7708435058594,
+      "epoch": 0.819,
+      "grad_norm": 6.465062212154395,
+      "kl": 0.65625,
+      "learning_rate": 1.8686785953528922e-07,
+      "loss": 0.1366,
+      "reward": 2.602375626564026,
+      "reward_std": 0.4529639333486557,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.026096642017364502,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1638
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 466.66668701171875,
+      "epoch": 0.8195,
+      "grad_norm": 30.96092236002016,
+      "kl": 3.671875,
+      "learning_rate": 1.8640455224588636e-07,
+      "loss": 0.4841,
+      "reward": 2.4283162355422974,
+      "reward_std": 0.6781846880912781,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.8680555820465088,
+      "rewards/repetition_penalty_reward": -0.017864545807242393,
+      "rewards/tag_count_reward": 0.8489583730697632,
+      "step": 1639
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.00001525878906,
+      "epoch": 0.82,
+      "grad_norm": 11.249771941826626,
+      "kl": 2.7109375,
+      "learning_rate": 1.8594235253127372e-07,
+      "loss": 0.2243,
+      "reward": 2.520586609840393,
+      "reward_std": 0.6222244799137115,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.019344151951372623,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1640
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.81251525878906,
+      "epoch": 0.8205,
+      "grad_norm": 4.567943396166058,
+      "kl": 0.7119140625,
+      "learning_rate": 1.8548126179939188e-07,
+      "loss": 0.0507,
+      "reward": 2.6988391876220703,
+      "reward_std": 0.5334462970495224,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.037272000685334206,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1641
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.87500762939453,
+      "epoch": 0.821,
+      "grad_norm": 5.377391931963283,
+      "kl": 0.59375,
+      "learning_rate": 1.850212814548031e-07,
+      "loss": 0.0612,
+      "reward": 2.5510483980178833,
+      "reward_std": 0.32321809232234955,
+      "rewards/accuracy_reward": 0.5833333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.021868368610739708,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1642
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 383.87501525878906,
+      "epoch": 0.8215,
+      "grad_norm": 10.395746122624809,
+      "kl": 2.015625,
+      "learning_rate": 1.8456241289868718e-07,
+      "loss": 0.3281,
+      "reward": 2.2994015216827393,
+      "reward_std": 0.4463878870010376,
+      "rewards/accuracy_reward": 0.4166666679084301,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.023515181615948677,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1643
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 309.7083435058594,
+      "epoch": 0.822,
+      "grad_norm": 9.658762972884702,
+      "kl": 0.984375,
+      "learning_rate": 1.8410465752883758e-07,
+      "loss": 0.3969,
+      "reward": 2.853973150253296,
+      "reward_std": 0.33096832036972046,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.017554799327626824,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1644
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 295.3958435058594,
+      "epoch": 0.8225,
+      "grad_norm": 40.51819002173804,
+      "kl": 1.42578125,
+      "learning_rate": 1.8364801673965642e-07,
+      "loss": 0.287,
+      "reward": 2.553430914878845,
+      "reward_std": 0.580101728439331,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.022958120796829462,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1645
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 342.06251525878906,
+      "epoch": 0.823,
+      "grad_norm": 7.09616651885635,
+      "kl": 1.55859375,
+      "learning_rate": 1.8319249192215055e-07,
+      "loss": 0.5539,
+      "reward": 2.7683907747268677,
+      "reward_std": 0.487443208694458,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.035428643226623535,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1646
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 525.1875305175781,
+      "epoch": 0.8235,
+      "grad_norm": 22.444784133823088,
+      "kl": 3.859375,
+      "learning_rate": 1.8273808446392785e-07,
+      "loss": 0.5689,
+      "reward": 2.390328526496887,
+      "reward_std": 0.5453556627035141,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.8611111640930176,
+      "rewards/repetition_penalty_reward": -0.0072408574633300304,
+      "rewards/tag_count_reward": 0.8281250298023224,
+      "step": 1647
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 402.1875,
+      "epoch": 0.824,
+      "grad_norm": 7.4531870630991515,
+      "kl": 2.71875,
+      "learning_rate": 1.822847957491922e-07,
+      "loss": 0.65,
+      "reward": 2.406116247177124,
+      "reward_std": 0.7253749072551727,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9097221791744232,
+      "rewards/repetition_penalty_reward": -0.019231081008911133,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1648
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 375.25001525878906,
+      "epoch": 0.8245,
+      "grad_norm": 4.1340736903663595,
+      "kl": 1.33984375,
+      "learning_rate": 1.8183262715873938e-07,
+      "loss": 0.3177,
+      "reward": 2.4100518226623535,
+      "reward_std": 0.6681110262870789,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.023975889198482037,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1649
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 306.4583435058594,
+      "epoch": 0.825,
+      "grad_norm": 7.486120901127916,
+      "kl": 0.978515625,
+      "learning_rate": 1.8138158006995363e-07,
+      "loss": 0.1757,
+      "reward": 2.8634815216064453,
+      "reward_std": 0.3373248726129532,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.042768435552716255,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1650
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 345.66668701171875,
+      "epoch": 0.8255,
+      "grad_norm": 9.259229368359508,
+      "kl": 1.505859375,
+      "learning_rate": 1.8093165585680253e-07,
+      "loss": 0.3075,
+      "reward": 2.573695659637451,
+      "reward_std": 0.47858020663261414,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.021790657192468643,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1651
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 308.62501525878906,
+      "epoch": 0.826,
+      "grad_norm": 5.948718639971257,
+      "kl": 0.59765625,
+      "learning_rate": 1.804828558898332e-07,
+      "loss": -0.0333,
+      "reward": 2.6328206062316895,
+      "reward_std": 0.3835385888814926,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.032110003754496574,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1652
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 416.81251525878906,
+      "epoch": 0.8265,
+      "grad_norm": 5.733286623959369,
+      "kl": 1.5546875,
+      "learning_rate": 1.800351815361682e-07,
+      "loss": 0.4485,
+      "reward": 2.4559801816940308,
+      "reward_std": 0.6161761581897736,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.021450520493090153,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1653
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 379.1458435058594,
+      "epoch": 0.827,
+      "grad_norm": 5.107287414386508,
+      "kl": 1.21484375,
+      "learning_rate": 1.7958863415950112e-07,
+      "loss": 0.1107,
+      "reward": 2.421965479850769,
+      "reward_std": 0.3980814069509506,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.029423545114696026,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1654
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.18751525878906,
+      "epoch": 0.8275,
+      "grad_norm": 6.952303501359962,
+      "kl": 0.76171875,
+      "learning_rate": 1.7914321512009296e-07,
+      "loss": 0.0656,
+      "reward": 2.8843663930892944,
+      "reward_std": 0.22652451507747173,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.02014744747430086,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 1655
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 476.1666717529297,
+      "epoch": 0.828,
+      "grad_norm": 10.341849236868242,
+      "kl": 2.90234375,
+      "learning_rate": 1.7869892577476722e-07,
+      "loss": 0.4727,
+      "reward": 2.3343173265457153,
+      "reward_std": 0.7950380742549896,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.8472222983837128,
+      "rewards/repetition_penalty_reward": -0.033738273195922375,
+      "rewards/tag_count_reward": 0.8333333432674408,
+      "step": 1656
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 339.7083435058594,
+      "epoch": 0.8285,
+      "grad_norm": 7.640794732443528,
+      "kl": 1.568359375,
+      "learning_rate": 1.782557674769063e-07,
+      "loss": 0.2939,
+      "reward": 2.689374566078186,
+      "reward_std": 0.424824059009552,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.022430949844419956,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1657
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 525.875,
+      "epoch": 0.829,
+      "grad_norm": 11.914470370020899,
+      "kl": 2.82421875,
+      "learning_rate": 1.7781374157644713e-07,
+      "loss": 0.6357,
+      "reward": 2.147131085395813,
+      "reward_std": 0.5802744626998901,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.016063490882515907,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1658
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 406.5416717529297,
+      "epoch": 0.8295,
+      "grad_norm": 10.946757964721602,
+      "kl": 3.1015625,
+      "learning_rate": 1.773728494198775e-07,
+      "loss": 0.4053,
+      "reward": 2.3653860092163086,
+      "reward_std": 0.5210855007171631,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.021766817197203636,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1659
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 405.5208435058594,
+      "epoch": 0.83,
+      "grad_norm": 15.491914941420152,
+      "kl": 0.91796875,
+      "learning_rate": 1.7693309235023127e-07,
+      "loss": 0.2427,
+      "reward": 2.8351590633392334,
+      "reward_std": 0.33579762279987335,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03463265113532543,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1660
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 342.60418701171875,
+      "epoch": 0.8305,
+      "grad_norm": 13.255170819740858,
+      "kl": 1.369140625,
+      "learning_rate": 1.7649447170708466e-07,
+      "loss": 0.5412,
+      "reward": 2.8557145595550537,
+      "reward_std": 0.31544718984514475,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02796613797545433,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1661
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 392.9166717529297,
+      "epoch": 0.831,
+      "grad_norm": 10.898791772241301,
+      "kl": 2.2890625,
+      "learning_rate": 1.7605698882655233e-07,
+      "loss": 0.4671,
+      "reward": 2.5021458864212036,
+      "reward_std": 0.6049044132232666,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.027368071489036083,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1662
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 318.0833435058594,
+      "epoch": 0.8315,
+      "grad_norm": 7.104099769758903,
+      "kl": 1.33203125,
+      "learning_rate": 1.7562064504128281e-07,
+      "loss": 0.4717,
+      "reward": 2.5585243701934814,
+      "reward_std": 0.5537194460630417,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.017864540684968233,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1663
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.4583435058594,
+      "epoch": 0.832,
+      "grad_norm": 7.006000014838191,
+      "kl": 1.06640625,
+      "learning_rate": 1.7518544168045524e-07,
+      "loss": 0.3078,
+      "reward": 2.647615075111389,
+      "reward_std": 0.2714964300394058,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.025996237061917782,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1664
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 381.5625,
+      "epoch": 0.8325,
+      "grad_norm": 25.862084169657674,
+      "kl": 2.109375,
+      "learning_rate": 1.7475138006977437e-07,
+      "loss": 0.6256,
+      "reward": 2.650961399078369,
+      "reward_std": 0.5018740892410278,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.012233282905071974,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1665
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.9583435058594,
+      "epoch": 0.833,
+      "grad_norm": 7.442304533677901,
+      "kl": 1.189453125,
+      "learning_rate": 1.743184615314671e-07,
+      "loss": 0.1916,
+      "reward": 2.9172674417495728,
+      "reward_std": 0.1536797545850277,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.023704865016043186,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1666
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 488.93751525878906,
+      "epoch": 0.8335,
+      "grad_norm": 17.015693320793893,
+      "kl": 2.3828125,
+      "learning_rate": 1.7388668738427847e-07,
+      "loss": 0.5714,
+      "reward": 2.4601612091064453,
+      "reward_std": 0.6804801672697067,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.020741629414260387,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1667
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 343.0833435058594,
+      "epoch": 0.834,
+      "grad_norm": 6.622283982214126,
+      "kl": 1.80078125,
+      "learning_rate": 1.7345605894346726e-07,
+      "loss": 0.2748,
+      "reward": 2.1607359647750854,
+      "reward_std": 0.5849722027778625,
+      "rewards/accuracy_reward": 0.3541666716337204,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.02676413208246231,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1668
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 435.8125,
+      "epoch": 0.8345,
+      "grad_norm": 10.366061275777671,
+      "kl": 1.87890625,
+      "learning_rate": 1.7302657752080258e-07,
+      "loss": 0.5968,
+      "reward": 2.4938762187957764,
+      "reward_std": 0.6266940236091614,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.023485003039240837,
+      "rewards/tag_count_reward": 0.9062500298023224,
+      "step": 1669
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 306.9791717529297,
+      "epoch": 0.835,
+      "grad_norm": 6.471286087528944,
+      "kl": 0.912109375,
+      "learning_rate": 1.7259824442455923e-07,
+      "loss": 0.1973,
+      "reward": 2.6409337520599365,
+      "reward_std": 0.41041404008865356,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.023996802046895027,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1670
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 344.3958435058594,
+      "epoch": 0.8355,
+      "grad_norm": 8.18949050435243,
+      "kl": 2.07421875,
+      "learning_rate": 1.7217106095951412e-07,
+      "loss": 0.4581,
+      "reward": 2.5663615465164185,
+      "reward_std": 0.6968798041343689,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.015235766302794218,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1671
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 440.8958435058594,
+      "epoch": 0.836,
+      "grad_norm": 11.296052099828927,
+      "kl": 2.265625,
+      "learning_rate": 1.7174502842694212e-07,
+      "loss": 0.5144,
+      "reward": 2.2743382453918457,
+      "reward_std": 0.6736903786659241,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.01038410421460867,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1672
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 432.4583435058594,
+      "epoch": 0.8365,
+      "grad_norm": 11.962173628260595,
+      "kl": 2.888671875,
+      "learning_rate": 1.7132014812461227e-07,
+      "loss": 0.3413,
+      "reward": 2.4873476028442383,
+      "reward_std": 0.47372131049633026,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.021332964301109314,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1673
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 399.37501525878906,
+      "epoch": 0.837,
+      "grad_norm": 4.56941687729545,
+      "kl": 1.314453125,
+      "learning_rate": 1.7089642134678364e-07,
+      "loss": 0.165,
+      "reward": 2.6062146425247192,
+      "reward_std": 0.4681692570447922,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0344105139374733,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1674
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 418.12501525878906,
+      "epoch": 0.8375,
+      "grad_norm": 16.236737665323922,
+      "kl": 1.578125,
+      "learning_rate": 1.704738493842015e-07,
+      "loss": 0.5058,
+      "reward": 2.3385233879089355,
+      "reward_std": 0.7678199410438538,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9097222983837128,
+      "rewards/repetition_penalty_reward": -0.013907157350331545,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1675
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 344.7916717529297,
+      "epoch": 0.838,
+      "grad_norm": 8.549333095195593,
+      "kl": 2.1328125,
+      "learning_rate": 1.7005243352409333e-07,
+      "loss": 0.175,
+      "reward": 2.302179217338562,
+      "reward_std": 0.8448854684829712,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.8819444477558136,
+      "rewards/repetition_penalty_reward": -0.03289027698338032,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1676
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.9791717529297,
+      "epoch": 0.8385,
+      "grad_norm": 4.575278870826187,
+      "kl": 1.41015625,
+      "learning_rate": 1.6963217505016475e-07,
+      "loss": 0.1386,
+      "reward": 2.2787574529647827,
+      "reward_std": 0.3729303479194641,
+      "rewards/accuracy_reward": 0.3750000111758709,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.016381369438022375,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1677
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 343.5208435058594,
+      "epoch": 0.839,
+      "grad_norm": 4.195422017395303,
+      "kl": 1.482421875,
+      "learning_rate": 1.6921307524259625e-07,
+      "loss": 0.1819,
+      "reward": 2.7099480628967285,
+      "reward_std": 0.41559091210365295,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.022690830752253532,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1678
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 359.0208435058594,
+      "epoch": 0.8395,
+      "grad_norm": 11.950641286729567,
+      "kl": 0.943359375,
+      "learning_rate": 1.6879513537803839e-07,
+      "loss": 0.453,
+      "reward": 2.5649707317352295,
+      "reward_std": 0.39247044920921326,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.02183501608669758,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1679
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 447.4166717529297,
+      "epoch": 0.84,
+      "grad_norm": 8.687575451435944,
+      "kl": 1.419921875,
+      "learning_rate": 1.6837835672960831e-07,
+      "loss": 0.5576,
+      "reward": 2.619246244430542,
+      "reward_std": 0.6550075113773346,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.028323073871433735,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1680
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 277.83333587646484,
+      "epoch": 0.8405,
+      "grad_norm": 7.023805921675465,
+      "kl": 1.224609375,
+      "learning_rate": 1.6796274056688637e-07,
+      "loss": 0.0823,
+      "reward": 2.508470058441162,
+      "reward_std": 0.3732000142335892,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.014099352061748505,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 1681
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 349.25001525878906,
+      "epoch": 0.841,
+      "grad_norm": 6.786127817511959,
+      "kl": 1.1328125,
+      "learning_rate": 1.6754828815591131e-07,
+      "loss": 0.29,
+      "reward": 2.645016670227051,
+      "reward_std": 0.5615668296813965,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.018177741672843695,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1682
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 533.6458435058594,
+      "epoch": 0.8415,
+      "grad_norm": 37.07630929388199,
+      "kl": 5.0625,
+      "learning_rate": 1.6713500075917694e-07,
+      "loss": 0.5224,
+      "reward": 1.9782096147537231,
+      "reward_std": 0.7190957963466644,
+      "rewards/accuracy_reward": 0.291666679084301,
+      "rewards/reasoning_steps_reward": 0.8750000596046448,
+      "rewards/repetition_penalty_reward": -0.026998871937394142,
+      "rewards/tag_count_reward": 0.8385416865348816,
+      "step": 1683
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 509.54168701171875,
+      "epoch": 0.842,
+      "grad_norm": 25.945658463095125,
+      "kl": 2.25390625,
+      "learning_rate": 1.6672287963562852e-07,
+      "loss": 0.5761,
+      "reward": 2.5873239040374756,
+      "reward_std": 0.6155489385128021,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.8888890147209167,
+      "rewards/repetition_penalty_reward": -0.030731705483049154,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1684
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 443.7916717529297,
+      "epoch": 0.8425,
+      "grad_norm": 12.106839765407578,
+      "kl": 1.5078125,
+      "learning_rate": 1.6631192604065852e-07,
+      "loss": 0.4876,
+      "reward": 2.326740026473999,
+      "reward_std": 0.4968672841787338,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.02395441848784685,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1685
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.0833435058594,
+      "epoch": 0.843,
+      "grad_norm": 40.28917535782033,
+      "kl": 1.04296875,
+      "learning_rate": 1.659021412261026e-07,
+      "loss": 0.3504,
+      "reward": 2.5606080293655396,
+      "reward_std": 0.3592444807291031,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.014044871088117361,
+      "rewards/tag_count_reward": 0.9843750298023224,
+      "step": 1686
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 361.2708435058594,
+      "epoch": 0.8435,
+      "grad_norm": 6.73864611491594,
+      "kl": 1.017578125,
+      "learning_rate": 1.6549352644023668e-07,
+      "loss": 0.2952,
+      "reward": 2.494922637939453,
+      "reward_std": 0.2953641563653946,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.02764677256345749,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1687
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 417.3333435058594,
+      "epoch": 0.844,
+      "grad_norm": 8.47333082669667,
+      "kl": 1.14453125,
+      "learning_rate": 1.6508608292777203e-07,
+      "loss": 0.3733,
+      "reward": 2.648282527923584,
+      "reward_std": 0.3749905973672867,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.025328767485916615,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1688
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.3125,
+      "epoch": 0.8445,
+      "grad_norm": 10.455138386711887,
+      "kl": 1.40234375,
+      "learning_rate": 1.646798119298523e-07,
+      "loss": 0.2977,
+      "reward": 2.5954939126968384,
+      "reward_std": 0.37971626222133636,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.026033984497189522,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1689
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 399.25001525878906,
+      "epoch": 0.845,
+      "grad_norm": 10.80867448506658,
+      "kl": 1.30078125,
+      "learning_rate": 1.6427471468404952e-07,
+      "loss": 0.3745,
+      "reward": 2.5844664573669434,
+      "reward_std": 0.5553542077541351,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.03532520309090614,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1690
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 328.75,
+      "epoch": 0.8455,
+      "grad_norm": 10.515772073918267,
+      "kl": 1.103515625,
+      "learning_rate": 1.6387079242435995e-07,
+      "loss": 0.4825,
+      "reward": 2.847057342529297,
+      "reward_std": 0.3490803837776184,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.010581327602267265,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1691
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 380.50001525878906,
+      "epoch": 0.846,
+      "grad_norm": 9.354849527674562,
+      "kl": 1.859375,
+      "learning_rate": 1.6346804638120098e-07,
+      "loss": 0.3717,
+      "reward": 2.734411835670471,
+      "reward_std": 0.4372542053461075,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.010379912098869681,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1692
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 461.54168701171875,
+      "epoch": 0.8465,
+      "grad_norm": 10.81389100794866,
+      "kl": 1.71875,
+      "learning_rate": 1.6306647778140697e-07,
+      "loss": 0.8252,
+      "reward": 2.553888440132141,
+      "reward_std": 0.7343409359455109,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.022500536404550076,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1693
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 650.8333435058594,
+      "epoch": 0.847,
+      "grad_norm": 18.03694586453546,
+      "kl": 2.8359375,
+      "learning_rate": 1.6266608784822542e-07,
+      "loss": 0.5358,
+      "reward": 1.9123651385307312,
+      "reward_std": 0.5860169529914856,
+      "rewards/accuracy_reward": 0.18750000558793545,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.02166277915239334,
+      "rewards/tag_count_reward": 0.8437500298023224,
+      "step": 1694
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 634.6041870117188,
+      "epoch": 0.8475,
+      "grad_norm": 13.570448263572654,
+      "kl": 3.578125,
+      "learning_rate": 1.6226687780131337e-07,
+      "loss": 0.9188,
+      "reward": 2.1532905101776123,
+      "reward_std": 0.8254519104957581,
+      "rewards/accuracy_reward": 0.4791666716337204,
+      "rewards/reasoning_steps_reward": 0.8888888955116272,
+      "rewards/repetition_penalty_reward": -0.01684834063053131,
+      "rewards/tag_count_reward": 0.8020833432674408,
+      "step": 1695
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 366.7708435058594,
+      "epoch": 0.848,
+      "grad_norm": 9.145183407264888,
+      "kl": 0.953125,
+      "learning_rate": 1.6186884885673413e-07,
+      "loss": 0.443,
+      "reward": 2.3256582021713257,
+      "reward_std": 0.47854430973529816,
+      "rewards/accuracy_reward": 0.4166666716337204,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.016355796717107296,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1696
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 324.1458435058594,
+      "epoch": 0.8485,
+      "grad_norm": 13.420921446394319,
+      "kl": 1.810546875,
+      "learning_rate": 1.6147200222695275e-07,
+      "loss": 0.2714,
+      "reward": 2.748153805732727,
+      "reward_std": 0.5222579091787338,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.026151607744395733,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1697
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 379.4791717529297,
+      "epoch": 0.849,
+      "grad_norm": 16.712319193882028,
+      "kl": 1.3515625,
+      "learning_rate": 1.610763391208329e-07,
+      "loss": 0.4492,
+      "reward": 2.480373501777649,
+      "reward_std": 0.4376864656805992,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.019626670517027378,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1698
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 335.68751525878906,
+      "epoch": 0.8495,
+      "grad_norm": 8.460670702699383,
+      "kl": 1.15234375,
+      "learning_rate": 1.6068186074363307e-07,
+      "loss": 0.4626,
+      "reward": 2.559247851371765,
+      "reward_std": 0.5322316884994507,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.01887742755934596,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1699
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 600.6875305175781,
+      "epoch": 0.85,
+      "grad_norm": 15.150759771270096,
+      "kl": 2.6953125,
+      "learning_rate": 1.6028856829700258e-07,
+      "loss": 0.5681,
+      "reward": 2.267993927001953,
+      "reward_std": 0.7472598850727081,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.8958334028720856,
+      "rewards/repetition_penalty_reward": -0.018464550375938416,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1700
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 429.2916717529297,
+      "epoch": 0.8505,
+      "grad_norm": 13.06168321828285,
+      "kl": 1.9921875,
+      "learning_rate": 1.5989646297897876e-07,
+      "loss": 0.3076,
+      "reward": 2.5890194177627563,
+      "reward_std": 0.5929211974143982,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.030772419180721045,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1701
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 471.2916717529297,
+      "epoch": 0.851,
+      "grad_norm": 5.600339328654864,
+      "kl": 1.7421875,
+      "learning_rate": 1.5950554598398228e-07,
+      "loss": 0.523,
+      "reward": 2.2353298664093018,
+      "reward_std": 0.478582501411438,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.026822968386113644,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1702
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 443.70835876464844,
+      "epoch": 0.8515,
+      "grad_norm": 10.128996657974916,
+      "kl": 1.298828125,
+      "learning_rate": 1.59115818502814e-07,
+      "loss": 0.6875,
+      "reward": 2.5388582944869995,
+      "reward_std": 0.3540804013609886,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.04273904860019684,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1703
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 291.5833435058594,
+      "epoch": 0.852,
+      "grad_norm": 8.164831996051532,
+      "kl": 1.609375,
+      "learning_rate": 1.5872728172265146e-07,
+      "loss": 0.0751,
+      "reward": 2.435483932495117,
+      "reward_std": 0.7406750321388245,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.02111327089369297,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1704
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 278.9375,
+      "epoch": 0.8525,
+      "grad_norm": 10.858353319143736,
+      "kl": 0.80859375,
+      "learning_rate": 1.5833993682704515e-07,
+      "loss": 0.2179,
+      "reward": 2.5048502683639526,
+      "reward_std": 0.344282865524292,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.02119144331663847,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1705
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 446.3541717529297,
+      "epoch": 0.853,
+      "grad_norm": 5.175917884900346,
+      "kl": 1.5703125,
+      "learning_rate": 1.579537849959148e-07,
+      "loss": 0.3352,
+      "reward": 2.296080231666565,
+      "reward_std": 0.6603887677192688,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.03378090541809797,
+      "rewards/tag_count_reward": 0.9270833730697632,
+      "step": 1706
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 358.0625,
+      "epoch": 0.8535,
+      "grad_norm": 5.15758064164511,
+      "kl": 1.333984375,
+      "learning_rate": 1.5756882740554578e-07,
+      "loss": 0.3919,
+      "reward": 2.4720072746276855,
+      "reward_std": 0.41646583657711744,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.02972903847694397,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1707
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 458.00001525878906,
+      "epoch": 0.854,
+      "grad_norm": 17.280373579635196,
+      "kl": 1.3203125,
+      "learning_rate": 1.5718506522858572e-07,
+      "loss": 0.5052,
+      "reward": 2.531672954559326,
+      "reward_std": 0.6426400542259216,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.029090996831655502,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1708
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 551.5833435058594,
+      "epoch": 0.8545,
+      "grad_norm": 7.826892273231986,
+      "kl": 2.5625,
+      "learning_rate": 1.5680249963404065e-07,
+      "loss": 0.8511,
+      "reward": 2.2978310585021973,
+      "reward_std": 0.8033435940742493,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.8888888955116272,
+      "rewards/repetition_penalty_reward": -0.03897454775869846,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1709
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 312.68751525878906,
+      "epoch": 0.855,
+      "grad_norm": 5.338178479303831,
+      "kl": 0.853515625,
+      "learning_rate": 1.5642113178727193e-07,
+      "loss": 0.2184,
+      "reward": 2.568590521812439,
+      "reward_std": 0.4426009804010391,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.01821515103802085,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1710
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 431.75001525878906,
+      "epoch": 0.8555,
+      "grad_norm": 5.124987048089,
+      "kl": 1.8359375,
+      "learning_rate": 1.56040962849992e-07,
+      "loss": 0.4555,
+      "reward": 2.6776022911071777,
+      "reward_std": 0.5898395925760269,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.01684217620640993,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1711
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 439.25001525878906,
+      "epoch": 0.856,
+      "grad_norm": 7.502268383392246,
+      "kl": 2.271484375,
+      "learning_rate": 1.5566199398026147e-07,
+      "loss": 0.2842,
+      "reward": 2.5188140869140625,
+      "reward_std": 0.5468065068125725,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.02285270020365715,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1712
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 527.1458587646484,
+      "epoch": 0.8565,
+      "grad_norm": 8.654666055300138,
+      "kl": 1.9453125,
+      "learning_rate": 1.5528422633248516e-07,
+      "loss": 0.3654,
+      "reward": 2.467902421951294,
+      "reward_std": 0.7317368686199188,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.0407780222594738,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1713
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 520.5208587646484,
+      "epoch": 0.857,
+      "grad_norm": 5.651385244412516,
+      "kl": 2.1875,
+      "learning_rate": 1.5490766105740876e-07,
+      "loss": 0.6816,
+      "reward": 2.492395281791687,
+      "reward_std": 0.4476168677210808,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.01454922091215849,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1714
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 411.875,
+      "epoch": 0.8575,
+      "grad_norm": 9.849249493770895,
+      "kl": 1.8671875,
+      "learning_rate": 1.5453229930211563e-07,
+      "loss": 0.6325,
+      "reward": 2.4678452014923096,
+      "reward_std": 0.4792132079601288,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.02173815481364727,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1715
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 421.7708435058594,
+      "epoch": 0.858,
+      "grad_norm": 15.301189903888302,
+      "kl": 2.1796875,
+      "learning_rate": 1.5415814221002265e-07,
+      "loss": 0.2572,
+      "reward": 2.3374699354171753,
+      "reward_std": 0.4645020067691803,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.023641261272132397,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1716
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 325.18750762939453,
+      "epoch": 0.8585,
+      "grad_norm": 6.930860696826815,
+      "kl": 1.078125,
+      "learning_rate": 1.5378519092087712e-07,
+      "loss": 0.246,
+      "reward": 2.599347233772278,
+      "reward_std": 0.4288046956062317,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.013500066474080086,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1717
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 296.0208435058594,
+      "epoch": 0.859,
+      "grad_norm": 6.893289627917303,
+      "kl": 0.708984375,
+      "learning_rate": 1.5341344657075354e-07,
+      "loss": 0.0453,
+      "reward": 2.7252657413482666,
+      "reward_std": 0.38875049352645874,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.0212620310485363,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1718
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 377.62501525878906,
+      "epoch": 0.8595,
+      "grad_norm": 5.587920983816194,
+      "kl": 1.4921875,
+      "learning_rate": 1.5304291029204954e-07,
+      "loss": 0.4535,
+      "reward": 2.5673869848251343,
+      "reward_std": 0.7304165959358215,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.022890993393957615,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1719
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 476.1041717529297,
+      "epoch": 0.86,
+      "grad_norm": 5.569410661498516,
+      "kl": 1.2109375,
+      "learning_rate": 1.5267358321348285e-07,
+      "loss": 0.4086,
+      "reward": 2.5152324438095093,
+      "reward_std": 0.5040201544761658,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.035114867612719536,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1720
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 509.18751525878906,
+      "epoch": 0.8605,
+      "grad_norm": 9.547003786726727,
+      "kl": 2.05078125,
+      "learning_rate": 1.5230546646008792e-07,
+      "loss": 0.3872,
+      "reward": 2.3030202388763428,
+      "reward_std": 0.302436888217926,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9166667461395264,
+      "rewards/repetition_penalty_reward": -0.030313138850033283,
+      "rewards/tag_count_reward": 0.8958333730697632,
+      "step": 1721
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 349.2708435058594,
+      "epoch": 0.861,
+      "grad_norm": 4.655955530469267,
+      "kl": 1.0625,
+      "learning_rate": 1.5193856115321224e-07,
+      "loss": 0.2487,
+      "reward": 2.71955668926239,
+      "reward_std": 0.43974626809358597,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.02349890023469925,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1722
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 318.7083435058594,
+      "epoch": 0.8615,
+      "grad_norm": 3.532375552852986,
+      "kl": 0.6015625,
+      "learning_rate": 1.5157286841051285e-07,
+      "loss": 0.0244,
+      "reward": 2.9468058347702026,
+      "reward_std": 0.07718131458386779,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.032360827550292015,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1723
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.9166717529297,
+      "epoch": 0.862,
+      "grad_norm": 11.354635521002548,
+      "kl": 0.9140625,
+      "learning_rate": 1.5120838934595337e-07,
+      "loss": 0.2791,
+      "reward": 2.9219456911087036,
+      "reward_std": 0.1730342721566558,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.01381842652335763,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1724
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 328.5833435058594,
+      "epoch": 0.8625,
+      "grad_norm": 6.092509316447617,
+      "kl": 1.162109375,
+      "learning_rate": 1.5084512506980023e-07,
+      "loss": 0.1386,
+      "reward": 2.519277572631836,
+      "reward_std": 0.3455023765563965,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.027597556822001934,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1725
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 442.3125305175781,
+      "epoch": 0.863,
+      "grad_norm": 6.529119686425871,
+      "kl": 1.9296875,
+      "learning_rate": 1.5048307668861947e-07,
+      "loss": 0.451,
+      "reward": 2.341726541519165,
+      "reward_std": 0.5323204696178436,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.02632923796772957,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1726
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 502.6458435058594,
+      "epoch": 0.8635,
+      "grad_norm": 11.398809580135172,
+      "kl": 1.892578125,
+      "learning_rate": 1.5012224530527297e-07,
+      "loss": 0.2973,
+      "reward": 2.4390887022018433,
+      "reward_std": 0.264788331463933,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.029661373235285282,
+      "rewards/tag_count_reward": 0.90625,
+      "step": 1727
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 328.5208435058594,
+      "epoch": 0.864,
+      "grad_norm": 7.130945265677619,
+      "kl": 1.052734375,
+      "learning_rate": 1.4976263201891613e-07,
+      "loss": 0.2046,
+      "reward": 2.761821746826172,
+      "reward_std": 0.5065120309591293,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.017692207358777523,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1728
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 373.8333435058594,
+      "epoch": 0.8645,
+      "grad_norm": 7.459954968059489,
+      "kl": 1.54296875,
+      "learning_rate": 1.4940423792499306e-07,
+      "loss": 0.4104,
+      "reward": 2.648692011833191,
+      "reward_std": 0.5437482595443726,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.02318317536264658,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1729
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 354.7083435058594,
+      "epoch": 0.865,
+      "grad_norm": 8.31639308568495,
+      "kl": 1.419921875,
+      "learning_rate": 1.4904706411523448e-07,
+      "loss": 0.277,
+      "reward": 2.6593313217163086,
+      "reward_std": 0.50356225669384,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.02122433064505458,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1730
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 424.2708435058594,
+      "epoch": 0.8655,
+      "grad_norm": 6.80725176888497,
+      "kl": 1.158203125,
+      "learning_rate": 1.4869111167765372e-07,
+      "loss": 0.4757,
+      "reward": 2.588240623474121,
+      "reward_std": 0.5882539600133896,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.019398383796215057,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1731
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 403.8541717529297,
+      "epoch": 0.866,
+      "grad_norm": 10.635761448282858,
+      "kl": 1.3984375,
+      "learning_rate": 1.483363816965435e-07,
+      "loss": 0.5591,
+      "reward": 2.674522638320923,
+      "reward_std": 0.6385620087385178,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.016449684277176857,
+      "rewards/tag_count_reward": 0.9270833730697632,
+      "step": 1732
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 430.35418701171875,
+      "epoch": 0.8665,
+      "grad_norm": 14.223874714304245,
+      "kl": 1.41015625,
+      "learning_rate": 1.479828752524731e-07,
+      "loss": 0.3242,
+      "reward": 2.7146769762039185,
+      "reward_std": 0.4512895792722702,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.021434213500469923,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1733
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 399.75001525878906,
+      "epoch": 0.867,
+      "grad_norm": 8.449702624326806,
+      "kl": 1.3359375,
+      "learning_rate": 1.4763059342228434e-07,
+      "loss": 0.4984,
+      "reward": 2.7380974292755127,
+      "reward_std": 0.6073452234268188,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.030999962240457535,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1734
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 333.4166717529297,
+      "epoch": 0.8675,
+      "grad_norm": 4.452898713963573,
+      "kl": 0.64453125,
+      "learning_rate": 1.4727953727908877e-07,
+      "loss": 0.1877,
+      "reward": 2.59316086769104,
+      "reward_std": 0.3150123804807663,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.030102994292974472,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1735
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 376.7291717529297,
+      "epoch": 0.868,
+      "grad_norm": 6.27878798387463,
+      "kl": 1.08203125,
+      "learning_rate": 1.469297078922642e-07,
+      "loss": 0.4336,
+      "reward": 2.635421872138977,
+      "reward_std": 0.31346043944358826,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.04166135564446449,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1736
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 377.8541717529297,
+      "epoch": 0.8685,
+      "grad_norm": 4.719740434148159,
+      "kl": 1.3046875,
+      "learning_rate": 1.4658110632745174e-07,
+      "loss": 0.2813,
+      "reward": 2.764721632003784,
+      "reward_std": 0.43931836541742086,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.025209171697497368,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1737
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 343.62501525878906,
+      "epoch": 0.869,
+      "grad_norm": 5.45166053562493,
+      "kl": 1.02734375,
+      "learning_rate": 1.4623373364655223e-07,
+      "loss": 0.3308,
+      "reward": 2.8263684511184692,
+      "reward_std": 0.26489363610744476,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.027798308990895748,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1738
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.4375,
+      "epoch": 0.8695,
+      "grad_norm": 11.37070470078536,
+      "kl": 1.609375,
+      "learning_rate": 1.45887590907723e-07,
+      "loss": 0.3803,
+      "reward": 2.567717671394348,
+      "reward_std": 0.7005654871463776,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.019088009372353554,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1739
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 534.7291717529297,
+      "epoch": 0.87,
+      "grad_norm": 15.389689880178889,
+      "kl": 2.359375,
+      "learning_rate": 1.4554267916537495e-07,
+      "loss": 0.8358,
+      "reward": 2.2046847343444824,
+      "reward_std": 0.6854039132595062,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.909722238779068,
+      "rewards/repetition_penalty_reward": -0.017537596635520458,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1740
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 422.6041717529297,
+      "epoch": 0.8705,
+      "grad_norm": 13.823047244979128,
+      "kl": 1.76953125,
+      "learning_rate": 1.4519899947016888e-07,
+      "loss": 0.5483,
+      "reward": 2.4729528427124023,
+      "reward_std": 0.536044716835022,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.014894509688019753,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1741
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 407.02085876464844,
+      "epoch": 0.871,
+      "grad_norm": 10.689522653892224,
+      "kl": 1.3583984375,
+      "learning_rate": 1.448565528690129e-07,
+      "loss": 0.4994,
+      "reward": 2.4924492835998535,
+      "reward_std": 0.46057581901550293,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.02838408574461937,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1742
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 329.5208435058594,
+      "epoch": 0.8715,
+      "grad_norm": 4.741298253263598,
+      "kl": 0.931640625,
+      "learning_rate": 1.4451534040505881e-07,
+      "loss": 0.1265,
+      "reward": 2.8301846981048584,
+      "reward_std": 0.25126277655363083,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03613481484353542,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1743
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 393.8541717529297,
+      "epoch": 0.872,
+      "grad_norm": 17.650431883419923,
+      "kl": 2.62890625,
+      "learning_rate": 1.4417536311769885e-07,
+      "loss": 0.4878,
+      "reward": 2.3627763986587524,
+      "reward_std": 0.4748759865760803,
+      "rewards/accuracy_reward": 0.5000000223517418,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.029584777541458607,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1744
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 368.0833435058594,
+      "epoch": 0.8725,
+      "grad_norm": 4.887810570436975,
+      "kl": 1.171875,
+      "learning_rate": 1.438366220425628e-07,
+      "loss": 0.4411,
+      "reward": 2.735856294631958,
+      "reward_std": 0.47497308254241943,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.03671339526772499,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1745
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 356.5416717529297,
+      "epoch": 0.873,
+      "grad_norm": 5.959026182042601,
+      "kl": 1.115234375,
+      "learning_rate": 1.4349911821151462e-07,
+      "loss": 0.4461,
+      "reward": 2.7494075298309326,
+      "reward_std": 0.3780593601986766,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.03010638989508152,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1746
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 505.04168701171875,
+      "epoch": 0.8735,
+      "grad_norm": 9.461995383287107,
+      "kl": 2.03125,
+      "learning_rate": 1.4316285265264978e-07,
+      "loss": 0.6131,
+      "reward": 2.443665385246277,
+      "reward_std": 0.32284732908010483,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.021612409502267838,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1747
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 505.41668701171875,
+      "epoch": 0.874,
+      "grad_norm": 5.960798471551781,
+      "kl": 1.921875,
+      "learning_rate": 1.4282782639029128e-07,
+      "loss": 0.6131,
+      "reward": 2.305721640586853,
+      "reward_std": 0.7335332632064819,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.018931284546852112,
+      "rewards/tag_count_reward": 0.8802083730697632,
+      "step": 1748
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 230.1666717529297,
+      "epoch": 0.8745,
+      "grad_norm": 5.536492769825701,
+      "kl": 0.98046875,
+      "learning_rate": 1.4249404044498727e-07,
+      "loss": 0.1038,
+      "reward": 2.941461443901062,
+      "reward_std": 0.18287718016654253,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.008191618835553527,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1749
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 355.0208435058594,
+      "epoch": 0.875,
+      "grad_norm": 12.021308076552485,
+      "kl": 0.67578125,
+      "learning_rate": 1.4216149583350755e-07,
+      "loss": 0.2597,
+      "reward": 2.925000309944153,
+      "reward_std": 0.15377740375697613,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03506927099078894,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 1750
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 332.375,
+      "epoch": 0.8755,
+      "grad_norm": 18.87149191181025,
+      "kl": 0.92578125,
+      "learning_rate": 1.418301935688408e-07,
+      "loss": 0.2581,
+      "reward": 2.644270658493042,
+      "reward_std": 0.24090787768363953,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.022396131418645382,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1751
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 285.0416793823242,
+      "epoch": 0.876,
+      "grad_norm": 4.302085628117804,
+      "kl": 0.828125,
+      "learning_rate": 1.4150013466019114e-07,
+      "loss": 0.16,
+      "reward": 2.77993905544281,
+      "reward_std": 0.36863449215888977,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.015199816785752773,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1752
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 327.93751525878906,
+      "epoch": 0.8765,
+      "grad_norm": 10.151412864739319,
+      "kl": 1.16796875,
+      "learning_rate": 1.4117132011297528e-07,
+      "loss": 0.4048,
+      "reward": 2.7850375175476074,
+      "reward_std": 0.502088338136673,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.015309692360460758,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1753
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.8333435058594,
+      "epoch": 0.877,
+      "grad_norm": 6.5516024659716425,
+      "kl": 0.76953125,
+      "learning_rate": 1.4084375092881917e-07,
+      "loss": 0.294,
+      "reward": 2.7152035236358643,
+      "reward_std": 0.4315020889043808,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.02611603494733572,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1754
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 480.45835876464844,
+      "epoch": 0.8775,
+      "grad_norm": 11.39389155107235,
+      "kl": 1.8203125,
+      "learning_rate": 1.405174281055556e-07,
+      "loss": 0.4952,
+      "reward": 2.452518105506897,
+      "reward_std": 0.5552013963460922,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.011023662984371185,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1755
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 323.8333435058594,
+      "epoch": 0.878,
+      "grad_norm": 4.28402802913973,
+      "kl": 0.873046875,
+      "learning_rate": 1.4019235263722034e-07,
+      "loss": 0.1875,
+      "reward": 2.5597667694091797,
+      "reward_std": 0.40155889838933945,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.030511243268847466,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1756
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 483.0208435058594,
+      "epoch": 0.8785,
+      "grad_norm": 13.359830394312729,
+      "kl": 2.0078125,
+      "learning_rate": 1.3986852551404962e-07,
+      "loss": 0.499,
+      "reward": 2.397765874862671,
+      "reward_std": 0.4639376848936081,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.024109240621328354,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1757
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 434.4375,
+      "epoch": 0.879,
+      "grad_norm": 7.144462129204785,
+      "kl": 1.4609375,
+      "learning_rate": 1.395459477224772e-07,
+      "loss": 0.3825,
+      "reward": 2.5313947200775146,
+      "reward_std": 0.4858998954296112,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0311053404584527,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1758
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 309.25001525878906,
+      "epoch": 0.8795,
+      "grad_norm": 5.497608019794996,
+      "kl": 0.890625,
+      "learning_rate": 1.3922462024513075e-07,
+      "loss": 0.3099,
+      "reward": 2.6326241493225098,
+      "reward_std": 0.32827917486429214,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.009737157728523016,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1759
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 383.22918701171875,
+      "epoch": 0.88,
+      "grad_norm": 7.950064862238582,
+      "kl": 1.3828125,
+      "learning_rate": 1.3890454406082956e-07,
+      "loss": 0.4321,
+      "reward": 2.515586256980896,
+      "reward_std": 0.5082004070281982,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666567325592,
+      "rewards/repetition_penalty_reward": -0.020872057415544987,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1760
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 337.18751525878906,
+      "epoch": 0.8805,
+      "grad_norm": 4.154164825967078,
+      "kl": 0.740234375,
+      "learning_rate": 1.385857201445813e-07,
+      "loss": 0.2849,
+      "reward": 2.466183662414551,
+      "reward_std": 0.2627424318343401,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.030344081111252308,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1761
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 357.7916717529297,
+      "epoch": 0.881,
+      "grad_norm": 8.822719078728802,
+      "kl": 1.125,
+      "learning_rate": 1.3826814946757888e-07,
+      "loss": 0.1549,
+      "reward": 2.106475353240967,
+      "reward_std": 0.4267265051603317,
+      "rewards/accuracy_reward": 0.2083333358168602,
+      "rewards/reasoning_steps_reward": 0.9791666269302368,
+      "rewards/repetition_penalty_reward": -0.03935814555734396,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1762
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 341.6041717529297,
+      "epoch": 0.8815,
+      "grad_norm": 4.2275590882335266,
+      "kl": 0.904296875,
+      "learning_rate": 1.3795183299719753e-07,
+      "loss": 0.1513,
+      "reward": 2.4813464879989624,
+      "reward_std": 0.5057013630867004,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03601475339382887,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1763
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.9583435058594,
+      "epoch": 0.882,
+      "grad_norm": 5.592735227584482,
+      "kl": 0.982421875,
+      "learning_rate": 1.3763677169699217e-07,
+      "loss": -0.0045,
+      "reward": 2.6361879110336304,
+      "reward_std": 0.3717052489519119,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.021798397414386272,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1764
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 379.3541717529297,
+      "epoch": 0.8825,
+      "grad_norm": 4.766850312236698,
+      "kl": 1.126953125,
+      "learning_rate": 1.3732296652669417e-07,
+      "loss": 0.2914,
+      "reward": 2.6185107231140137,
+      "reward_std": 0.407680407166481,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03253097087144852,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1765
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 348.00001525878906,
+      "epoch": 0.883,
+      "grad_norm": 4.908003911372998,
+      "kl": 1.236328125,
+      "learning_rate": 1.370104184422085e-07,
+      "loss": 0.4774,
+      "reward": 2.7331286668777466,
+      "reward_std": 0.5909359902143478,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.03423246555030346,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1766
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 424.31251525878906,
+      "epoch": 0.8835,
+      "grad_norm": 4.218081542001781,
+      "kl": 1.5625,
+      "learning_rate": 1.3669912839561083e-07,
+      "loss": 0.4539,
+      "reward": 2.497738838195801,
+      "reward_std": 0.3815242201089859,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.028302965685725212,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1767
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 355.7708435058594,
+      "epoch": 0.884,
+      "grad_norm": 11.659886994958134,
+      "kl": 1.17578125,
+      "learning_rate": 1.3638909733514452e-07,
+      "loss": 0.3457,
+      "reward": 2.616908311843872,
+      "reward_std": 0.5863338112831116,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.021980691701173782,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1768
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 435.93751525878906,
+      "epoch": 0.8845,
+      "grad_norm": 10.949518936212161,
+      "kl": 1.16015625,
+      "learning_rate": 1.3608032620521803e-07,
+      "loss": 0.5473,
+      "reward": 2.750458240509033,
+      "reward_std": 0.5212399363517761,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.025583509355783463,
+      "rewards/tag_count_reward": 0.9427083730697632,
+      "step": 1769
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 354.81251525878906,
+      "epoch": 0.885,
+      "grad_norm": 10.36571400939137,
+      "kl": 1.19921875,
+      "learning_rate": 1.3577281594640182e-07,
+      "loss": 0.7841,
+      "reward": 2.636287212371826,
+      "reward_std": 0.4996710419654846,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.023435143288224936,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1770
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 452.375,
+      "epoch": 0.8855,
+      "grad_norm": 9.00198300414043,
+      "kl": 1.53125,
+      "learning_rate": 1.354665674954255e-07,
+      "loss": 0.4026,
+      "reward": 2.6022530794143677,
+      "reward_std": 0.33922192733734846,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.027955124154686928,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1771
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 364.68750762939453,
+      "epoch": 0.886,
+      "grad_norm": 7.279570069897522,
+      "kl": 0.958984375,
+      "learning_rate": 1.351615817851748e-07,
+      "loss": 0.297,
+      "reward": 2.481359362602234,
+      "reward_std": 0.33179375529289246,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.02732121106237173,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1772
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 390.12501525878906,
+      "epoch": 0.8865,
+      "grad_norm": 10.916967026611704,
+      "kl": 1.41015625,
+      "learning_rate": 1.3485785974468913e-07,
+      "loss": 0.3221,
+      "reward": 2.6607974767684937,
+      "reward_std": 0.38768453896045685,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.035383082926273346,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1773
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 440.12501525878906,
+      "epoch": 0.887,
+      "grad_norm": 12.121221337164327,
+      "kl": 2.30078125,
+      "learning_rate": 1.345554022991586e-07,
+      "loss": 0.5623,
+      "reward": 2.4204729795455933,
+      "reward_std": 0.6897162795066833,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.03265202045440674,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1774
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 318.7083435058594,
+      "epoch": 0.8875,
+      "grad_norm": 10.982971637870204,
+      "kl": 1.44140625,
+      "learning_rate": 1.3425421036992097e-07,
+      "loss": 0.3624,
+      "reward": 2.7605329751968384,
+      "reward_std": 0.5273626148700714,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.020717153325676918,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1775
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 468.8958435058594,
+      "epoch": 0.888,
+      "grad_norm": 10.54577580044071,
+      "kl": 1.63671875,
+      "learning_rate": 1.3395428487445914e-07,
+      "loss": 0.39,
+      "reward": 2.5657081604003906,
+      "reward_std": 0.4877399206161499,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0332502406090498,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1776
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.4791717529297,
+      "epoch": 0.8885,
+      "grad_norm": 5.822202062218948,
+      "kl": 0.548828125,
+      "learning_rate": 1.3365562672639807e-07,
+      "loss": 0.0764,
+      "reward": 2.9392874240875244,
+      "reward_std": 0.10572412749752402,
+      "rewards/accuracy_reward": 0.9583333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.019045927096158266,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1777
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 367.0208435058594,
+      "epoch": 0.889,
+      "grad_norm": 12.670512447704958,
+      "kl": 1.421875,
+      "learning_rate": 1.3335823683550237e-07,
+      "loss": 0.7225,
+      "reward": 2.640958547592163,
+      "reward_std": 0.7428161203861237,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.017027711495757103,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1778
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 403.1458435058594,
+      "epoch": 0.8895,
+      "grad_norm": 9.688022329481187,
+      "kl": 1.75,
+      "learning_rate": 1.3306211610767327e-07,
+      "loss": 0.8464,
+      "reward": 2.530915379524231,
+      "reward_std": 0.6522108912467957,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.021168189123272896,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1779
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 362.0208435058594,
+      "epoch": 0.89,
+      "grad_norm": 13.017237996007623,
+      "kl": 1.33984375,
+      "learning_rate": 1.3276726544494571e-07,
+      "loss": 0.4707,
+      "reward": 2.7710498571395874,
+      "reward_std": 0.40325474739074707,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.025825275108218193,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1780
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 389.7291717529297,
+      "epoch": 0.8905,
+      "grad_norm": 5.387878254985672,
+      "kl": 1.51171875,
+      "learning_rate": 1.3247368574548605e-07,
+      "loss": 0.4883,
+      "reward": 2.2095279693603516,
+      "reward_std": 0.547009214758873,
+      "rewards/accuracy_reward": 0.3750000223517418,
+      "rewards/reasoning_steps_reward": 0.9236110746860504,
+      "rewards/repetition_penalty_reward": -0.02137490874156356,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1781
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 348.8541717529297,
+      "epoch": 0.891,
+      "grad_norm": 4.93435481117121,
+      "kl": 1.052734375,
+      "learning_rate": 1.3218137790358892e-07,
+      "loss": 0.356,
+      "reward": 2.8138599395751953,
+      "reward_std": 0.295873099938035,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.029890276491642,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1782
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 503.6250305175781,
+      "epoch": 0.8915,
+      "grad_norm": 18.667124890654204,
+      "kl": 1.90625,
+      "learning_rate": 1.3189034280967474e-07,
+      "loss": 0.4897,
+      "reward": 2.317874312400818,
+      "reward_std": 0.3144010305404663,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9166667461395264,
+      "rewards/repetition_penalty_reward": -0.051917336881160736,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1783
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 509.16668701171875,
+      "epoch": 0.892,
+      "grad_norm": 137.39498749282393,
+      "kl": 6.7734375,
+      "learning_rate": 1.316005813502869e-07,
+      "loss": 0.6333,
+      "reward": 2.432840585708618,
+      "reward_std": 0.6739647388458252,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.025492852553725243,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1784
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 349.2291717529297,
+      "epoch": 0.8925,
+      "grad_norm": 5.858027527176804,
+      "kl": 1.3203125,
+      "learning_rate": 1.31312094408089e-07,
+      "loss": 0.3111,
+      "reward": 2.6231387853622437,
+      "reward_std": 0.44334521889686584,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.0192225007340312,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1785
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 402.58335876464844,
+      "epoch": 0.893,
+      "grad_norm": 4.541492575563727,
+      "kl": 1.345703125,
+      "learning_rate": 1.3102488286186234e-07,
+      "loss": 0.4852,
+      "reward": 2.777048349380493,
+      "reward_std": 0.41408103704452515,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.028507346287369728,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1786
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 483.72918701171875,
+      "epoch": 0.8935,
+      "grad_norm": 6.934337961180237,
+      "kl": 2.015625,
+      "learning_rate": 1.30738947586503e-07,
+      "loss": 0.6592,
+      "reward": 2.56930148601532,
+      "reward_std": 0.6575948297977448,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.031392961740493774,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1787
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 290.25,
+      "epoch": 0.894,
+      "grad_norm": 6.966975877164384,
+      "kl": 1.09765625,
+      "learning_rate": 1.3045428945301953e-07,
+      "loss": 0.3265,
+      "reward": 2.8180272579193115,
+      "reward_std": 0.34140200912952423,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.017042224761098623,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1788
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 304.2708435058594,
+      "epoch": 0.8945,
+      "grad_norm": 6.272772314297587,
+      "kl": 1.0859375,
+      "learning_rate": 1.3017090932852998e-07,
+      "loss": 0.1709,
+      "reward": 2.655430555343628,
+      "reward_std": 0.332093209028244,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.014708399772644043,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1789
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 458.9375305175781,
+      "epoch": 0.895,
+      "grad_norm": 8.12356036354293,
+      "kl": 1.83203125,
+      "learning_rate": 1.2988880807625927e-07,
+      "loss": 0.6271,
+      "reward": 2.6602718830108643,
+      "reward_std": 0.5047450065612793,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03070042561739683,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1790
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 418.4166717529297,
+      "epoch": 0.8955,
+      "grad_norm": 8.535870766876632,
+      "kl": 1.236328125,
+      "learning_rate": 1.2960798655553673e-07,
+      "loss": 0.3997,
+      "reward": 2.491165280342102,
+      "reward_std": 0.27948543429374695,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9097222089767456,
+      "rewards/repetition_penalty_reward": -0.017515364568680525,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1791
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 321.7708435058594,
+      "epoch": 0.896,
+      "grad_norm": 6.5441149884787,
+      "kl": 1.076171875,
+      "learning_rate": 1.2932844562179352e-07,
+      "loss": 0.3451,
+      "reward": 2.7869484424591064,
+      "reward_std": 0.531157523393631,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.023815508000552654,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1792
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 296.8958435058594,
+      "epoch": 0.8965,
+      "grad_norm": 8.808586232096019,
+      "kl": 0.890625,
+      "learning_rate": 1.2905018612655974e-07,
+      "loss": 0.0777,
+      "reward": 2.684244394302368,
+      "reward_std": 0.2914978265762329,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.029297261498868465,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1793
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 589.0208435058594,
+      "epoch": 0.897,
+      "grad_norm": 5.648691995344081,
+      "kl": 2.07421875,
+      "learning_rate": 1.2877320891746201e-07,
+      "loss": 0.7666,
+      "reward": 2.3528552055358887,
+      "reward_std": 0.4950469881296158,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.0290891882032156,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1794
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 479.2708435058594,
+      "epoch": 0.8975,
+      "grad_norm": 5.153259121443274,
+      "kl": 1.52734375,
+      "learning_rate": 1.284975148382211e-07,
+      "loss": 0.4386,
+      "reward": 2.5888208150863647,
+      "reward_std": 0.37291720509529114,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.036179195158183575,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1795
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 330.6041717529297,
+      "epoch": 0.898,
+      "grad_norm": 9.368121569370517,
+      "kl": 1.115234375,
+      "learning_rate": 1.2822310472864885e-07,
+      "loss": 0.2721,
+      "reward": 2.7130978107452393,
+      "reward_std": 0.25604604184627533,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.023013423196971416,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1796
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 347.7083435058594,
+      "epoch": 0.8985,
+      "grad_norm": 12.211005971271671,
+      "kl": 1.15234375,
+      "learning_rate": 1.2794997942464603e-07,
+      "loss": 0.3333,
+      "reward": 2.739255905151367,
+      "reward_std": 0.45043814182281494,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.026369189843535423,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1797
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 422.35418701171875,
+      "epoch": 0.899,
+      "grad_norm": 10.548113579660656,
+      "kl": 2.203125,
+      "learning_rate": 1.2767813975819983e-07,
+      "loss": 0.3646,
+      "reward": 2.240885376930237,
+      "reward_std": 0.4887467324733734,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.8888889253139496,
+      "rewards/repetition_penalty_reward": -0.02300364524126053,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1798
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 417.1666717529297,
+      "epoch": 0.8995,
+      "grad_norm": 4.435343161658372,
+      "kl": 1.275390625,
+      "learning_rate": 1.274075865573809e-07,
+      "loss": 0.3645,
+      "reward": 2.541459321975708,
+      "reward_std": 0.3853719085454941,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.03666570037603378,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1799
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 572.6666870117188,
+      "epoch": 0.9,
+      "grad_norm": 12.274646361156972,
+      "kl": 2.36328125,
+      "learning_rate": 1.2713832064634125e-07,
+      "loss": 0.5502,
+      "reward": 2.4783600568771362,
+      "reward_std": 0.6872712820768356,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.014695549616590142,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1800
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 435.7083435058594,
+      "epoch": 0.9005,
+      "grad_norm": 6.174999936072762,
+      "kl": 1.251953125,
+      "learning_rate": 1.2687034284531145e-07,
+      "loss": 0.219,
+      "reward": 2.5370699167251587,
+      "reward_std": 0.4170294851064682,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.027166323270648718,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1801
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 423.16668701171875,
+      "epoch": 0.901,
+      "grad_norm": 8.371287383648244,
+      "kl": 1.984375,
+      "learning_rate": 1.2660365397059856e-07,
+      "loss": 0.6964,
+      "reward": 2.5543148517608643,
+      "reward_std": 0.6248021870851517,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.02728235349059105,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1802
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 443.08335876464844,
+      "epoch": 0.9015,
+      "grad_norm": 4.999316997210919,
+      "kl": 1.60546875,
+      "learning_rate": 1.263382548345829e-07,
+      "loss": 0.4042,
+      "reward": 2.5936447381973267,
+      "reward_std": 0.49106191098690033,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0313554760068655,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1803
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 438.00001525878906,
+      "epoch": 0.902,
+      "grad_norm": 13.183151595537693,
+      "kl": 2.0703125,
+      "learning_rate": 1.260741462457165e-07,
+      "loss": 0.3495,
+      "reward": 2.561880946159363,
+      "reward_std": 0.6570396423339844,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.011035696603357792,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1804
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 480.4375,
+      "epoch": 0.9025,
+      "grad_norm": 7.229161374154403,
+      "kl": 2.3125,
+      "learning_rate": 1.258113290085197e-07,
+      "loss": 0.405,
+      "reward": 2.4197758436203003,
+      "reward_std": 0.7334087789058685,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.03161316737532616,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1805
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 374.2291717529297,
+      "epoch": 0.903,
+      "grad_norm": 8.888192201627456,
+      "kl": 1.01171875,
+      "learning_rate": 1.2554980392357956e-07,
+      "loss": 0.5509,
+      "reward": 2.8383933305740356,
+      "reward_std": 0.3369765877723694,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.027926414273679256,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1806
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.31251525878906,
+      "epoch": 0.9035,
+      "grad_norm": 7.34813847799913,
+      "kl": 0.875,
+      "learning_rate": 1.2528957178754676e-07,
+      "loss": 0.1249,
+      "reward": 2.66613507270813,
+      "reward_std": 0.42864419519901276,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.026573394425213337,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1807
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.7291717529297,
+      "epoch": 0.904,
+      "grad_norm": 6.090013410671144,
+      "kl": 0.939453125,
+      "learning_rate": 1.2503063339313356e-07,
+      "loss": 0.4151,
+      "reward": 2.8806700706481934,
+      "reward_std": 0.2488960325717926,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.025580264627933502,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1808
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 430.2083435058594,
+      "epoch": 0.9045,
+      "grad_norm": 13.573272613458863,
+      "kl": 1.37890625,
+      "learning_rate": 1.2477298952911116e-07,
+      "loss": 0.5799,
+      "reward": 2.1159850358963013,
+      "reward_std": 0.4026012271642685,
+      "rewards/accuracy_reward": 0.2291666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.014223407953977585,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1809
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 404.0833435058594,
+      "epoch": 0.905,
+      "grad_norm": 8.300631259188407,
+      "kl": 2.8203125,
+      "learning_rate": 1.2451664098030743e-07,
+      "loss": 0.5028,
+      "reward": 2.1978728771209717,
+      "reward_std": 0.8104668259620667,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.8333333432674408,
+      "rewards/repetition_penalty_reward": -0.026085459627211094,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1810
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.7291717529297,
+      "epoch": 0.9055,
+      "grad_norm": 16.892046533776394,
+      "kl": 1.1953125,
+      "learning_rate": 1.242615885276046e-07,
+      "loss": 0.5455,
+      "reward": 2.822051763534546,
+      "reward_std": 0.5525839030742645,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.01822614297270775,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1811
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 369.9583435058594,
+      "epoch": 0.906,
+      "grad_norm": 6.741532569071179,
+      "kl": 0.923828125,
+      "learning_rate": 1.2400783294793668e-07,
+      "loss": 0.3853,
+      "reward": 2.762416362762451,
+      "reward_std": 0.38367322087287903,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.025778336450457573,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1812
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 438.8333435058594,
+      "epoch": 0.9065,
+      "grad_norm": 7.797105914296452,
+      "kl": 1.375,
+      "learning_rate": 1.2375537501428706e-07,
+      "loss": 0.3492,
+      "reward": 2.350590944290161,
+      "reward_std": 0.4562319219112396,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.015728731639683247,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1813
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 515.3958435058594,
+      "epoch": 0.907,
+      "grad_norm": 16.218331905338545,
+      "kl": 1.94921875,
+      "learning_rate": 1.235042154956865e-07,
+      "loss": 0.4092,
+      "reward": 2.4498504400253296,
+      "reward_std": 0.2465880587697029,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.024108044803142548,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1814
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 463.27085876464844,
+      "epoch": 0.9075,
+      "grad_norm": 4.573817188249182,
+      "kl": 1.50390625,
+      "learning_rate": 1.232543551572103e-07,
+      "loss": 0.6756,
+      "reward": 2.6501909494400024,
+      "reward_std": 0.4004078805446625,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.025156395509839058,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1815
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 484.3333435058594,
+      "epoch": 0.908,
+      "grad_norm": 21.11225799549021,
+      "kl": 2.875,
+      "learning_rate": 1.2300579475997657e-07,
+      "loss": 0.6141,
+      "reward": 2.0554409623146057,
+      "reward_std": 0.6282560527324677,
+      "rewards/accuracy_reward": 0.354166679084301,
+      "rewards/reasoning_steps_reward": 0.854166716337204,
+      "rewards/repetition_penalty_reward": -0.022684063762426376,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1816
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 390.3958435058594,
+      "epoch": 0.9085,
+      "grad_norm": 4.270688484326959,
+      "kl": 1.458984375,
+      "learning_rate": 1.227585350611433e-07,
+      "loss": 0.4887,
+      "reward": 2.5446548461914062,
+      "reward_std": 0.4144492670893669,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.026525692082941532,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1817
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 510.0833435058594,
+      "epoch": 0.909,
+      "grad_norm": 10.195671953035172,
+      "kl": 1.78125,
+      "learning_rate": 1.2251257681390645e-07,
+      "loss": 0.5759,
+      "reward": 2.590057134628296,
+      "reward_std": 0.579290121793747,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02799857221543789,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1818
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 365.5208435058594,
+      "epoch": 0.9095,
+      "grad_norm": 14.858149785129092,
+      "kl": 1.1015625,
+      "learning_rate": 1.2226792076749734e-07,
+      "loss": 0.3952,
+      "reward": 2.7983046770095825,
+      "reward_std": 0.5069041550159454,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.031556460075080395,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1819
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 498.27085876464844,
+      "epoch": 0.91,
+      "grad_norm": 13.474699117275431,
+      "kl": 2.5390625,
+      "learning_rate": 1.220245676671809e-07,
+      "loss": 0.6901,
+      "reward": 2.2409090995788574,
+      "reward_std": 0.5821706056594849,
+      "rewards/accuracy_reward": 0.3958333358168602,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.01777158584445715,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1820
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 321.2083435058594,
+      "epoch": 0.9105,
+      "grad_norm": 5.583742600416253,
+      "kl": 1.087890625,
+      "learning_rate": 1.2178251825425282e-07,
+      "loss": 0.3231,
+      "reward": 2.433494746685028,
+      "reward_std": 0.37168148159980774,
+      "rewards/accuracy_reward": 0.5208333544433117,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.01963021163828671,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1821
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 437.35418701171875,
+      "epoch": 0.911,
+      "grad_norm": 15.808710038962724,
+      "kl": 1.85546875,
+      "learning_rate": 1.2154177326603763e-07,
+      "loss": 0.3345,
+      "reward": 1.9915515780448914,
+      "reward_std": 0.4922345131635666,
+      "rewards/accuracy_reward": 0.1666666716337204,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.017128958366811275,
+      "rewards/tag_count_reward": 0.9114583730697632,
+      "step": 1822
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 461.66668701171875,
+      "epoch": 0.9115,
+      "grad_norm": 4.487639145726479,
+      "kl": 1.5859375,
+      "learning_rate": 1.2130233343588623e-07,
+      "loss": 0.6766,
+      "reward": 2.4462087154388428,
+      "reward_std": 0.48641470074653625,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.013860756065696478,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1823
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 427.5208435058594,
+      "epoch": 0.912,
+      "grad_norm": 5.206907795171928,
+      "kl": 1.11328125,
+      "learning_rate": 1.2106419949317388e-07,
+      "loss": 0.298,
+      "reward": 2.293039083480835,
+      "reward_std": 0.4323887377977371,
+      "rewards/accuracy_reward": 0.3958333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222983837128,
+      "rewards/repetition_penalty_reward": -0.022933254949748516,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1824
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 455.7916717529297,
+      "epoch": 0.9125,
+      "grad_norm": 3.5810924881234034,
+      "kl": 1.447265625,
+      "learning_rate": 1.2082737216329792e-07,
+      "loss": 0.4868,
+      "reward": 2.722281336784363,
+      "reward_std": 0.5233409157954156,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.02424659300595522,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1825
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 366.6875,
+      "epoch": 0.913,
+      "grad_norm": 10.549190317425618,
+      "kl": 1.34765625,
+      "learning_rate": 1.2059185216767543e-07,
+      "loss": 0.6111,
+      "reward": 2.620617389678955,
+      "reward_std": 0.4774337261915207,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.018271582201123238,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1826
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.50001525878906,
+      "epoch": 0.9135,
+      "grad_norm": 8.459023548215319,
+      "kl": 0.80859375,
+      "learning_rate": 1.203576402237412e-07,
+      "loss": 0.186,
+      "reward": 2.81669545173645,
+      "reward_std": 0.3177480548620224,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.046151867136359215,
+      "rewards/tag_count_reward": 0.9947916865348816,
+      "step": 1827
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 435.4791717529297,
+      "epoch": 0.914,
+      "grad_norm": 10.059114870695714,
+      "kl": 1.42578125,
+      "learning_rate": 1.2012473704494537e-07,
+      "loss": 0.8145,
+      "reward": 2.4705541133880615,
+      "reward_std": 0.6846470832824707,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.024237760342657566,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1828
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 396.3541717529297,
+      "epoch": 0.9145,
+      "grad_norm": 7.729647467876799,
+      "kl": 1.3515625,
+      "learning_rate": 1.1989314334075144e-07,
+      "loss": 0.3771,
+      "reward": 2.610179305076599,
+      "reward_std": 0.4845607876777649,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.025237280875444412,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1829
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 404.16668701171875,
+      "epoch": 0.915,
+      "grad_norm": 6.446666593330831,
+      "kl": 1.466796875,
+      "learning_rate": 1.1966285981663407e-07,
+      "loss": 0.735,
+      "reward": 2.597218871116638,
+      "reward_std": 0.3849441111087799,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.01562830712646246,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1830
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 400.66668701171875,
+      "epoch": 0.9155,
+      "grad_norm": 7.699253526367492,
+      "kl": 1.23046875,
+      "learning_rate": 1.1943388717407668e-07,
+      "loss": 0.5849,
+      "reward": 2.241459369659424,
+      "reward_std": 0.3354404419660568,
+      "rewards/accuracy_reward": 0.35416667722165585,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.02069345023483038,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1831
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 341.31251525878906,
+      "epoch": 0.916,
+      "grad_norm": 3.9150353521261585,
+      "kl": 1.09375,
+      "learning_rate": 1.1920622611056974e-07,
+      "loss": 0.5796,
+      "reward": 2.800022602081299,
+      "reward_std": 0.3668653219938278,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.02810249850153923,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1832
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 333.31251525878906,
+      "epoch": 0.9165,
+      "grad_norm": 6.331026738180072,
+      "kl": 1.453125,
+      "learning_rate": 1.1897987731960835e-07,
+      "loss": 0.5084,
+      "reward": 2.8525108098983765,
+      "reward_std": 0.3399234637618065,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.01901690987870097,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1833
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 427.2708435058594,
+      "epoch": 0.917,
+      "grad_norm": 7.909012344076658,
+      "kl": 1.515625,
+      "learning_rate": 1.1875484149069004e-07,
+      "loss": 0.9793,
+      "reward": 2.418446898460388,
+      "reward_std": 0.4566657245159149,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.013844884466379881,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1834
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 326.4375,
+      "epoch": 0.9175,
+      "grad_norm": 15.090270389095783,
+      "kl": 1.81640625,
+      "learning_rate": 1.1853111930931312e-07,
+      "loss": 0.3326,
+      "reward": 2.7400245666503906,
+      "reward_std": 0.5180048495531082,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.01692010648548603,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1835
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.33333587646484,
+      "epoch": 0.918,
+      "grad_norm": 14.385789840879672,
+      "kl": 1.265625,
+      "learning_rate": 1.1830871145697412e-07,
+      "loss": 0.276,
+      "reward": 2.505857825279236,
+      "reward_std": 0.44202379882335663,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.018447946291416883,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1836
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 474.4583435058594,
+      "epoch": 0.9185,
+      "grad_norm": 22.396059856717127,
+      "kl": 1.96484375,
+      "learning_rate": 1.1808761861116589e-07,
+      "loss": 0.9942,
+      "reward": 2.2753371000289917,
+      "reward_std": 0.6203365921974182,
+      "rewards/accuracy_reward": 0.4375000149011612,
+      "rewards/reasoning_steps_reward": 0.9374999701976776,
+      "rewards/repetition_penalty_reward": -0.016329674050211906,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1837
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 442.62501525878906,
+      "epoch": 0.919,
+      "grad_norm": 15.472496124027192,
+      "kl": 1.697265625,
+      "learning_rate": 1.1786784144537563e-07,
+      "loss": 0.6291,
+      "reward": 2.5869717597961426,
+      "reward_std": 0.6588756740093231,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9444444179534912,
+      "rewards/repetition_penalty_reward": -0.034556107595562935,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1838
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 360.6666717529297,
+      "epoch": 0.9195,
+      "grad_norm": 6.6879952807989715,
+      "kl": 1.015625,
+      "learning_rate": 1.1764938062908261e-07,
+      "loss": 0.3457,
+      "reward": 2.6010701656341553,
+      "reward_std": 0.44879356026649475,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02740222029387951,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1839
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 565.2083587646484,
+      "epoch": 0.92,
+      "grad_norm": 14.940386937141579,
+      "kl": 2.140625,
+      "learning_rate": 1.1743223682775649e-07,
+      "loss": 0.7608,
+      "reward": 2.381098985671997,
+      "reward_std": 0.6445316672325134,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9236111640930176,
+      "rewards/repetition_penalty_reward": -0.021678834222257137,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1840
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 427.81251525878906,
+      "epoch": 0.9205,
+      "grad_norm": 7.0097527574764635,
+      "kl": 1.4765625,
+      "learning_rate": 1.172164107028549e-07,
+      "loss": 0.5432,
+      "reward": 2.608567476272583,
+      "reward_std": 0.7569788694381714,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.019904857501387596,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1841
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 314.10418701171875,
+      "epoch": 0.921,
+      "grad_norm": 5.8284852901692314,
+      "kl": 0.796875,
+      "learning_rate": 1.1700190291182158e-07,
+      "loss": 0.2706,
+      "reward": 2.492654800415039,
+      "reward_std": 0.44027116894721985,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.028178581036627293,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1842
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 383.10418701171875,
+      "epoch": 0.9215,
+      "grad_norm": 11.442490301965112,
+      "kl": 0.736328125,
+      "learning_rate": 1.1678871410808454e-07,
+      "loss": 0.2935,
+      "reward": 2.5915920734405518,
+      "reward_std": 0.4983036518096924,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03340807091444731,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1843
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 519.125,
+      "epoch": 0.922,
+      "grad_norm": 14.031885386954286,
+      "kl": 1.98046875,
+      "learning_rate": 1.1657684494105386e-07,
+      "loss": 0.2441,
+      "reward": 2.374902606010437,
+      "reward_std": 0.2330544777214527,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.019194713328033686,
+      "rewards/tag_count_reward": 0.8802083432674408,
+      "step": 1844
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 438.875,
+      "epoch": 0.9225,
+      "grad_norm": 5.420069805796557,
+      "kl": 1.453125,
+      "learning_rate": 1.1636629605611966e-07,
+      "loss": 0.7929,
+      "reward": 2.498497247695923,
+      "reward_std": 0.4927578568458557,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.013655558926984668,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1845
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 339.35418701171875,
+      "epoch": 0.923,
+      "grad_norm": 12.990900222582072,
+      "kl": 1.109375,
+      "learning_rate": 1.1615706809465051e-07,
+      "loss": 0.4619,
+      "reward": 2.6084574460983276,
+      "reward_std": 0.43079179525375366,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.02175095770508051,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1846
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 318.1041717529297,
+      "epoch": 0.9235,
+      "grad_norm": 3.6306004992091196,
+      "kl": 0.671875,
+      "learning_rate": 1.1594916169399087e-07,
+      "loss": 0.0754,
+      "reward": 2.619461178779602,
+      "reward_std": 0.2529575452208519,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.01942774746567011,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1847
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.2083435058594,
+      "epoch": 0.924,
+      "grad_norm": 12.262592006970328,
+      "kl": 1.064453125,
+      "learning_rate": 1.1574257748745986e-07,
+      "loss": 0.0744,
+      "reward": 2.376179814338684,
+      "reward_std": 0.5322179198265076,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.028334129601716995,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1848
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 389.12501525878906,
+      "epoch": 0.9245,
+      "grad_norm": 4.963783636537933,
+      "kl": 1.248046875,
+      "learning_rate": 1.1553731610434876e-07,
+      "loss": 0.308,
+      "reward": 2.5988192558288574,
+      "reward_std": 0.4921792149543762,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.031389085575938225,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1849
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 416.8125,
+      "epoch": 0.925,
+      "grad_norm": 11.2545524933753,
+      "kl": 1.724609375,
+      "learning_rate": 1.1533337816991931e-07,
+      "loss": 0.1345,
+      "reward": 2.312873065471649,
+      "reward_std": 0.45963311195373535,
+      "rewards/accuracy_reward": 0.5000000223517418,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.018724264577031136,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1850
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 400.43751525878906,
+      "epoch": 0.9255,
+      "grad_norm": 6.837253831426617,
+      "kl": 1.1015625,
+      "learning_rate": 1.1513076430540177e-07,
+      "loss": 0.6187,
+      "reward": 2.7513691186904907,
+      "reward_std": 0.4550359845161438,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.035089388489723206,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1851
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 349.2708435058594,
+      "epoch": 0.926,
+      "grad_norm": 11.15586838979647,
+      "kl": 0.6845703125,
+      "learning_rate": 1.1492947512799328e-07,
+      "loss": 0.3386,
+      "reward": 2.8566954135894775,
+      "reward_std": 0.34863437712192535,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.02872132882475853,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1852
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.2291717529297,
+      "epoch": 0.9265,
+      "grad_norm": 7.058419371924635,
+      "kl": 1.044921875,
+      "learning_rate": 1.1472951125085547e-07,
+      "loss": 0.2017,
+      "reward": 2.3934552669525146,
+      "reward_std": 0.22491255030035973,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02668381668627262,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1853
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 335.3958435058594,
+      "epoch": 0.927,
+      "grad_norm": 8.473916574225584,
+      "kl": 1.1015625,
+      "learning_rate": 1.1453087328311299e-07,
+      "loss": 0.3028,
+      "reward": 2.344975471496582,
+      "reward_std": 0.41645242273807526,
+      "rewards/accuracy_reward": 0.4583333544433117,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.026552507653832436,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1854
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 267.3541717529297,
+      "epoch": 0.9275,
+      "grad_norm": 4.44552019323035,
+      "kl": 0.564453125,
+      "learning_rate": 1.1433356182985158e-07,
+      "loss": 0.0992,
+      "reward": 2.769820213317871,
+      "reward_std": 0.2516600340604782,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02184649184346199,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1855
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 496.75,
+      "epoch": 0.928,
+      "grad_norm": 6.704669428410124,
+      "kl": 1.333984375,
+      "learning_rate": 1.1413757749211602e-07,
+      "loss": 0.494,
+      "reward": 2.636578679084778,
+      "reward_std": 0.5088631808757782,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.03008817508816719,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1856
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 361.9166793823242,
+      "epoch": 0.9285,
+      "grad_norm": 5.283796075893927,
+      "kl": 1.150390625,
+      "learning_rate": 1.1394292086690874e-07,
+      "loss": 0.5762,
+      "reward": 2.6141971349716187,
+      "reward_std": 0.36808090656995773,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.019483450800180435,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1857
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.7708435058594,
+      "epoch": 0.929,
+      "grad_norm": 13.090857743748684,
+      "kl": 0.96875,
+      "learning_rate": 1.137495925471875e-07,
+      "loss": 0.2827,
+      "reward": 2.789952039718628,
+      "reward_std": 0.39836449921131134,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.019075598567724228,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1858
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 289.7708435058594,
+      "epoch": 0.9295,
+      "grad_norm": 12.791650212488964,
+      "kl": 1.384765625,
+      "learning_rate": 1.1355759312186396e-07,
+      "loss": 0.0501,
+      "reward": 2.5625548362731934,
+      "reward_std": 0.6163710951805115,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.027723138220608234,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1859
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 343.18751525878906,
+      "epoch": 0.93,
+      "grad_norm": 3.6089686265116567,
+      "kl": 1.65234375,
+      "learning_rate": 1.1336692317580158e-07,
+      "loss": 0.4677,
+      "reward": 2.5494874715805054,
+      "reward_std": 0.5432776808738708,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.013012669514864683,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1860
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 363.47918701171875,
+      "epoch": 0.9305,
+      "grad_norm": 4.537501307347617,
+      "kl": 1.193359375,
+      "learning_rate": 1.1317758328981414e-07,
+      "loss": 0.4101,
+      "reward": 2.4980164766311646,
+      "reward_std": 0.415087066590786,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.01760861463844776,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1861
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 552.3541870117188,
+      "epoch": 0.931,
+      "grad_norm": 26.096314917799425,
+      "kl": 3.484375,
+      "learning_rate": 1.1298957404066381e-07,
+      "loss": 0.5845,
+      "reward": 2.1797789335250854,
+      "reward_std": 0.6568257510662079,
+      "rewards/accuracy_reward": 0.5833333730697632,
+      "rewards/reasoning_steps_reward": 0.7986111044883728,
+      "rewards/repetition_penalty_reward": -0.014665620867162943,
+      "rewards/tag_count_reward": 0.8125,
+      "step": 1862
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 424.1875,
+      "epoch": 0.9315,
+      "grad_norm": 8.73387409565459,
+      "kl": 1.814453125,
+      "learning_rate": 1.1280289600105928e-07,
+      "loss": 0.415,
+      "reward": 2.678765296936035,
+      "reward_std": 0.4467965252697468,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.03130429983139038,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1863
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 415.8750305175781,
+      "epoch": 0.932,
+      "grad_norm": 8.913281266427834,
+      "kl": 1.458984375,
+      "learning_rate": 1.1261754973965422e-07,
+      "loss": 0.2109,
+      "reward": 2.7937296628952026,
+      "reward_std": 0.3128529414534569,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.02918718010187149,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1864
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 433.8333435058594,
+      "epoch": 0.9325,
+      "grad_norm": 6.364728489318106,
+      "kl": 1.65625,
+      "learning_rate": 1.1243353582104555e-07,
+      "loss": 0.4969,
+      "reward": 2.676972270011902,
+      "reward_std": 0.4657895863056183,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.029625079594552517,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1865
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 387.31251525878906,
+      "epoch": 0.933,
+      "grad_norm": 13.019299595053573,
+      "kl": 1.962890625,
+      "learning_rate": 1.1225085480577158e-07,
+      "loss": 0.4171,
+      "reward": 2.349880337715149,
+      "reward_std": 0.5604653209447861,
+      "rewards/accuracy_reward": 0.5625000298023224,
+      "rewards/reasoning_steps_reward": 0.881944477558136,
+      "rewards/repetition_penalty_reward": -0.021647500805556774,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1866
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 293.06251525878906,
+      "epoch": 0.9335,
+      "grad_norm": 8.091047902175246,
+      "kl": 0.810546875,
+      "learning_rate": 1.1206950725031034e-07,
+      "loss": 0.2837,
+      "reward": 2.5777370929718018,
+      "reward_std": 0.35544297099113464,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.022957434877753258,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 1867
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.00000762939453,
+      "epoch": 0.934,
+      "grad_norm": 6.992390350124804,
+      "kl": 0.6953125,
+      "learning_rate": 1.1188949370707787e-07,
+      "loss": 0.0188,
+      "reward": 2.7592689990997314,
+      "reward_std": 0.17366931587457657,
+      "rewards/accuracy_reward": 0.7708333730697632,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.011564383283257484,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1868
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 571.7708435058594,
+      "epoch": 0.9345,
+      "grad_norm": 21.933887010235615,
+      "kl": 2.62109375,
+      "learning_rate": 1.117108147244268e-07,
+      "loss": 1.082,
+      "reward": 2.3994613885879517,
+      "reward_std": 0.9821091592311859,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.8333333432674408,
+      "rewards/repetition_penalty_reward": -0.01720538828521967,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1869
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 377.81251525878906,
+      "epoch": 0.935,
+      "grad_norm": 12.81237348732685,
+      "kl": 1.54296875,
+      "learning_rate": 1.1153347084664419e-07,
+      "loss": 0.5125,
+      "reward": 2.731147050857544,
+      "reward_std": 0.4906059801578522,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.02058906713500619,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1870
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 328.7916717529297,
+      "epoch": 0.9355,
+      "grad_norm": 13.632759689287054,
+      "kl": 1.46875,
+      "learning_rate": 1.1135746261395021e-07,
+      "loss": 0.456,
+      "reward": 2.5591378211975098,
+      "reward_std": 0.518625944852829,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.0224593966268003,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1871
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 470.2708435058594,
+      "epoch": 0.936,
+      "grad_norm": 7.995683940604791,
+      "kl": 1.75390625,
+      "learning_rate": 1.1118279056249653e-07,
+      "loss": 0.5945,
+      "reward": 2.374450922012329,
+      "reward_std": 0.7306532114744186,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.028326891362667084,
+      "rewards/tag_count_reward": 0.8958333730697632,
+      "step": 1872
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 379.47918701171875,
+      "epoch": 0.9365,
+      "grad_norm": 8.269844718099108,
+      "kl": 1.5546875,
+      "learning_rate": 1.1100945522436453e-07,
+      "loss": 0.5047,
+      "reward": 2.2810773849487305,
+      "reward_std": 0.45627573132514954,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333134651184,
+      "rewards/repetition_penalty_reward": -0.021005945280194283,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1873
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 247.1041717529297,
+      "epoch": 0.937,
+      "grad_norm": 4.879841129317827,
+      "kl": 0.658203125,
+      "learning_rate": 1.1083745712756364e-07,
+      "loss": 0.0382,
+      "reward": 2.8617427349090576,
+      "reward_std": 0.1286549223586917,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.013257297687232494,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1874
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 315.62501525878906,
+      "epoch": 0.9375,
+      "grad_norm": 5.772943696708483,
+      "kl": 1.052734375,
+      "learning_rate": 1.1066679679602998e-07,
+      "loss": 0.1094,
+      "reward": 2.2595136165618896,
+      "reward_std": 0.6155846416950226,
+      "rewards/accuracy_reward": 0.375,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.02000036695972085,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1875
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 562.0625152587891,
+      "epoch": 0.938,
+      "grad_norm": 12.685880213504184,
+      "kl": 2.75,
+      "learning_rate": 1.1049747474962444e-07,
+      "loss": 0.3528,
+      "reward": 2.470315456390381,
+      "reward_std": 0.3357051908969879,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.01926790364086628,
+      "rewards/tag_count_reward": 0.8645833730697632,
+      "step": 1876
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 399.00001525878906,
+      "epoch": 0.9385,
+      "grad_norm": 10.487921547989083,
+      "kl": 1.80078125,
+      "learning_rate": 1.1032949150413137e-07,
+      "loss": 0.2994,
+      "reward": 2.3582355976104736,
+      "reward_std": 0.48615631461143494,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.025445050559937954,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1877
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.2708435058594,
+      "epoch": 0.939,
+      "grad_norm": 13.345702679349012,
+      "kl": 0.615234375,
+      "learning_rate": 1.1016284757125685e-07,
+      "loss": 0.0566,
+      "reward": 2.74407958984375,
+      "reward_std": 0.42066872119903564,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.023281564470380545,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1878
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 381.2291717529297,
+      "epoch": 0.9395,
+      "grad_norm": 5.3082591079335915,
+      "kl": 1.546875,
+      "learning_rate": 1.099975434586272e-07,
+      "loss": 0.4069,
+      "reward": 2.447960615158081,
+      "reward_std": 0.7231379747390747,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.022525550797581673,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1879
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 306.3541717529297,
+      "epoch": 0.94,
+      "grad_norm": 6.836011499419833,
+      "kl": 1.158203125,
+      "learning_rate": 1.0983357966978745e-07,
+      "loss": 0.2872,
+      "reward": 2.6273016929626465,
+      "reward_std": 0.5883876979351044,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.013323335442692041,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1880
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 385.75001525878906,
+      "epoch": 0.9405,
+      "grad_norm": 5.637240680811261,
+      "kl": 1.46875,
+      "learning_rate": 1.096709567041997e-07,
+      "loss": 0.3754,
+      "reward": 2.4940316677093506,
+      "reward_std": 0.5485326498746872,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.02332939486950636,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1881
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 521.2083587646484,
+      "epoch": 0.941,
+      "grad_norm": 12.973036573210113,
+      "kl": 2.0078125,
+      "learning_rate": 1.0950967505724175e-07,
+      "loss": 0.3535,
+      "reward": 2.4711567163467407,
+      "reward_std": 0.4558318704366684,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.018426863476634026,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1882
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 384.2291717529297,
+      "epoch": 0.9415,
+      "grad_norm": 13.634252652179764,
+      "kl": 1.35546875,
+      "learning_rate": 1.0934973522020537e-07,
+      "loss": 0.5317,
+      "reward": 2.6534253358840942,
+      "reward_std": 0.546121746301651,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.020185869187116623,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1883
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 415.3541717529297,
+      "epoch": 0.942,
+      "grad_norm": 7.550722470094486,
+      "kl": 1.59375,
+      "learning_rate": 1.0919113768029517e-07,
+      "loss": 0.5054,
+      "reward": 2.2631616592407227,
+      "reward_std": 0.5261791199445724,
+      "rewards/accuracy_reward": 0.416666679084301,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.021560687571763992,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1884
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 407.9375,
+      "epoch": 0.9425,
+      "grad_norm": 5.164750350944079,
+      "kl": 1.28515625,
+      "learning_rate": 1.0903388292062668e-07,
+      "loss": 0.6485,
+      "reward": 2.5368224382400513,
+      "reward_std": 0.6539618074893951,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.016997200436890125,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1885
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 282.2916717529297,
+      "epoch": 0.943,
+      "grad_norm": 3.8431346982306622,
+      "kl": 0.546875,
+      "learning_rate": 1.0887797142022521e-07,
+      "loss": 0.0839,
+      "reward": 2.8730608224868774,
+      "reward_std": 0.1409488208591938,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.02277265675365925,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1886
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 459.41668701171875,
+      "epoch": 0.9435,
+      "grad_norm": 10.504402038451365,
+      "kl": 1.92578125,
+      "learning_rate": 1.0872340365402415e-07,
+      "loss": 0.7645,
+      "reward": 2.4082196950912476,
+      "reward_std": 0.7337851822376251,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.902777761220932,
+      "rewards/repetition_penalty_reward": -0.020599967800080776,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1887
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 287.29168701171875,
+      "epoch": 0.944,
+      "grad_norm": 12.459861460775484,
+      "kl": 1.25,
+      "learning_rate": 1.0857018009286381e-07,
+      "loss": 0.1306,
+      "reward": 2.6860339641571045,
+      "reward_std": 0.3863665908575058,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.029244041070342064,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1888
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 351.0833435058594,
+      "epoch": 0.9445,
+      "grad_norm": 6.926656405954893,
+      "kl": 1.1328125,
+      "learning_rate": 1.0841830120348969e-07,
+      "loss": 0.3466,
+      "reward": 2.717449426651001,
+      "reward_std": 0.533571720123291,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.020397868007421494,
+      "rewards/tag_count_reward": 0.9739583730697632,
+      "step": 1889
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 362.81250762939453,
+      "epoch": 0.945,
+      "grad_norm": 7.389820923834504,
+      "kl": 1.1484375,
+      "learning_rate": 1.0826776744855121e-07,
+      "loss": 0.22,
+      "reward": 2.472840905189514,
+      "reward_std": 0.3649376714602113,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.018478597048670053,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1890
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.6041793823242,
+      "epoch": 0.9455,
+      "grad_norm": 6.258707268362366,
+      "kl": 0.951171875,
+      "learning_rate": 1.0811857928660037e-07,
+      "loss": 0.2564,
+      "reward": 2.780505061149597,
+      "reward_std": 0.317706068046391,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.01116166659630835,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1891
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 467.6458435058594,
+      "epoch": 0.946,
+      "grad_norm": 6.7122965752871995,
+      "kl": 1.60546875,
+      "learning_rate": 1.0797073717209013e-07,
+      "loss": 0.432,
+      "reward": 2.605563759803772,
+      "reward_std": 0.6417441666126251,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.015964028425514698,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1892
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.9791717529297,
+      "epoch": 0.9465,
+      "grad_norm": 6.7300045965088735,
+      "kl": 0.9375,
+      "learning_rate": 1.0782424155537314e-07,
+      "loss": 0.2671,
+      "reward": 2.7784035205841064,
+      "reward_std": 0.41331613063812256,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.020207691006362438,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1893
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 548.9166717529297,
+      "epoch": 0.947,
+      "grad_norm": 11.551989318050323,
+      "kl": 1.875,
+      "learning_rate": 1.0767909288270063e-07,
+      "loss": 0.5684,
+      "reward": 2.411288857460022,
+      "reward_std": 0.6332357153296471,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.01753074210137129,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1894
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 663.4375305175781,
+      "epoch": 0.9475,
+      "grad_norm": 11.540534185730696,
+      "kl": 2.625,
+      "learning_rate": 1.0753529159622047e-07,
+      "loss": 0.8717,
+      "reward": 2.090899109840393,
+      "reward_std": 0.831304669380188,
+      "rewards/accuracy_reward": 0.3958333358168602,
+      "rewards/reasoning_steps_reward": 0.8958333432674408,
+      "rewards/repetition_penalty_reward": -0.018475967459380627,
+      "rewards/tag_count_reward": 0.8177083730697632,
+      "step": 1895
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 409.9791717529297,
+      "epoch": 0.948,
+      "grad_norm": 4.503871661371665,
+      "kl": 1.1015625,
+      "learning_rate": 1.0739283813397639e-07,
+      "loss": 0.3769,
+      "reward": 2.603825807571411,
+      "reward_std": 0.39157669246196747,
+      "rewards/accuracy_reward": 0.7083333730697632,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.026382511481642723,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1896
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 481.1875305175781,
+      "epoch": 0.9485,
+      "grad_norm": 14.294347009213048,
+      "kl": 1.68359375,
+      "learning_rate": 1.0725173292990626e-07,
+      "loss": 1.1626,
+      "reward": 2.711683988571167,
+      "reward_std": 0.6804588735103607,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.008802221855148673,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1897
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 573.5208435058594,
+      "epoch": 0.949,
+      "grad_norm": 7.82241175608747,
+      "kl": 2.578125,
+      "learning_rate": 1.0711197641384115e-07,
+      "loss": 0.8068,
+      "reward": 2.370753049850464,
+      "reward_std": 0.9385839700698853,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.847222238779068,
+      "rewards/repetition_penalty_reward": -0.01813590247184038,
+      "rewards/tag_count_reward": 0.8333333730697632,
+      "step": 1898
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 467.0833435058594,
+      "epoch": 0.9495,
+      "grad_norm": 10.691850459801657,
+      "kl": 1.5625,
+      "learning_rate": 1.0697356901150353e-07,
+      "loss": 0.7529,
+      "reward": 2.5153530836105347,
+      "reward_std": 0.8187885880470276,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9097222089767456,
+      "rewards/repetition_penalty_reward": -0.02457757294178009,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1899
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 548.8958435058594,
+      "epoch": 0.95,
+      "grad_norm": 6.406869092852505,
+      "kl": 2.15234375,
+      "learning_rate": 1.068365111445064e-07,
+      "loss": 0.6496,
+      "reward": 2.331941604614258,
+      "reward_std": 0.6170355081558228,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.0291697159409523,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1900
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 595.0000228881836,
+      "epoch": 0.9505,
+      "grad_norm": 17.669263047736123,
+      "kl": 2.453125,
+      "learning_rate": 1.0670080323035176e-07,
+      "loss": 0.3439,
+      "reward": 2.2904654145240784,
+      "reward_std": 0.47801604866981506,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.01509033516049385,
+      "rewards/tag_count_reward": 0.875,
+      "step": 1901
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 541.0208435058594,
+      "epoch": 0.951,
+      "grad_norm": 4.730967706498896,
+      "kl": 2.078125,
+      "learning_rate": 1.0656644568242946e-07,
+      "loss": 0.5377,
+      "reward": 2.2303082942962646,
+      "reward_std": 0.7414398193359375,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.8888888955116272,
+      "rewards/repetition_penalty_reward": -0.028372248634696007,
+      "rewards/tag_count_reward": 0.890625,
+      "step": 1902
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 455.4583435058594,
+      "epoch": 0.9515,
+      "grad_norm": 10.877460950090613,
+      "kl": 1.27734375,
+      "learning_rate": 1.0643343891001591e-07,
+      "loss": 0.7656,
+      "reward": 2.5471227169036865,
+      "reward_std": 0.412681981921196,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.024057872593402863,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1903
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 361.4583435058594,
+      "epoch": 0.952,
+      "grad_norm": 12.3885461588749,
+      "kl": 1.26953125,
+      "learning_rate": 1.063017833182728e-07,
+      "loss": 0.7706,
+      "reward": 2.5880606174468994,
+      "reward_std": 0.756420761346817,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.01610612729564309,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1904
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 311.9166717529297,
+      "epoch": 0.9525,
+      "grad_norm": 10.093151031262789,
+      "kl": 1.091796875,
+      "learning_rate": 1.0617147930824586e-07,
+      "loss": 0.237,
+      "reward": 2.755587577819824,
+      "reward_std": 0.5843307077884674,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.016982081811875105,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 1905
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 426.7708435058594,
+      "epoch": 0.953,
+      "grad_norm": 8.18441727922782,
+      "kl": 1.35546875,
+      "learning_rate": 1.0604252727686379e-07,
+      "loss": 0.6793,
+      "reward": 2.3721930980682373,
+      "reward_std": 0.6670109927654266,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.023640274070203304,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 1906
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 332.4583435058594,
+      "epoch": 0.9535,
+      "grad_norm": 5.6481670754710835,
+      "kl": 0.818359375,
+      "learning_rate": 1.0591492761693674e-07,
+      "loss": 0.2792,
+      "reward": 2.5908457040786743,
+      "reward_std": 0.29737700521945953,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.022001695819199085,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1907
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 324.2083435058594,
+      "epoch": 0.954,
+      "grad_norm": 16.28687030981918,
+      "kl": 1.1875,
+      "learning_rate": 1.0578868071715544e-07,
+      "loss": 0.4517,
+      "reward": 2.8130040168762207,
+      "reward_std": 0.47449105978012085,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.01859321352094412,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1908
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 443.3333435058594,
+      "epoch": 0.9545,
+      "grad_norm": 6.091387449777901,
+      "kl": 1.541015625,
+      "learning_rate": 1.0566378696208987e-07,
+      "loss": 0.6145,
+      "reward": 2.5944347381591797,
+      "reward_std": 0.6039746999740601,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.020148571580648422,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1909
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 468.6666717529297,
+      "epoch": 0.955,
+      "grad_norm": 5.333170833207152,
+      "kl": 1.3046875,
+      "learning_rate": 1.0554024673218806e-07,
+      "loss": 0.5553,
+      "reward": 2.528602719306946,
+      "reward_std": 0.48105429112911224,
+      "rewards/accuracy_reward": 0.6458333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03910571709275246,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1910
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 298.7083435058594,
+      "epoch": 0.9555,
+      "grad_norm": 5.486537631035801,
+      "kl": 1.017578125,
+      "learning_rate": 1.054180604037749e-07,
+      "loss": 0.1708,
+      "reward": 2.73952579498291,
+      "reward_std": 0.3539083171635866,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03130759112536907,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1911
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 327.4791717529297,
+      "epoch": 0.956,
+      "grad_norm": 9.305310519927465,
+      "kl": 0.92578125,
+      "learning_rate": 1.0529722834905125e-07,
+      "loss": 0.2203,
+      "reward": 2.675445556640625,
+      "reward_std": 0.3721562922000885,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.031151846051216125,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1912
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 364.5416717529297,
+      "epoch": 0.9565,
+      "grad_norm": 4.1629931880553315,
+      "kl": 1.1962890625,
+      "learning_rate": 1.0517775093609241e-07,
+      "loss": 0.5035,
+      "reward": 2.6332170963287354,
+      "reward_std": 0.21848932653665543,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03865810390561819,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1913
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 388.91668701171875,
+      "epoch": 0.957,
+      "grad_norm": 6.405039325043405,
+      "kl": 1.72265625,
+      "learning_rate": 1.0505962852884739e-07,
+      "loss": 0.5846,
+      "reward": 2.712180018424988,
+      "reward_std": 0.5378515720367432,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02045894879847765,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1914
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 412.83335876464844,
+      "epoch": 0.9575,
+      "grad_norm": 16.248265342729503,
+      "kl": 1.7421875,
+      "learning_rate": 1.0494286148713744e-07,
+      "loss": 0.9792,
+      "reward": 2.438920497894287,
+      "reward_std": 0.568135529756546,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.014204645762220025,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1915
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 490.60418701171875,
+      "epoch": 0.958,
+      "grad_norm": 13.583903329253506,
+      "kl": 1.8203125,
+      "learning_rate": 1.0482745016665526e-07,
+      "loss": 0.7488,
+      "reward": 2.5179425477981567,
+      "reward_std": 0.6293874979019165,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555522441864,
+      "rewards/repetition_penalty_reward": -0.037613097578287125,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 1916
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 268.9375,
+      "epoch": 0.9585,
+      "grad_norm": 3.8538643099347962,
+      "kl": 0.916015625,
+      "learning_rate": 1.0471339491896373e-07,
+      "loss": 0.3297,
+      "reward": 2.9169150590896606,
+      "reward_std": 0.2011337815783918,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.011904461309313774,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1917
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 373.7083435058594,
+      "epoch": 0.959,
+      "grad_norm": 8.19149127254569,
+      "kl": 1.080078125,
+      "learning_rate": 1.0460069609149496e-07,
+      "loss": 0.3442,
+      "reward": 2.6776570081710815,
+      "reward_std": 0.4415852725505829,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.02546807937324047,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1918
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 425.1041717529297,
+      "epoch": 0.9595,
+      "grad_norm": 7.078016567150192,
+      "kl": 1.62109375,
+      "learning_rate": 1.044893540275491e-07,
+      "loss": 0.5886,
+      "reward": 2.3200796842575073,
+      "reward_std": 0.72013059258461,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.02193425875157118,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1919
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 376.0416717529297,
+      "epoch": 0.96,
+      "grad_norm": 3.7664864308299464,
+      "kl": 1.33203125,
+      "learning_rate": 1.0437936906629334e-07,
+      "loss": 0.6744,
+      "reward": 2.767207145690918,
+      "reward_std": 0.5426684468984604,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.01925129722803831,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1920
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 493.375,
+      "epoch": 0.9605,
+      "grad_norm": 8.937614698943863,
+      "kl": 2.099609375,
+      "learning_rate": 1.0427074154276104e-07,
+      "loss": 0.5657,
+      "reward": 2.1462767124176025,
+      "reward_std": 0.6453874707221985,
+      "rewards/accuracy_reward": 0.3750000149011612,
+      "rewards/reasoning_steps_reward": 0.916666716337204,
+      "rewards/repetition_penalty_reward": -0.020390215329825878,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1921
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 473.91668701171875,
+      "epoch": 0.961,
+      "grad_norm": 13.323431549636778,
+      "kl": 2.11328125,
+      "learning_rate": 1.0416347178785039e-07,
+      "loss": 0.846,
+      "reward": 2.5011810064315796,
+      "reward_std": 0.6241348683834076,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9236111342906952,
+      "rewards/repetition_penalty_reward": -0.02138859312981367,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1922
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 350.37501525878906,
+      "epoch": 0.9615,
+      "grad_norm": 7.917549405794224,
+      "kl": 1.013671875,
+      "learning_rate": 1.0405756012832367e-07,
+      "loss": 0.4971,
+      "reward": 2.590930223464966,
+      "reward_std": 0.4288761019706726,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02191714197397232,
+      "rewards/tag_count_reward": 0.9739583730697632,
+      "step": 1923
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 309.7916717529297,
+      "epoch": 0.962,
+      "grad_norm": 9.512873967750881,
+      "kl": 1.076171875,
+      "learning_rate": 1.0395300688680625e-07,
+      "loss": 0.5754,
+      "reward": 2.356972813606262,
+      "reward_std": 0.4319635033607483,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.011082816403359175,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1924
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 451.14585876464844,
+      "epoch": 0.9625,
+      "grad_norm": 7.649494528326308,
+      "kl": 1.37890625,
+      "learning_rate": 1.0384981238178533e-07,
+      "loss": 0.7388,
+      "reward": 2.306758999824524,
+      "reward_std": 0.5872194170951843,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.019629981368780136,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1925
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.37501525878906,
+      "epoch": 0.963,
+      "grad_norm": 7.3664247239012175,
+      "kl": 0.87109375,
+      "learning_rate": 1.0374797692760933e-07,
+      "loss": 0.29,
+      "reward": 2.7261266708374023,
+      "reward_std": 0.4131031781435013,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02040126919746399,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1926
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 457.72918701171875,
+      "epoch": 0.9635,
+      "grad_norm": 9.740860339646286,
+      "kl": 1.5390625,
+      "learning_rate": 1.036475008344867e-07,
+      "loss": 0.8369,
+      "reward": 2.6725724935531616,
+      "reward_std": 0.5475434064865112,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.02360812947154045,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1927
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 526.0208435058594,
       "epoch": 0.964,
-      "grad_norm": 72.11734538341996,
-      "kl": 6.40625,
+      "grad_norm": 7.32137984381808,
+      "kl": 2.19140625,
       "learning_rate": 1.0354838440848501e-07,
-      "loss": 0.7886,
-      "reward": 2.2568776607513428,
-      "reward_std": 0.48433394730091095,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.017428037710487843,
-      "rewards/tag_count_reward": 0.8645833730697632,
-      "step": 964
+      "loss": 0.8225,
+      "reward": 2.199436902999878,
+      "reward_std": 0.6961483359336853,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.008896507322788239,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1928
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 312.00001525878906,
+      "epoch": 0.9645,
+      "grad_norm": 7.548627408249209,
+      "kl": 1.587890625,
+      "learning_rate": 1.0345062795153009e-07,
+      "loss": 0.22,
+      "reward": 2.5896737575531006,
+      "reward_std": 0.4959706515073776,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.024909449741244316,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1929
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 414.0833435058594,
+      "completion_length": 540.4375152587891,
       "epoch": 0.965,
-      "grad_norm": 22.298889854844614,
-      "kl": 3.390625,
+      "grad_norm": 4.690334600576642,
+      "kl": 1.890625,
       "learning_rate": 1.0335423176140511e-07,
-      "loss": 0.4795,
-      "reward": 2.5144201517105103,
-      "reward_std": 0.6362143456935883,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.958333432674408,
-      "rewards/repetition_penalty_reward": -0.016829868778586388,
-      "rewards/tag_count_reward": 0.9270833432674408,
-      "step": 965
+      "loss": 0.6585,
+      "reward": 2.1660616397857666,
+      "reward_std": 0.6110673546791077,
+      "rewards/accuracy_reward": 0.3750000149011612,
+      "rewards/reasoning_steps_reward": 0.9236111044883728,
+      "rewards/repetition_penalty_reward": -0.01796635054051876,
+      "rewards/tag_count_reward": 0.8854166865348816,
+      "step": 1930
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 483.3958435058594,
+      "epoch": 0.9655,
+      "grad_norm": 4.681975241875133,
+      "kl": 1.26953125,
+      "learning_rate": 1.0325919613174951e-07,
+      "loss": 0.6228,
+      "reward": 2.6714521646499634,
+      "reward_std": 0.6622795760631561,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9583334028720856,
+      "rewards/repetition_penalty_reward": -0.02125630248337984,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1931
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 511.43751525878906,
+      "completion_length": 317.0208435058594,
       "epoch": 0.966,
-      "grad_norm": 21.828332739197077,
-      "kl": 5.1015625,
+      "grad_norm": 5.82480472157766,
+      "kl": 1.294921875,
       "learning_rate": 1.0316552135205837e-07,
-      "loss": 0.8053,
-      "reward": 2.537219762802124,
-      "reward_std": 0.5350492745637894,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.020071997307240963,
-      "rewards/tag_count_reward": 0.8906250298023224,
-      "step": 966
+      "loss": 0.2464,
+      "reward": 2.7479259967803955,
+      "reward_std": 0.4953030524775386,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.010754720773547888,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1932
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 331.5416717529297,
+      "epoch": 0.9665,
+      "grad_norm": 6.724427674491569,
+      "kl": 1.34765625,
+      "learning_rate": 1.0307320770768129e-07,
+      "loss": 0.0875,
+      "reward": 2.34666109085083,
+      "reward_std": 0.47824424505233765,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.02660275436937809,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1933
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 437.9166717529297,
+      "completion_length": 532.9583435058594,
       "epoch": 0.967,
-      "grad_norm": 36.06941047437207,
-      "kl": 4.046875,
+      "grad_norm": 12.406632178922738,
+      "kl": 2.44140625,
       "learning_rate": 1.029822554798216e-07,
-      "loss": 0.7437,
-      "reward": 2.4883724451065063,
-      "reward_std": 0.756380021572113,
-      "rewards/accuracy_reward": 0.6875,
-      "rewards/reasoning_steps_reward": 0.8958333432674408,
-      "rewards/repetition_penalty_reward": -0.016835974529385567,
-      "rewards/tag_count_reward": 0.9218750298023224,
-      "step": 967
+      "loss": 0.7136,
+      "reward": 2.3831958770751953,
+      "reward_std": 0.7278265357017517,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/reasoning_steps_reward": 0.902777761220932,
+      "rewards/repetition_penalty_reward": -0.014373642392456532,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 1934
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 361.47918701171875,
+      "completion_length": 274.7083435058594,
+      "epoch": 0.9675,
+      "grad_norm": 10.363127459700797,
+      "kl": 0.884765625,
+      "learning_rate": 1.0289266494553565e-07,
+      "loss": 0.1206,
+      "reward": 2.2719321250915527,
+      "reward_std": 0.19579820428043604,
+      "rewards/accuracy_reward": 0.2916666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.019734575413167477,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1935
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 292.5208435058594,
       "epoch": 0.968,
-      "grad_norm": 21.352797175646803,
-      "kl": 1.361328125,
+      "grad_norm": 3.7229224037837403,
+      "kl": 1.138671875,
       "learning_rate": 1.0280443637773163e-07,
-      "loss": 0.2428,
-      "reward": 2.3535938262939453,
-      "reward_std": 0.34989143908023834,
-      "rewards/accuracy_reward": 0.4166666865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.017933969385921955,
+      "loss": 0.1786,
+      "reward": 2.8120529651641846,
+      "reward_std": 0.497484490275383,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9444445371627808,
+      "rewards/repetition_penalty_reward": -0.01780821569263935,
       "rewards/tag_count_reward": 0.96875,
-      "step": 968
+      "step": 1936
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 419.14585876464844,
+      "epoch": 0.9685,
+      "grad_norm": 6.842482946515753,
+      "kl": 1.5859375,
+      "learning_rate": 1.0271757004516918e-07,
+      "loss": 0.9308,
+      "reward": 2.6955126523971558,
+      "reward_std": 0.6962899565696716,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.016293123364448547,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1937
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 397.4583435058594,
+      "completion_length": 431.81251525878906,
       "epoch": 0.969,
-      "grad_norm": 35.203792265802775,
-      "kl": 2.890625,
+      "grad_norm": 4.528182947858323,
+      "kl": 1.62890625,
       "learning_rate": 1.0263206621245807e-07,
-      "loss": 0.7547,
-      "reward": 2.354074716567993,
-      "reward_std": 0.3742067515850067,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.9722222089767456,
-      "rewards/repetition_penalty_reward": -0.00877267774194479,
+      "loss": 0.5407,
+      "reward": 2.654048204421997,
+      "reward_std": 0.5423067063093185,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9305555820465088,
+      "rewards/repetition_penalty_reward": -0.016090850345790386,
+      "rewards/tag_count_reward": 0.9062500298023224,
+      "step": 1938
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 419.2916717529297,
+      "epoch": 0.9695,
+      "grad_norm": 10.72838395202056,
+      "kl": 1.0546875,
+      "learning_rate": 1.0254792514005792e-07,
+      "loss": 0.661,
+      "reward": 2.7793266773223877,
+      "reward_std": 0.44574533961713314,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.0418538823723793,
       "rewards/tag_count_reward": 0.953125,
-      "step": 969
+      "step": 1939
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 422.56251525878906,
+      "completion_length": 409.1875,
       "epoch": 0.97,
-      "grad_norm": 38.23995456003063,
-      "kl": 4.59375,
+      "grad_norm": 5.836202176867066,
+      "kl": 1.3515625,
       "learning_rate": 1.0246514708427701e-07,
-      "loss": 0.5101,
-      "reward": 2.6607662439346313,
-      "reward_std": 0.5710070729255676,
-      "rewards/accuracy_reward": 0.8125,
-      "rewards/reasoning_steps_reward": 0.9305556118488312,
-      "rewards/repetition_penalty_reward": -0.014581102412194014,
-      "rewards/tag_count_reward": 0.9322916865348816,
-      "step": 970
+      "loss": 0.6253,
+      "reward": 2.5530662536621094,
+      "reward_std": 0.36100663244724274,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.021586645394563675,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1940
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 352.56251525878906,
+      "epoch": 0.9705,
+      "grad_norm": 4.689568620130673,
+      "kl": 1.009765625,
+      "learning_rate": 1.0238373229727166e-07,
+      "loss": 0.2688,
+      "reward": 2.7452911138534546,
+      "reward_std": 0.22191456332802773,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.013389479368925095,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1941
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 518.6875305175781,
+      "completion_length": 301.7083435058594,
       "epoch": 0.971,
-      "grad_norm": 21.69225757308708,
-      "kl": 5.203125,
+      "grad_norm": 6.599491959106032,
+      "kl": 0.931640625,
       "learning_rate": 1.0230368102704531e-07,
-      "loss": 0.8592,
-      "reward": 2.31989324092865,
-      "reward_std": 0.6834602952003479,
-      "rewards/accuracy_reward": 0.5000000298023224,
-      "rewards/reasoning_steps_reward": 0.9583334028720856,
-      "rewards/repetition_penalty_reward": -0.018648307770490646,
-      "rewards/tag_count_reward": 0.8802083432674408,
-      "step": 971
+      "loss": 0.3311,
+      "reward": 2.8716362714767456,
+      "reward_std": 0.38382330536842346,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.018988667987287045,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1942
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 561.4375305175781,
+      "epoch": 0.9715,
+      "grad_norm": 8.295763130048279,
+      "kl": 2.08203125,
+      "learning_rate": 1.022249935174482e-07,
+      "loss": 1.1474,
+      "reward": 2.3098472356796265,
+      "reward_std": 0.7834429144859314,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.8611111640930176,
+      "rewards/repetition_penalty_reward": -0.020014054141938686,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1943
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 512.9375,
+      "completion_length": 446.3125305175781,
       "epoch": 0.972,
-      "grad_norm": 35.530778404533244,
-      "kl": 6.0859375,
+      "grad_norm": 8.663254346647918,
+      "kl": 1.36328125,
       "learning_rate": 1.0214767000817596e-07,
-      "loss": 0.8544,
-      "reward": 2.3701701164245605,
-      "reward_std": 0.8559737205505371,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.8958334028720856,
-      "rewards/repetition_penalty_reward": -0.02045489940792322,
-      "rewards/tag_count_reward": 0.8697916865348816,
-      "step": 972
+      "loss": 0.5033,
+      "reward": 2.553582787513733,
+      "reward_std": 0.5564675778150558,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.03322279639542103,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1944
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 524.1666870117188,
+      "completion_length": 355.62501525878906,
+      "epoch": 0.9725,
+      "grad_norm": 4.153430704410522,
+      "kl": 1.0390625,
+      "learning_rate": 1.0207171073476951e-07,
+      "loss": 0.247,
+      "reward": 2.5982961654663086,
+      "reward_std": 0.31510170828551054,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03712050523608923,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1945
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 365.6458435058594,
       "epoch": 0.973,
-      "grad_norm": 30.66580558458817,
-      "kl": 8.59375,
+      "grad_norm": 6.995041095518159,
+      "kl": 1.1484375,
       "learning_rate": 1.01997115928614e-07,
-      "loss": 1.0968,
-      "reward": 2.2603888511657715,
-      "reward_std": 0.8733722567558289,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.8888888955116272,
-      "rewards/repetition_penalty_reward": -0.019125062506645918,
-      "rewards/tag_count_reward": 0.8072916865348816,
-      "step": 973
+      "loss": 0.5551,
+      "reward": 2.7796308994293213,
+      "reward_std": 0.35975193604826927,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.018980273976922035,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1946
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 431.12501525878906,
+      "epoch": 0.9735,
+      "grad_norm": 4.410695064600253,
+      "kl": 1.57421875,
+      "learning_rate": 1.0192388581693806e-07,
+      "loss": 0.6488,
+      "reward": 2.5324418544769287,
+      "reward_std": 0.6425078958272934,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.033530326560139656,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1947
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 598.7916870117188,
+      "completion_length": 421.6041717529297,
       "epoch": 0.974,
-      "grad_norm": 21.256888216031047,
-      "kl": 6.8125,
+      "grad_norm": 10.096856330968942,
+      "kl": 1.796875,
       "learning_rate": 1.0185202062281336e-07,
-      "loss": 1.205,
-      "reward": 2.3526848554611206,
-      "reward_std": 0.7947600483894348,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.9305555522441864,
-      "rewards/repetition_penalty_reward": -0.02578743826597929,
-      "rewards/tag_count_reward": 0.8229166865348816,
-      "step": 974
+      "loss": 0.3495,
+      "reward": 2.2948466539382935,
+      "reward_std": 0.3842976242303848,
+      "rewards/accuracy_reward": 0.5000000204890966,
+      "rewards/reasoning_steps_reward": 0.9027777910232544,
+      "rewards/repetition_penalty_reward": -0.019389580003917217,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 1948
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 413.22918701171875,
+      "epoch": 0.9745,
+      "grad_norm": 4.028820810607798,
+      "kl": 1.22265625,
+      "learning_rate": 1.0178152056515371e-07,
+      "loss": 0.7743,
+      "reward": 2.539444088935852,
+      "reward_std": 0.5672547519207001,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.0161115531809628,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 1949
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 607.625,
+      "completion_length": 301.3958435058594,
       "epoch": 0.975,
-      "grad_norm": 19.685066419282403,
-      "kl": 6.890625,
+      "grad_norm": 7.863861707517645,
+      "kl": 1.05859375,
       "learning_rate": 1.017123858587145e-07,
-      "loss": 0.9204,
-      "reward": 2.2289880514144897,
-      "reward_std": 0.5278165340423584,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.9236111044883728,
-      "rewards/repetition_penalty_reward": -0.027956443838775158,
-      "rewards/tag_count_reward": 0.8125000298023224,
-      "step": 975
+      "loss": 0.1674,
+      "reward": 2.773741364479065,
+      "reward_std": 0.4449689909815788,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.023133596405386925,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1950
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 585.8541870117188,
+      "completion_length": 621.6041870117188,
+      "epoch": 0.9755,
+      "grad_norm": 496.05634241089564,
+      "kl": 2.5625,
+      "learning_rate": 1.0164461671409212e-07,
+      "loss": 1.1447,
+      "reward": 2.413469076156616,
+      "reward_std": 0.8790780007839203,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.854166716337204,
+      "rewards/repetition_penalty_reward": -0.018822629936039448,
+      "rewards/tag_count_reward": 0.8489583730697632,
+      "step": 1951
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 449.4166717529297,
       "epoch": 0.976,
-      "grad_norm": 21.89405265517972,
-      "kl": 7.140625,
+      "grad_norm": 4.37873588820322,
+      "kl": 1.5078125,
       "learning_rate": 1.0157821333772304e-07,
-      "loss": 1.0002,
-      "reward": 2.272312641143799,
-      "reward_std": 0.7230339646339417,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.9375,
-      "rewards/repetition_penalty_reward": -0.01935403887182474,
-      "rewards/tag_count_reward": 0.7916666865348816,
-      "step": 976
+      "loss": 0.6414,
+      "reward": 2.761192560195923,
+      "reward_std": 0.3843380808830261,
+      "rewards/accuracy_reward": 0.8958333730697632,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.025265809148550034,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1952
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 313.8958435058594,
+      "epoch": 0.9765,
+      "grad_norm": 7.709851287682846,
+      "kl": 1.125,
+      "learning_rate": 1.0151317593188354e-07,
+      "loss": 0.4473,
+      "reward": 2.742506980895996,
+      "reward_std": 0.28937215672340244,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.009229286457411945,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1953
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 517.0833435058594,
+      "completion_length": 585.6041717529297,
       "epoch": 0.977,
-      "grad_norm": 19.225554130007215,
-      "kl": 4.0234375,
+      "grad_norm": 7.1975720249775526,
+      "kl": 2.0546875,
       "learning_rate": 1.014495046946888e-07,
-      "loss": 0.8685,
-      "reward": 2.6582794189453125,
-      "reward_std": 0.5991591513156891,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.022276088129729033,
-      "rewards/tag_count_reward": 0.8958333432674408,
-      "step": 977
+      "loss": 0.6728,
+      "reward": 2.314143478870392,
+      "reward_std": 0.5873344540596008,
+      "rewards/accuracy_reward": 0.5208333432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.026134257204830647,
+      "rewards/tag_count_reward": 0.8541666865348816,
+      "step": 1954
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 461.8958435058594,
+      "epoch": 0.9775,
+      "grad_norm": 4.880043762137739,
+      "kl": 1.34765625,
+      "learning_rate": 1.0138719982009242e-07,
+      "loss": 0.6723,
+      "reward": 2.327051043510437,
+      "reward_std": 0.6554215252399445,
+      "rewards/accuracy_reward": 0.4791666865348816,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.021907367277890444,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 1955
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 436.8333435058594,
+      "completion_length": 450.6041717529297,
       "epoch": 0.978,
-      "grad_norm": 40.49511138880158,
-      "kl": 2.71875,
+      "grad_norm": 4.714021383660871,
+      "kl": 1.515625,
       "learning_rate": 1.013262614978859e-07,
-      "loss": 0.8484,
-      "reward": 2.702500820159912,
-      "reward_std": 0.6443986296653748,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.024929875042289495,
-      "rewards/tag_count_reward": 0.9218750298023224,
-      "step": 978
+      "loss": 0.3862,
+      "reward": 2.6686251163482666,
+      "reward_std": 0.49254344403743744,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111640930176,
+      "rewards/repetition_penalty_reward": -0.03102776501327753,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1956
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 469.5416717529297,
+      "completion_length": 374.9166717529297,
+      "epoch": 0.9785,
+      "grad_norm": 4.593658438724619,
+      "kl": 1.19140625,
+      "learning_rate": 1.0126668991369792e-07,
+      "loss": 0.3453,
+      "reward": 2.5224099159240723,
+      "reward_std": 0.4157126843929291,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.024465198628604412,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1957
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.00001525878906,
       "epoch": 0.979,
-      "grad_norm": 31.345278883427696,
-      "kl": 3.375,
+      "grad_norm": 8.529770354299277,
+      "kl": 0.966796875,
       "learning_rate": 1.0120848524899386e-07,
-      "loss": 0.9379,
-      "reward": 2.366294264793396,
-      "reward_std": 0.646638810634613,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.0278029702603817,
-      "rewards/tag_count_reward": 0.8802083730697632,
-      "step": 979
+      "loss": 0.3444,
+      "reward": 2.6569780111312866,
+      "reward_std": 0.3028900623321533,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.006216405134182423,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1958
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 473.10418701171875,
+      "epoch": 0.9795,
+      "grad_norm": 5.589279353041215,
+      "kl": 1.9375,
+      "learning_rate": 1.0115164768107522e-07,
+      "loss": 0.7369,
+      "reward": 2.5905479192733765,
+      "reward_std": 0.7043185234069824,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.92361119389534,
+      "rewards/repetition_penalty_reward": -0.025771519169211388,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1959
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 448.4583435058594,
+      "completion_length": 253.41667938232422,
       "epoch": 0.98,
-      "grad_norm": 14.372103163574602,
-      "kl": 2.703125,
+      "grad_norm": 4.817368128682914,
+      "kl": 0.62109375,
       "learning_rate": 1.0109617738307911e-07,
-      "loss": 0.4298,
-      "reward": 2.4362382888793945,
-      "reward_std": 0.5597494542598724,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9583333730697632,
-      "rewards/repetition_penalty_reward": -0.03251180611550808,
-      "rewards/tag_count_reward": 0.9062500298023224,
-      "step": 980
+      "loss": 0.0064,
+      "reward": 2.6383495330810547,
+      "reward_std": 0.2589970678091049,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555820465088,
+      "rewards/repetition_penalty_reward": -0.0213725995272398,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1960
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 366.6041717529297,
+      "epoch": 0.9805,
+      "grad_norm": 10.352217018184477,
+      "kl": 1.03125,
+      "learning_rate": 1.0104207452397761e-07,
+      "loss": 0.6742,
+      "reward": 2.5156456232070923,
+      "reward_std": 0.5356017798185349,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.979166716337204,
+      "rewards/repetition_penalty_reward": -0.015604355372488499,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1961
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 340.18750762939453,
+      "epoch": 0.981,
+      "grad_norm": 5.328236970495506,
+      "kl": 0.94921875,
+      "learning_rate": 1.0098933926857752e-07,
+      "loss": 0.3793,
+      "reward": 2.6204140186309814,
+      "reward_std": 0.24081332981586456,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.9375000596046448,
+      "rewards/repetition_penalty_reward": -0.02541936282068491,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1962
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 710.3125305175781,
+      "epoch": 0.9815,
+      "grad_norm": 17.365217085343485,
+      "kl": 2.859375,
+      "learning_rate": 1.0093797177751944e-07,
+      "loss": 1.1361,
+      "reward": 2.3238528966903687,
+      "reward_std": 0.8956755101680756,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.8472222685813904,
+      "rewards/repetition_penalty_reward": -0.018161091022193432,
+      "rewards/tag_count_reward": 0.8281250298023224,
+      "step": 1963
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 372.0208435058594,
-      "epoch": 0.981,
-      "grad_norm": 40.416913288237026,
-      "kl": 2.2578125,
-      "learning_rate": 1.0098933926857752e-07,
-      "loss": 0.4458,
-      "reward": 2.7070990800857544,
-      "reward_std": 0.5603702366352081,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9444444477558136,
-      "rewards/repetition_penalty_reward": -0.018595370464026928,
-      "rewards/tag_count_reward": 0.9270833432674408,
-      "step": 981
+      "completion_length": 306.7916717529297,
+      "epoch": 0.982,
+      "grad_norm": 5.0392308029500725,
+      "kl": 0.810546875,
+      "learning_rate": 1.0088797220727779e-07,
+      "loss": 0.3007,
+      "reward": 2.6658294200897217,
+      "reward_std": 0.208114517852664,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.026879037730395794,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1964
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 503.1666717529297,
-      "epoch": 0.982,
-      "grad_norm": 23.377929734927985,
-      "kl": 4.28125,
-      "learning_rate": 1.0088797220727779e-07,
-      "loss": 1.1209,
-      "reward": 2.3728084564208984,
-      "reward_std": 0.7684433162212372,
-      "rewards/accuracy_reward": 0.5625000298023224,
-      "rewards/reasoning_steps_reward": 0.951388955116272,
-      "rewards/repetition_penalty_reward": -0.03170543722808361,
-      "rewards/tag_count_reward": 0.8906250298023224,
-      "step": 982
+      "completion_length": 291.56251525878906,
+      "epoch": 0.9825,
+      "grad_norm": 7.335928965442494,
+      "kl": 0.9375,
+      "learning_rate": 1.0083934071015988e-07,
+      "loss": 0.0943,
+      "reward": 2.629801034927368,
+      "reward_std": 0.2985463812947273,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02297679055482149,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1965
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 457.6250305175781,
+      "completion_length": 289.5416793823242,
       "epoch": 0.983,
-      "grad_norm": 23.345141824922493,
-      "kl": 3.3046875,
+      "grad_norm": 8.189186451005964,
+      "kl": 1.376953125,
       "learning_rate": 1.007920774343056e-07,
-      "loss": 0.7212,
-      "reward": 2.3101412057876587,
-      "reward_std": 0.5889628529548645,
-      "rewards/accuracy_reward": 0.4375,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.016247691586613655,
-      "rewards/tag_count_reward": 0.8958333432674408,
-      "step": 983
+      "loss": 0.051,
+      "reward": 2.853655695915222,
+      "reward_std": 0.26282477006316185,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.04391396418213844,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1966
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 345.4583435058594,
+      "epoch": 0.9835,
+      "grad_norm": 8.561365760313354,
+      "kl": 2.14453125,
+      "learning_rate": 1.0074618252368726e-07,
+      "loss": 0.5123,
+      "reward": 2.6916539669036865,
+      "reward_std": 0.6342622339725494,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.02015175297856331,
+      "rewards/tag_count_reward": 0.9270833432674408,
+      "step": 1967
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 370.7708435058594,
+      "completion_length": 316.7291717529297,
       "epoch": 0.984,
-      "grad_norm": 42.40361638319742,
-      "kl": 2.703125,
+      "grad_norm": 5.9460907756296475,
+      "kl": 1.0078125,
       "learning_rate": 1.0070165611810855e-07,
-      "loss": 0.4711,
-      "reward": 2.4439754486083984,
-      "reward_std": 0.4780399203300476,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.02130233682692051,
+      "loss": 0.3107,
+      "reward": 2.4366408586502075,
+      "reward_std": 0.44192972034215927,
+      "rewards/accuracy_reward": 0.5000000223517418,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.016484168358147144,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1968
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 359.85418701171875,
+      "epoch": 0.9845,
+      "grad_norm": 3.5890920698204845,
+      "kl": 1.2109375,
+      "learning_rate": 1.0065849835320473e-07,
+      "loss": 0.3065,
+      "reward": 2.7752894163131714,
+      "reward_std": 0.41077224910259247,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.03721063770353794,
       "rewards/tag_count_reward": 0.9583333432674408,
-      "step": 984
+      "step": 1969
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 569.6666870117188,
+      "completion_length": 339.6041717529297,
       "epoch": 0.985,
-      "grad_norm": 41.35640722099136,
-      "kl": 6.78125,
+      "grad_norm": 4.603168506552496,
+      "kl": 0.732421875,
       "learning_rate": 1.0061670936044178e-07,
-      "loss": 1.0564,
-      "reward": 2.4486454725265503,
-      "reward_std": 0.6525047123432159,
-      "rewards/accuracy_reward": 0.6458333432674408,
+      "loss": 0.1274,
+      "reward": 2.7143149375915527,
+      "reward_std": 0.3336441293358803,
+      "rewards/accuracy_reward": 0.7708333432674408,
       "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.020104597322642803,
-      "rewards/tag_count_reward": 0.8437500298023224,
-      "step": 985
+      "rewards/repetition_penalty_reward": -0.020060266833752394,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1970
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 320.125,
+      "epoch": 0.9855,
+      "grad_norm": 8.307150756656938,
+      "kl": 0.787109375,
+      "learning_rate": 1.0057628926711624e-07,
+      "loss": 0.308,
+      "reward": 2.7315473556518555,
+      "reward_std": 0.42569366097450256,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.888888955116272,
+      "rewards/repetition_penalty_reward": -0.016716606449335814,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1971
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 408.7916717529297,
+      "completion_length": 348.2291717529297,
       "epoch": 0.986,
-      "grad_norm": 47.381447307488855,
-      "kl": 4.13671875,
+      "grad_norm": 8.306320282400852,
+      "kl": 1.294921875,
       "learning_rate": 1.005372381963547e-07,
-      "loss": 0.6591,
-      "reward": 2.723273992538452,
-      "reward_std": 0.3431188315153122,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9861111640930176,
-      "rewards/repetition_penalty_reward": -0.023253954481333494,
-      "rewards/tag_count_reward": 0.9270833730697632,
-      "step": 986
+      "loss": 0.2106,
+      "reward": 2.2555699348449707,
+      "reward_std": 0.2516388399526477,
+      "rewards/accuracy_reward": 0.3333333358168602,
+      "rewards/reasoning_steps_reward": 0.9791666269302368,
+      "rewards/repetition_penalty_reward": -0.01526356441900134,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1972
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 401.1041717529297,
+      "completion_length": 338.18751525878906,
+      "epoch": 0.9865,
+      "grad_norm": 5.448252350820911,
+      "kl": 1.0390625,
+      "learning_rate": 1.0049955626711354e-07,
+      "loss": 0.2096,
+      "reward": 2.6255258321762085,
+      "reward_std": 0.24584404285997152,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9722222089767456,
+      "rewards/repetition_penalty_reward": -0.013363123405724764,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1973
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 401.6458435058594,
       "epoch": 0.987,
-      "grad_norm": 64.30330810008361,
-      "kl": 2.298828125,
+      "grad_norm": 6.478360593950035,
+      "kl": 1.6171875,
       "learning_rate": 1.0046324359417842e-07,
-      "loss": 0.4446,
-      "reward": 2.5345929861068726,
-      "reward_std": 0.383854431565851,
-      "rewards/accuracy_reward": 0.625,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.026170868426561356,
-      "rewards/tag_count_reward": 0.9635416865348816,
-      "step": 987
+      "loss": 0.3997,
+      "reward": 2.263159990310669,
+      "reward_std": 0.6182901561260223,
+      "rewards/accuracy_reward": 0.4166666716337204,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.026770692318677902,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1974
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 285.00001525878906,
+      "epoch": 0.9875,
+      "grad_norm": 9.459666746367304,
+      "kl": 0.9453125,
+      "learning_rate": 1.0042830028816399e-07,
+      "loss": 0.1263,
+      "reward": 2.572731137275696,
+      "reward_std": 0.37718044966459274,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.014074573758989573,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1975
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 508.54168701171875,
+      "completion_length": 276.2708435058594,
       "epoch": 0.988,
-      "grad_norm": 18.399449871940522,
-      "kl": 5.25,
+      "grad_norm": 4.3871213558384845,
+      "kl": 0.861328125,
       "learning_rate": 1.0039472645551372e-07,
-      "loss": 0.7395,
-      "reward": 2.3243101835250854,
-      "reward_std": 0.6121802628040314,
-      "rewards/accuracy_reward": 0.520833358168602,
-      "rewards/reasoning_steps_reward": 0.9375000298023224,
-      "rewards/repetition_penalty_reward": -0.035064928233623505,
-      "rewards/tag_count_reward": 0.9010416865348816,
-      "step": 988
+      "loss": 0.0517,
+      "reward": 2.809394598007202,
+      "reward_std": 0.18590081203728914,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.010049775708466768,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1976
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 232.52084350585938,
+      "epoch": 0.9885,
+      "grad_norm": 8.470772145588477,
+      "kl": 0.68359375,
+      "learning_rate": 1.0036252219849932e-07,
+      "loss": 0.0676,
+      "reward": 2.9624515771865845,
+      "reward_std": 0.08178183203563094,
+      "rewards/accuracy_reward": 0.9791666865348816,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.01671524066478014,
+      "rewards/tag_count_reward": 1.0,
+      "step": 1977
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 360.4583435058594,
+      "completion_length": 424.5208435058594,
       "epoch": 0.989,
-      "grad_norm": 23.549397740992113,
-      "kl": 1.0419921875,
+      "grad_norm": 5.674149225074641,
+      "kl": 1.59375,
       "learning_rate": 1.0033168761522048e-07,
-      "loss": 0.3115,
-      "reward": 2.6984081268310547,
-      "reward_std": 0.3867932856082916,
+      "loss": 0.499,
+      "reward": 2.5746582746505737,
+      "reward_std": 0.4873664379119873,
       "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.015133682638406754,
-      "rewards/tag_count_reward": 0.984375,
-      "step": 989
+      "rewards/reasoning_steps_reward": 0.951388955116272,
+      "rewards/repetition_penalty_reward": -0.02777231764048338,
+      "rewards/tag_count_reward": 0.921875,
+      "step": 1978
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 415.16668701171875,
+      "completion_length": 412.8958435058594,
+      "epoch": 0.9895,
+      "grad_norm": 5.813467245598919,
+      "kl": 1.2734375,
+      "learning_rate": 1.0030222279960469e-07,
+      "loss": 0.6803,
+      "reward": 2.581472635269165,
+      "reward_std": 0.5482227504253387,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.02790247928351164,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1979
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 454.9375,
       "epoch": 0.99,
-      "grad_norm": 33.19104469693633,
-      "kl": 4.4765625,
+      "grad_norm": 6.839115137637116,
+      "kl": 1.765625,
       "learning_rate": 1.002741278414069e-07,
-      "loss": 0.6084,
-      "reward": 2.583367347717285,
-      "reward_std": 0.5392884910106659,
+      "loss": 0.643,
+      "reward": 2.5423187017440796,
+      "reward_std": 0.4872732013463974,
       "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.019063206389546394,
-      "rewards/tag_count_reward": 0.9218750298023224,
-      "step": 990
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.021917639300227165,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1980
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 426.7083435058594,
+      "epoch": 0.9905,
+      "grad_norm": 12.357127431821414,
+      "kl": 1.1328125,
+      "learning_rate": 1.002474028262093e-07,
+      "loss": 0.6164,
+      "reward": 2.5697481632232666,
+      "reward_std": 0.3590293526649475,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.0205296752974391,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1981
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 623.3125305175781,
+      "completion_length": 442.4583435058594,
       "epoch": 0.991,
-      "grad_norm": 80.63884599695989,
-      "kl": 9.921875,
+      "grad_norm": 8.104455391245587,
+      "kl": 1.765625,
       "learning_rate": 1.0022204783542078e-07,
-      "loss": 1.1105,
-      "reward": 2.0245230197906494,
-      "reward_std": 0.5311869978904724,
-      "rewards/accuracy_reward": 0.3333333432674408,
-      "rewards/reasoning_steps_reward": 0.9097222089767456,
-      "rewards/repetition_penalty_reward": -0.010199269745498896,
-      "rewards/tag_count_reward": 0.7916666865348816,
-      "step": 991
+      "loss": 1.1146,
+      "reward": 2.4505070447921753,
+      "reward_std": 0.7328682243824005,
+      "rewards/accuracy_reward": 0.6666666716337204,
+      "rewards/reasoning_steps_reward": 0.9097222685813904,
+      "rewards/repetition_penalty_reward": -0.02692350000143051,
+      "rewards/tag_count_reward": 0.9010416865348816,
+      "step": 1982
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 360.2083435058594,
+      "epoch": 0.9915,
+      "grad_norm": 7.862692984107953,
+      "kl": 0.900390625,
+      "learning_rate": 1.001980629462772e-07,
+      "loss": 0.4929,
+      "reward": 2.743220090866089,
+      "reward_std": 0.480159193277359,
+      "rewards/accuracy_reward": 0.8125000298023224,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02414090372622013,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 1983
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 506.25,
+      "completion_length": 435.0,
       "epoch": 0.992,
-      "grad_norm": 18.011334199256787,
-      "kl": 5.390625,
+      "grad_norm": 8.155193647156521,
+      "kl": 1.6015625,
       "learning_rate": 1.0017544823184055e-07,
-      "loss": 0.9891,
-      "reward": 2.471301794052124,
-      "reward_std": 0.5606184005737305,
-      "rewards/accuracy_reward": 0.6458333432674408,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.014809360727667809,
-      "rewards/tag_count_reward": 0.8958333730697632,
-      "step": 992
+      "loss": 0.4418,
+      "reward": 2.546668291091919,
+      "reward_std": 0.5175595879554749,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9375,
+      "rewards/repetition_penalty_reward": -0.0210402044467628,
+      "rewards/tag_count_reward": 0.9218750298023224,
+      "step": 1984
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 307.6458435058594,
+      "epoch": 0.9925,
+      "grad_norm": 11.349064132553734,
+      "kl": 1.390625,
+      "learning_rate": 1.0015420376099923e-07,
+      "loss": 0.395,
+      "reward": 2.6141462326049805,
+      "reward_std": 0.3557308465242386,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.01953439600765705,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 1985
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 409.43751525878906,
+      "completion_length": 390.625,
       "epoch": 0.993,
-      "grad_norm": 28.657341564218928,
-      "kl": 1.66796875,
+      "grad_norm": 3.8219642488097927,
+      "kl": 1.1328125,
       "learning_rate": 1.001343295984676e-07,
-      "loss": 0.5813,
-      "reward": 2.359801173210144,
-      "reward_std": 0.4328533709049225,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.972222238779068,
-      "rewards/repetition_penalty_reward": -0.01867112284526229,
-      "rewards/tag_count_reward": 0.9687500298023224,
-      "step": 993
+      "loss": 0.4817,
+      "reward": 2.7117778062820435,
+      "reward_std": 0.5155874937772751,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.034749857150018215,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 1986
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 606.7916870117188,
+      "epoch": 0.9935,
+      "grad_norm": 14.173929864366983,
+      "kl": 2.625,
+      "learning_rate": 1.0011582580478576e-07,
+      "loss": 0.6239,
+      "reward": 2.241394519805908,
+      "reward_std": 0.7578141689300537,
+      "rewards/accuracy_reward": 0.5000000298023224,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.025966664776206017,
+      "rewards/tag_count_reward": 0.8645833432674408,
+      "step": 1987
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 437.9583435058594,
+      "completion_length": 324.3541793823242,
       "epoch": 0.994,
-      "grad_norm": 14.15639088003171,
-      "kl": 2.4453125,
+      "grad_norm": 5.893395799651384,
+      "kl": 1.6328125,
       "learning_rate": 1.0009869243631952e-07,
-      "loss": 0.6178,
-      "reward": 2.6296534538269043,
-      "reward_std": 0.6427464187145233,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.012707796646282077,
-      "rewards/tag_count_reward": 0.9479166865348816,
-      "step": 994
+      "loss": 0.4473,
+      "reward": 2.572134852409363,
+      "reward_std": 0.43454277515411377,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.016406969632953405,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 1988
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 365.5416793823242,
+      "epoch": 0.9945,
+      "grad_norm": 10.350146304014267,
+      "kl": 0.990234375,
+      "learning_rate": 1.000829295452601e-07,
+      "loss": 0.4707,
+      "reward": 2.8346047401428223,
+      "reward_std": 0.31515760254114866,
+      "rewards/accuracy_reward": 0.8958333432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.02650655061006546,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 1989
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 422.2083435058594,
+      "completion_length": 510.5208435058594,
       "epoch": 0.995,
-      "grad_norm": 21.464668161928536,
-      "kl": 2.0234375,
+      "grad_norm": 10.609712135513492,
+      "kl": 1.94140625,
       "learning_rate": 1.0006853717962393e-07,
-      "loss": 0.5771,
-      "reward": 2.529154658317566,
-      "reward_std": 0.3804011940956116,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.015984368277713656,
-      "rewards/tag_count_reward": 0.9479166865348816,
-      "step": 995
+      "loss": 0.4845,
+      "reward": 2.408261299133301,
+      "reward_std": 0.6599603295326233,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.010141530307009816,
+      "rewards/tag_count_reward": 0.8906250298023224,
+      "step": 1990
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 329.87501525878906,
+      "epoch": 0.9955,
+      "grad_norm": 8.570082105600441,
+      "kl": 1.505859375,
+      "learning_rate": 1.0005551538325274e-07,
+      "loss": 0.2282,
+      "reward": 2.8254886865615845,
+      "reward_std": 0.2351871496066451,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02173378225415945,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 1991
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 450.04168701171875,
+      "completion_length": 315.1666717529297,
       "epoch": 0.996,
-      "grad_norm": 29.84614498881874,
-      "kl": 3.0390625,
+      "grad_norm": 6.189976928471374,
+      "kl": 0.90234375,
       "learning_rate": 1.000438641958131e-07,
-      "loss": 0.9596,
-      "reward": 2.2574844360351562,
-      "reward_std": 0.5413917303085327,
-      "rewards/accuracy_reward": 0.375,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.015085038263350725,
-      "rewards/tag_count_reward": 0.9322916865348816,
-      "step": 996
+      "loss": 0.2172,
+      "reward": 2.61415696144104,
+      "reward_std": 0.2505191368982196,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.022995917359367013,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1992
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 411.0208435058594,
+      "completion_length": 506.27085876464844,
+      "epoch": 0.9965,
+      "grad_norm": 9.237622275005121,
+      "kl": 4.1015625,
+      "learning_rate": 1.0003358365279661e-07,
+      "loss": 0.5888,
+      "reward": 2.3739354014396667,
+      "reward_std": 0.7014772593975067,
+      "rewards/accuracy_reward": 0.6041666716337204,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.020161897875368595,
+      "rewards/tag_count_reward": 0.8593750298023224,
+      "step": 1993
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 502.6458435058594,
       "epoch": 0.997,
-      "grad_norm": 22.404250638364193,
-      "kl": 4.38671875,
+      "grad_norm": 7.126787307275001,
+      "kl": 1.69140625,
       "learning_rate": 1.0002467378551954e-07,
-      "loss": 0.5886,
-      "reward": 2.3560396432876587,
-      "reward_std": 0.5860812664031982,
-      "rewards/accuracy_reward": 0.4791666865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.013752035796642303,
-      "rewards/tag_count_reward": 0.9114583730697632,
-      "step": 997
+      "loss": 0.7538,
+      "reward": 2.45101535320282,
+      "reward_std": 0.8071577250957489,
+      "rewards/accuracy_reward": 0.6458333730697632,
+      "rewards/reasoning_steps_reward": 0.9305556118488312,
+      "rewards/repetition_penalty_reward": -0.021206957288086414,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 1994
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 406.2083435058594,
+      "completion_length": 306.50001525878906,
+      "epoch": 0.9975,
+      "grad_norm": 4.0754860395849875,
+      "kl": 1.087890625,
+      "learning_rate": 1.000171346211229e-07,
+      "loss": 0.2516,
+      "reward": 2.844114303588867,
+      "reward_std": 0.32640238106250763,
+      "rewards/accuracy_reward": 0.9375000298023224,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.027413712814450264,
+      "rewards/tag_count_reward": 0.96875,
+      "step": 1995
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 275.5208435058594,
       "epoch": 0.998,
-      "grad_norm": 14.675909635758037,
-      "kl": 4.03125,
+      "grad_norm": 10.126201588856906,
+      "kl": 0.859375,
       "learning_rate": 1.0001096618257236e-07,
-      "loss": 0.5068,
-      "reward": 2.554211378097534,
-      "reward_std": 0.657451719045639,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9444444477558136,
-      "rewards/repetition_penalty_reward": -0.025649813003838062,
-      "rewards/tag_count_reward": 0.9062500298023224,
-      "step": 998
+      "loss": 0.3494,
+      "reward": 2.716609239578247,
+      "reward_std": 0.06195330573245883,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.017765806522220373,
+      "rewards/tag_count_reward": 0.984375,
+      "step": 1996
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 413.4791717529297,
+      "epoch": 0.9985,
+      "grad_norm": 8.06318792454859,
+      "kl": 1.3515625,
+      "learning_rate": 1.0000616848865797e-07,
+      "loss": 0.4413,
+      "reward": 2.815260648727417,
+      "reward_std": 0.36746685206890106,
+      "rewards/accuracy_reward": 0.9166666865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02675328589975834,
+      "rewards/tag_count_reward": 0.953125,
+      "step": 1997
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 529.0416717529297,
+      "completion_length": 338.9583435058594,
       "epoch": 0.999,
-      "grad_norm": 25.41791128417754,
-      "kl": 4.296875,
+      "grad_norm": 5.2700476094611615,
+      "kl": 1.13671875,
       "learning_rate": 1.0000274155399433e-07,
-      "loss": 0.7834,
-      "reward": 2.529424786567688,
-      "reward_std": 0.7073009014129639,
+      "loss": 0.3144,
+      "reward": 2.762750267982483,
+      "reward_std": 0.3325686603784561,
+      "rewards/accuracy_reward": 0.8333333730697632,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.020235874690115452,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 1998
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 535.2083435058594,
+      "epoch": 0.9995,
+      "grad_norm": 3.755049968060971,
+      "kl": 1.91015625,
+      "learning_rate": 1.0000068538902053e-07,
+      "loss": 0.7273,
+      "reward": 2.472402811050415,
+      "reward_std": 0.8260309398174286,
       "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.019186487421393394,
-      "rewards/tag_count_reward": 0.8958333432674408,
-      "step": 999
+      "rewards/reasoning_steps_reward": 0.92361119389534,
+      "rewards/repetition_penalty_reward": -0.013708289712667465,
+      "rewards/tag_count_reward": 0.8750000298023224,
+      "step": 1999
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 462.625,
+      "completion_length": 969.0625,
       "epoch": 1.0,
-      "grad_norm": 21.030282475146205,
-      "kl": 3.8828125,
+      "grad_norm": 41.67279675930041,
+      "kl": 2.5078125,
       "learning_rate": 1e-07,
-      "loss": 1.0549,
-      "reward": 2.609420418739319,
-      "reward_std": 0.6920914947986603,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9652778208255768,
-      "rewards/repetition_penalty_reward": -0.012107519898563623,
-      "rewards/tag_count_reward": 0.9270833432674408,
-      "step": 1000
+      "loss": 0.6262,
+      "reward": 2.2097694873809814,
+      "reward_std": 0.4844767898321152,
+      "rewards/accuracy_reward": 0.4166666865348816,
+      "rewards/reasoning_steps_reward": 0.944444477558136,
+      "rewards/repetition_penalty_reward": -0.010716728633269668,
+      "rewards/tag_count_reward": 0.8593750298023224,
+      "step": 2000
     },
     {
       "epoch": 1.0,
-      "step": 1000,
+      "step": 2000,
       "total_flos": 0.0,
-      "train_loss": 0.49418719741604583,
-      "train_runtime": 57748.1799,
-      "train_samples_per_second": 0.069,
-      "train_steps_per_second": 0.017
+      "train_loss": 1.366488031092214,
+      "train_runtime": 85280.8428,
+      "train_samples_per_second": 0.094,
+      "train_steps_per_second": 0.023
     }
   ],
   "logging_steps": 1,
-  "max_steps": 1000,
+  "max_steps": 2000,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 200,