diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,5353 +1,10681 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9985007496251874,
+  "epoch": 0.9988751406074241,
   "eval_steps": 500,
-  "global_step": 333,
+  "global_step": 666,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "clip_ratio": 0.0,
-      "completion_length": 125.27083587646484,
-      "epoch": 0.0029985007496251873,
-      "grad_norm": 2.316068982886434,
+      "completion_length": 150.88541984558105,
+      "epoch": 0.0014998125234345708,
+      "grad_norm": 2.014207151760731,
       "kl": 0.0,
-      "learning_rate": 5.882352941176471e-07,
-      "loss": -0.139,
-      "reward": 0.4294887036085129,
-      "reward_std": 0.6044371202588081,
-      "rewards/accuracy_reward": 0.09375000419095159,
-      "rewards/reasoning_steps_reward": 0.0920138955116272,
-      "rewards/repetition_penalty_reward": -0.05184812843799591,
-      "rewards/tag_count_reward": 0.295572929084301,
+      "learning_rate": 2.9850746268656716e-07,
+      "loss": -0.123,
+      "reward": 0.39659371972084045,
+      "reward_std": 0.6371014267206192,
+      "rewards/accuracy_reward": 0.1250000037252903,
+      "rewards/reasoning_steps_reward": 0.04861111380159855,
+      "rewards/repetition_penalty_reward": -0.05305909365415573,
+      "rewards/tag_count_reward": 0.2760416716337204,
       "step": 1
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 129.07812881469727,
-      "epoch": 0.005997001499250375,
-      "grad_norm": 2.528986939979676,
+      "completion_length": 139.43750381469727,
+      "epoch": 0.0029996250468691415,
+      "grad_norm": 2.2602676300267803,
       "kl": 0.0,
-      "learning_rate": 1.1764705882352942e-06,
-      "loss": -0.1784,
-      "reward": 0.4756753593683243,
-      "reward_std": 0.6413701921701431,
-      "rewards/accuracy_reward": 0.11979167070239782,
-      "rewards/reasoning_steps_reward": 0.1041666753590107,
-      "rewards/repetition_penalty_reward": -0.04776214715093374,
-      "rewards/tag_count_reward": 0.299479179084301,
+      "learning_rate": 5.970149253731343e-07,
+      "loss": -0.0563,
+      "reward": 0.29565349593758583,
+      "reward_std": 0.5291136056184769,
+      "rewards/accuracy_reward": 0.07812500046566129,
+      "rewards/reasoning_steps_reward": 0.039930558297783136,
+      "rewards/repetition_penalty_reward": -0.04505832493305206,
+      "rewards/tag_count_reward": 0.2226562537252903,
       "step": 2
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 116.48437881469727,
-      "epoch": 0.008995502248875561,
-      "grad_norm": 2.8435255560528976,
-      "kl": 0.00046443939208984375,
-      "learning_rate": 1.7647058823529414e-06,
-      "loss": -0.1512,
-      "reward": 0.4948234558105469,
-      "reward_std": 0.6535268872976303,
-      "rewards/accuracy_reward": 0.1510416679084301,
-      "rewards/reasoning_steps_reward": 0.07118056109175086,
-      "rewards/repetition_penalty_reward": -0.04380502179265022,
-      "rewards/tag_count_reward": 0.3164062649011612,
+      "completion_length": 153.2708396911621,
+      "epoch": 0.0044994375703037125,
+      "grad_norm": 2.1173606964196554,
+      "kl": 0.000335693359375,
+      "learning_rate": 8.955223880597015e-07,
+      "loss": -0.1131,
+      "reward": 0.4513879381120205,
+      "reward_std": 0.6719020158052444,
+      "rewards/accuracy_reward": 0.14583333721384406,
+      "rewards/reasoning_steps_reward": 0.07986111380159855,
+      "rewards/repetition_penalty_reward": -0.047744009643793106,
+      "rewards/tag_count_reward": 0.2734375037252903,
       "step": 3
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 113.15625190734863,
-      "epoch": 0.01199400299850075,
-      "grad_norm": 2.6137621112730782,
-      "kl": 0.0006656646728515625,
-      "learning_rate": 2.3529411764705885e-06,
-      "loss": -0.0962,
-      "reward": 0.39275161921977997,
-      "reward_std": 0.5565674006938934,
-      "rewards/accuracy_reward": 0.0937500037252903,
-      "rewards/reasoning_steps_reward": 0.0590277835726738,
-      "rewards/repetition_penalty_reward": -0.0386720122769475,
-      "rewards/tag_count_reward": 0.2786458358168602,
+      "completion_length": 143.56250381469727,
+      "epoch": 0.005999250093738283,
+      "grad_norm": 2.1085551568397918,
+      "kl": 0.00038623809814453125,
+      "learning_rate": 1.1940298507462686e-06,
+      "loss": -0.0492,
+      "reward": 0.2926744148135185,
+      "reward_std": 0.5166807025671005,
+      "rewards/accuracy_reward": 0.07812500046566129,
+      "rewards/reasoning_steps_reward": 0.0451388917863369,
+      "rewards/repetition_penalty_reward": -0.04933948162943125,
+      "rewards/tag_count_reward": 0.2187500037252903,
       "step": 4
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 122.72396278381348,
-      "epoch": 0.014992503748125937,
-      "grad_norm": 2.3547773173814814,
-      "kl": 0.0024394989013671875,
-      "learning_rate": 2.9411764705882355e-06,
-      "loss": -0.125,
-      "reward": 0.5311619490385056,
-      "reward_std": 0.61662757396698,
-      "rewards/accuracy_reward": 0.14062500139698386,
-      "rewards/reasoning_steps_reward": 0.0885416716337204,
-      "rewards/repetition_penalty_reward": -0.05737974401563406,
-      "rewards/tag_count_reward": 0.3593750074505806,
+      "completion_length": 139.6354217529297,
+      "epoch": 0.0074990626171728535,
+      "grad_norm": 2.0783939318892255,
+      "kl": 0.0004963874816894531,
+      "learning_rate": 1.4925373134328358e-06,
+      "loss": -0.0786,
+      "reward": 0.33502432703971863,
+      "reward_std": 0.5550277233123779,
+      "rewards/accuracy_reward": 0.08854166977107525,
+      "rewards/reasoning_steps_reward": 0.055555559694767,
+      "rewards/repetition_penalty_reward": -0.044749997556209564,
+      "rewards/tag_count_reward": 0.2356770858168602,
       "step": 5
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 138.2447967529297,
-      "epoch": 0.017991004497751123,
-      "grad_norm": 2.1062805811826757,
-      "kl": 0.055267333984375,
-      "learning_rate": 3.529411764705883e-06,
-      "loss": -0.0627,
-      "reward": 0.7481685727834702,
-      "reward_std": 0.7483109384775162,
-      "rewards/accuracy_reward": 0.2708333432674408,
-      "rewards/reasoning_steps_reward": 0.06076389271765947,
-      "rewards/repetition_penalty_reward": -0.06129324156790972,
-      "rewards/tag_count_reward": 0.4778645858168602,
+      "completion_length": 141.94271087646484,
+      "epoch": 0.008998875140607425,
+      "grad_norm": 2.140600840808783,
+      "kl": 0.001529693603515625,
+      "learning_rate": 1.791044776119403e-06,
+      "loss": -0.1144,
+      "reward": 0.3174091763794422,
+      "reward_std": 0.5731675401329994,
+      "rewards/accuracy_reward": 0.08854166697710752,
+      "rewards/reasoning_steps_reward": 0.04513889132067561,
+      "rewards/repetition_penalty_reward": -0.05325053818523884,
+      "rewards/tag_count_reward": 0.2369791753590107,
       "step": 6
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 150.7708396911621,
-      "epoch": 0.020989505247376312,
-      "grad_norm": 91.80938835633462,
-      "kl": 2.9296875,
-      "learning_rate": 4.11764705882353e-06,
-      "loss": 0.0604,
-      "reward": 1.0586681962013245,
-      "reward_std": 0.6929789483547211,
-      "rewards/accuracy_reward": 0.3489583358168602,
-      "rewards/reasoning_steps_reward": 0.1371527910232544,
-      "rewards/repetition_penalty_reward": -0.06676589138805866,
-      "rewards/tag_count_reward": 0.6393229365348816,
+      "completion_length": 139.45833778381348,
+      "epoch": 0.010498687664041995,
+      "grad_norm": 2.0946221841916577,
+      "kl": 0.012420654296875,
+      "learning_rate": 2.08955223880597e-06,
+      "loss": -0.0705,
+      "reward": 0.4593450725078583,
+      "reward_std": 0.5785461291670799,
+      "rewards/accuracy_reward": 0.09895833488553762,
+      "rewards/reasoning_steps_reward": 0.032986113568767905,
+      "rewards/repetition_penalty_reward": -0.05020354688167572,
+      "rewards/tag_count_reward": 0.377604179084301,
       "step": 7
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 154.0104217529297,
-      "epoch": 0.0239880059970015,
-      "grad_norm": 80.12581730059306,
-      "kl": 2.7734375,
-      "learning_rate": 4.705882352941177e-06,
-      "loss": 0.0618,
-      "reward": 1.2028335630893707,
-      "reward_std": 0.6846612095832825,
-      "rewards/accuracy_reward": 0.4687500074505806,
-      "rewards/reasoning_steps_reward": 0.1232638955116272,
-      "rewards/repetition_penalty_reward": -0.06626365892589092,
-      "rewards/tag_count_reward": 0.6770833730697632,
+      "completion_length": 160.0104217529297,
+      "epoch": 0.011998500187476566,
+      "grad_norm": 1.610019846974552,
+      "kl": 0.0306396484375,
+      "learning_rate": 2.3880597014925373e-06,
+      "loss": -0.0826,
+      "reward": 0.5668669790029526,
+      "reward_std": 0.695038303732872,
+      "rewards/accuracy_reward": 0.1458333358168602,
+      "rewards/reasoning_steps_reward": 0.08506944868713617,
+      "rewards/repetition_penalty_reward": -0.053358727134764194,
+      "rewards/tag_count_reward": 0.3893229216337204,
       "step": 8
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 161.4791717529297,
-      "epoch": 0.026986506746626688,
-      "grad_norm": 25.478259616973777,
-      "kl": 0.8330078125,
-      "learning_rate": 5.294117647058824e-06,
-      "loss": -0.0184,
-      "reward": 1.3118138909339905,
-      "reward_std": 0.7422287911176682,
-      "rewards/accuracy_reward": 0.4843750149011612,
-      "rewards/reasoning_steps_reward": 0.18055557273328304,
-      "rewards/repetition_penalty_reward": -0.07056461833417416,
-      "rewards/tag_count_reward": 0.7174479365348816,
+      "completion_length": 171.60937881469727,
+      "epoch": 0.013498312710911136,
+      "grad_norm": 19.671391688045556,
+      "kl": 0.759765625,
+      "learning_rate": 2.686567164179105e-06,
+      "loss": -0.1139,
+      "reward": 0.9202403426170349,
+      "reward_std": 0.7923817485570908,
+      "rewards/accuracy_reward": 0.3020833432674408,
+      "rewards/reasoning_steps_reward": 0.1649305671453476,
+      "rewards/repetition_penalty_reward": -0.07541941851377487,
+      "rewards/tag_count_reward": 0.5286458432674408,
       "step": 9
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 165.47916793823242,
-      "epoch": 0.029985007496251874,
-      "grad_norm": 3.3317610347487396,
-      "kl": 0.205810546875,
-      "learning_rate": 5.882352941176471e-06,
-      "loss": -0.0374,
-      "reward": 1.234879344701767,
-      "reward_std": 0.7370101809501648,
-      "rewards/accuracy_reward": 0.432291679084301,
-      "rewards/reasoning_steps_reward": 0.1458333469927311,
-      "rewards/repetition_penalty_reward": -0.06069365330040455,
-      "rewards/tag_count_reward": 0.7174479365348816,
+      "completion_length": 173.39062881469727,
+      "epoch": 0.014998125234345707,
+      "grad_norm": 116.39843362025961,
+      "kl": 3.4140625,
+      "learning_rate": 2.9850746268656716e-06,
+      "loss": 0.0701,
+      "reward": 0.9044180512428284,
+      "reward_std": 0.7367634326219559,
+      "rewards/accuracy_reward": 0.2968750074505806,
+      "rewards/reasoning_steps_reward": 0.06423611333593726,
+      "rewards/repetition_penalty_reward": -0.06867224909365177,
+      "rewards/tag_count_reward": 0.6119791716337204,
       "step": 10
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 178.56771087646484,
-      "epoch": 0.03298350824587706,
-      "grad_norm": 1.2483918606755664,
-      "kl": 0.08367919921875,
-      "learning_rate": 6.470588235294119e-06,
-      "loss": -0.064,
-      "reward": 1.2948878109455109,
-      "reward_std": 0.7816713899374008,
-      "rewards/accuracy_reward": 0.3489583432674408,
-      "rewards/reasoning_steps_reward": 0.2638889104127884,
-      "rewards/repetition_penalty_reward": -0.07577204331755638,
-      "rewards/tag_count_reward": 0.7578125149011612,
+      "completion_length": 165.9322967529297,
+      "epoch": 0.016497937757780277,
+      "grad_norm": 16.02276467950822,
+      "kl": 0.72412109375,
+      "learning_rate": 3.283582089552239e-06,
+      "loss": 0.0225,
+      "reward": 1.0981639921665192,
+      "reward_std": 0.7406069040298462,
+      "rewards/accuracy_reward": 0.416666679084301,
+      "rewards/reasoning_steps_reward": 0.07812500465661287,
+      "rewards/repetition_penalty_reward": -0.06459642946720123,
+      "rewards/tag_count_reward": 0.6679687649011612,
       "step": 11
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 193.20312881469727,
-      "epoch": 0.035982008995502246,
-      "grad_norm": 1.2630629002924778,
-      "kl": 0.05889892578125,
-      "learning_rate": 7.058823529411766e-06,
-      "loss": -0.0437,
-      "reward": 1.472415030002594,
-      "reward_std": 0.7569083422422409,
-      "rewards/accuracy_reward": 0.473958358168602,
-      "rewards/reasoning_steps_reward": 0.2899305671453476,
-      "rewards/repetition_penalty_reward": -0.061005206778645515,
-      "rewards/tag_count_reward": 0.7695312649011612,
+      "completion_length": 188.2135467529297,
+      "epoch": 0.01799775028121485,
+      "grad_norm": 5.745682328164949,
+      "kl": 0.293212890625,
+      "learning_rate": 3.582089552238806e-06,
+      "loss": 0.0254,
+      "reward": 1.068365916609764,
+      "reward_std": 0.7423798739910126,
+      "rewards/accuracy_reward": 0.3593750149011612,
+      "rewards/reasoning_steps_reward": 0.1111111156642437,
+      "rewards/repetition_penalty_reward": -0.0635785311460495,
+      "rewards/tag_count_reward": 0.661458358168602,
       "step": 12
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 209.9791717529297,
-      "epoch": 0.038980509745127435,
-      "grad_norm": 1.1307015314913393,
-      "kl": 0.04443359375,
-      "learning_rate": 7.647058823529411e-06,
-      "loss": -0.0193,
-      "reward": 1.6883811056613922,
-      "reward_std": 0.7850492298603058,
-      "rewards/accuracy_reward": 0.5625000223517418,
-      "rewards/reasoning_steps_reward": 0.392361119389534,
-      "rewards/repetition_penalty_reward": -0.06986541766673326,
-      "rewards/tag_count_reward": 0.8033854216337204,
+      "completion_length": 160.97916793823242,
+      "epoch": 0.01949756280464942,
+      "grad_norm": 4.696162571842331,
+      "kl": 0.228271484375,
+      "learning_rate": 3.8805970149253735e-06,
+      "loss": -0.0185,
+      "reward": 1.1401186883449554,
+      "reward_std": 0.7335168719291687,
+      "rewards/accuracy_reward": 0.3645833395421505,
+      "rewards/reasoning_steps_reward": 0.10416667349636555,
+      "rewards/repetition_penalty_reward": -0.05519386660307646,
+      "rewards/tag_count_reward": 0.7265625149011612,
       "step": 13
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 222.8385467529297,
-      "epoch": 0.041979010494752625,
-      "grad_norm": 1.0685540714109272,
-      "kl": 0.08837890625,
-      "learning_rate": 8.23529411764706e-06,
-      "loss": -0.0437,
-      "reward": 1.8284152746200562,
-      "reward_std": 0.7906496375799179,
-      "rewards/accuracy_reward": 0.5468750149011612,
-      "rewards/reasoning_steps_reward": 0.5121527835726738,
-      "rewards/repetition_penalty_reward": -0.07045630738139153,
-      "rewards/tag_count_reward": 0.8398437649011612,
+      "completion_length": 149.4947967529297,
+      "epoch": 0.02099737532808399,
+      "grad_norm": 4.440887208265822,
+      "kl": 0.205322265625,
+      "learning_rate": 4.17910447761194e-06,
+      "loss": 0.0925,
+      "reward": 1.1221649795770645,
+      "reward_std": 0.6743175089359283,
+      "rewards/accuracy_reward": 0.3958333469927311,
+      "rewards/reasoning_steps_reward": 0.039930558763444424,
+      "rewards/repetition_penalty_reward": -0.06490109767764807,
+      "rewards/tag_count_reward": 0.7513020932674408,
       "step": 14
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 244.41667556762695,
-      "epoch": 0.044977511244377814,
-      "grad_norm": 0.9074510920159916,
-      "kl": 0.08544921875,
-      "learning_rate": 8.823529411764707e-06,
-      "loss": -0.0056,
-      "reward": 2.034987300634384,
-      "reward_std": 0.6369659751653671,
-      "rewards/accuracy_reward": 0.5000000074505806,
-      "rewards/reasoning_steps_reward": 0.7569444626569748,
-      "rewards/repetition_penalty_reward": -0.10607173293828964,
-      "rewards/tag_count_reward": 0.884114608168602,
+      "completion_length": 164.8072967529297,
+      "epoch": 0.02249718785151856,
+      "grad_norm": 4.632359176117392,
+      "kl": 0.2489013671875,
+      "learning_rate": 4.477611940298508e-06,
+      "loss": 0.0844,
+      "reward": 1.2497529983520508,
+      "reward_std": 0.6609881520271301,
+      "rewards/accuracy_reward": 0.432291679084301,
+      "rewards/reasoning_steps_reward": 0.09027777938172221,
+      "rewards/repetition_penalty_reward": -0.0488581582903862,
+      "rewards/tag_count_reward": 0.7760416865348816,
       "step": 15
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 245.24480438232422,
-      "epoch": 0.047976011994003,
-      "grad_norm": 0.958833819569195,
-      "kl": 0.107177734375,
-      "learning_rate": 9.411764705882354e-06,
-      "loss": 0.064,
-      "reward": 2.3543498516082764,
-      "reward_std": 0.5774905681610107,
-      "rewards/accuracy_reward": 0.6822916865348816,
-      "rewards/reasoning_steps_reward": 0.833333358168602,
-      "rewards/repetition_penalty_reward": -0.09486887603998184,
-      "rewards/tag_count_reward": 0.9335937649011612,
+      "completion_length": 159.5104217529297,
+      "epoch": 0.023997000374953132,
+      "grad_norm": 1.5383616347232285,
+      "kl": 0.078369140625,
+      "learning_rate": 4.7761194029850745e-06,
+      "loss": 0.016,
+      "reward": 1.3225627541542053,
+      "reward_std": 0.6928769499063492,
+      "rewards/accuracy_reward": 0.5156250149011612,
+      "rewards/reasoning_steps_reward": 0.0642361156642437,
+      "rewards/repetition_penalty_reward": -0.04896502383053303,
+      "rewards/tag_count_reward": 0.7916667014360428,
       "step": 16
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 224.5833396911621,
-      "epoch": 0.050974512743628186,
-      "grad_norm": 32.488470313797116,
-      "kl": 0.901611328125,
-      "learning_rate": 1e-05,
-      "loss": 0.0729,
-      "reward": 2.285028785467148,
-      "reward_std": 0.5724412277340889,
-      "rewards/accuracy_reward": 0.5885416716337204,
-      "rewards/reasoning_steps_reward": 0.885416716337204,
-      "rewards/repetition_penalty_reward": -0.0938775297254324,
-      "rewards/tag_count_reward": 0.9049479365348816,
+      "completion_length": 163.390625,
+      "epoch": 0.0254968128983877,
+      "grad_norm": 1.3506574427021751,
+      "kl": 0.06396484375,
+      "learning_rate": 5.074626865671642e-06,
+      "loss": 0.0088,
+      "reward": 1.286087691783905,
+      "reward_std": 0.7203228771686554,
+      "rewards/accuracy_reward": 0.4531250223517418,
+      "rewards/reasoning_steps_reward": 0.12500000558793545,
+      "rewards/repetition_penalty_reward": -0.055058157071471214,
+      "rewards/tag_count_reward": 0.7630208432674408,
       "step": 17
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 207.55729293823242,
-      "epoch": 0.053973013493253376,
-      "grad_norm": 1.0958281725310235,
-      "kl": 0.14453125,
-      "learning_rate": 1.0588235294117648e-05,
-      "loss": -0.0121,
-      "reward": 2.4053784608840942,
-      "reward_std": 0.49531228840351105,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.8975695073604584,
-      "rewards/repetition_penalty_reward": -0.08724310249090195,
-      "rewards/tag_count_reward": 0.9492187649011612,
+      "completion_length": 178.83854293823242,
+      "epoch": 0.02699662542182227,
+      "grad_norm": 1.2350533486266992,
+      "kl": 0.0546875,
+      "learning_rate": 5.37313432835821e-06,
+      "loss": -0.0441,
+      "reward": 1.1719821691513062,
+      "reward_std": 0.7089956551790237,
+      "rewards/accuracy_reward": 0.354166679084301,
+      "rewards/reasoning_steps_reward": 0.1684027947485447,
+      "rewards/repetition_penalty_reward": -0.06282693240791559,
+      "rewards/tag_count_reward": 0.7122395932674408,
       "step": 18
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 215.67708587646484,
-      "epoch": 0.05697151424287856,
-      "grad_norm": 1.0829252560391265,
-      "kl": 0.15185546875,
-      "learning_rate": 1.1176470588235295e-05,
-      "loss": -0.0053,
-      "reward": 2.3993316292762756,
-      "reward_std": 0.49921783059835434,
-      "rewards/accuracy_reward": 0.6979166865348816,
-      "rewards/reasoning_steps_reward": 0.9097222685813904,
-      "rewards/repetition_penalty_reward": -0.08981773070991039,
-      "rewards/tag_count_reward": 0.8815104365348816,
+      "completion_length": 200.53125762939453,
+      "epoch": 0.028496437945256844,
+      "grad_norm": 1.121999866578993,
+      "kl": 0.059814453125,
+      "learning_rate": 5.671641791044776e-06,
+      "loss": -0.0504,
+      "reward": 1.23805570602417,
+      "reward_std": 0.6833942234516144,
+      "rewards/accuracy_reward": 0.3333333507180214,
+      "rewards/reasoning_steps_reward": 0.2274305708706379,
+      "rewards/repetition_penalty_reward": -0.07401033584028482,
+      "rewards/tag_count_reward": 0.7513020932674408,
       "step": 19
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 244.6354217529297,
-      "epoch": 0.05997001499250375,
-      "grad_norm": 1.1922172194368157,
-      "kl": 0.1689453125,
-      "learning_rate": 1.1764705882352942e-05,
-      "loss": 0.1106,
-      "reward": 2.2201938033103943,
-      "reward_std": 0.5487363934516907,
-      "rewards/accuracy_reward": 0.4947916716337204,
-      "rewards/reasoning_steps_reward": 0.944444477558136,
-      "rewards/repetition_penalty_reward": -0.09664654545485973,
-      "rewards/tag_count_reward": 0.8776042014360428,
+      "completion_length": 210.5833396911621,
+      "epoch": 0.029996250468691414,
+      "grad_norm": 2.277302963689603,
+      "kl": 0.07037353515625,
+      "learning_rate": 5.970149253731343e-06,
+      "loss": 0.0108,
+      "reward": 1.2767416834831238,
+      "reward_std": 0.7041554003953934,
+      "rewards/accuracy_reward": 0.3645833432674408,
+      "rewards/reasoning_steps_reward": 0.2361111268401146,
+      "rewards/repetition_penalty_reward": -0.06483818404376507,
+      "rewards/tag_count_reward": 0.7408854365348816,
       "step": 20
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 308.5833435058594,
-      "epoch": 0.06296851574212893,
-      "grad_norm": 0.9900748090272767,
-      "kl": 0.1285400390625,
-      "learning_rate": 1.235294117647059e-05,
-      "loss": 0.1992,
-      "reward": 2.4696518182754517,
-      "reward_std": 0.5737268030643463,
-      "rewards/accuracy_reward": 0.6979166865348816,
-      "rewards/reasoning_steps_reward": 0.97743059694767,
-      "rewards/repetition_penalty_reward": -0.10543505474925041,
-      "rewards/tag_count_reward": 0.899739608168602,
+      "completion_length": 206.15625381469727,
+      "epoch": 0.031496062992125984,
+      "grad_norm": 1.051251380495841,
+      "kl": 0.10986328125,
+      "learning_rate": 6.2686567164179116e-06,
+      "loss": -0.0285,
+      "reward": 1.291599839925766,
+      "reward_std": 0.6519357860088348,
+      "rewards/accuracy_reward": 0.3229166828095913,
+      "rewards/reasoning_steps_reward": 0.2812500260770321,
+      "rewards/repetition_penalty_reward": -0.07298353686928749,
+      "rewards/tag_count_reward": 0.7604166865348816,
       "step": 21
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 292.5416717529297,
-      "epoch": 0.06596701649175413,
-      "grad_norm": 0.9436079749621433,
-      "kl": 0.150146484375,
-      "learning_rate": 1.2941176470588238e-05,
-      "loss": 0.2047,
-      "reward": 2.5528025031089783,
-      "reward_std": 0.5636586248874664,
-      "rewards/accuracy_reward": 0.7447917014360428,
-      "rewards/reasoning_steps_reward": 0.9670138955116272,
-      "rewards/repetition_penalty_reward": -0.10822184756398201,
-      "rewards/tag_count_reward": 0.9492187649011612,
+      "completion_length": 248.79687881469727,
+      "epoch": 0.032995875515560553,
+      "grad_norm": 0.9759846536663065,
+      "kl": 0.0806884765625,
+      "learning_rate": 6.567164179104478e-06,
+      "loss": -0.0597,
+      "reward": 1.6020659506320953,
+      "reward_std": 0.7493701279163361,
+      "rewards/accuracy_reward": 0.4322916865348816,
+      "rewards/reasoning_steps_reward": 0.5225694701075554,
+      "rewards/repetition_penalty_reward": -0.07805563136935234,
+      "rewards/tag_count_reward": 0.7252604365348816,
       "step": 22
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 299.51563262939453,
-      "epoch": 0.06896551724137931,
-      "grad_norm": 3.8939799298403286,
-      "kl": 0.23046875,
-      "learning_rate": 1.3529411764705885e-05,
-      "loss": 0.1481,
-      "reward": 2.538209021091461,
-      "reward_std": 0.5256823599338531,
-      "rewards/accuracy_reward": 0.7135416865348816,
-      "rewards/reasoning_steps_reward": 0.9809028208255768,
-      "rewards/repetition_penalty_reward": -0.10154793784022331,
-      "rewards/tag_count_reward": 0.9453125298023224,
+      "completion_length": 262.93750762939453,
+      "epoch": 0.03449568803899512,
+      "grad_norm": 0.8719127454418464,
+      "kl": 0.0849609375,
+      "learning_rate": 6.865671641791045e-06,
+      "loss": -0.0692,
+      "reward": 1.6442042291164398,
+      "reward_std": 0.6580450236797333,
+      "rewards/accuracy_reward": 0.3645833432674408,
+      "rewards/reasoning_steps_reward": 0.614583395421505,
+      "rewards/repetition_penalty_reward": -0.09017078392207623,
+      "rewards/tag_count_reward": 0.755208358168602,
       "step": 23
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 331.1458511352539,
-      "epoch": 0.07196401799100449,
-      "grad_norm": 1.4799872075073064,
-      "kl": 0.28955078125,
-      "learning_rate": 1.4117647058823532e-05,
-      "loss": 0.2011,
-      "reward": 2.411967933177948,
-      "reward_std": 0.7274067103862762,
-      "rewards/accuracy_reward": 0.6822916716337204,
-      "rewards/reasoning_steps_reward": 0.9496527910232544,
-      "rewards/repetition_penalty_reward": -0.09758076071739197,
-      "rewards/tag_count_reward": 0.8776041716337204,
+      "completion_length": 255.6718864440918,
+      "epoch": 0.0359955005624297,
+      "grad_norm": 1.7487963120056436,
+      "kl": 0.219970703125,
+      "learning_rate": 7.164179104477612e-06,
+      "loss": -0.0599,
+      "reward": 1.8432446718215942,
+      "reward_std": 0.7495845705270767,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.6840277910232544,
+      "rewards/repetition_penalty_reward": -0.08817902393639088,
+      "rewards/tag_count_reward": 0.7473958432674408,
       "step": 24
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 348.09375762939453,
-      "epoch": 0.07496251874062969,
-      "grad_norm": 1.8716934303767792,
-      "kl": 0.21533203125,
-      "learning_rate": 1.4705882352941179e-05,
-      "loss": 0.2849,
-      "reward": 2.35833877325058,
-      "reward_std": 0.6804773062467575,
-      "rewards/accuracy_reward": 0.6093750223517418,
-      "rewards/reasoning_steps_reward": 0.9583333879709244,
-      "rewards/repetition_penalty_reward": -0.10780714452266693,
-      "rewards/tag_count_reward": 0.8984375149011612,
+      "completion_length": 292.53125762939453,
+      "epoch": 0.03749531308586427,
+      "grad_norm": 0.9847582761123604,
+      "kl": 0.1385498046875,
+      "learning_rate": 7.46268656716418e-06,
+      "loss": -0.0537,
+      "reward": 1.8635078966617584,
+      "reward_std": 0.6817265152931213,
+      "rewards/accuracy_reward": 0.3958333358168602,
+      "rewards/reasoning_steps_reward": 0.7881944626569748,
+      "rewards/repetition_penalty_reward": -0.10697820782661438,
+      "rewards/tag_count_reward": 0.7864583432674408,
       "step": 25
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 403.9166793823242,
-      "epoch": 0.07796101949025487,
-      "grad_norm": 10.274951887282484,
-      "kl": 0.6142578125,
-      "learning_rate": 1.5294117647058822e-05,
-      "loss": 0.3997,
-      "reward": 2.2658557891845703,
-      "reward_std": 0.7715227752923965,
-      "rewards/accuracy_reward": 0.6093750223517418,
-      "rewards/reasoning_steps_reward": 0.94618059694767,
-      "rewards/repetition_penalty_reward": -0.0852727573364973,
-      "rewards/tag_count_reward": 0.7955729514360428,
+      "completion_length": 291.9322967529297,
+      "epoch": 0.03899512560929884,
+      "grad_norm": 0.8461790712043904,
+      "kl": 0.1474609375,
+      "learning_rate": 7.761194029850747e-06,
+      "loss": 0.0,
+      "reward": 1.9677788615226746,
+      "reward_std": 0.653272807598114,
+      "rewards/accuracy_reward": 0.4739583432674408,
+      "rewards/reasoning_steps_reward": 0.8489583730697632,
+      "rewards/repetition_penalty_reward": -0.10774205438792706,
+      "rewards/tag_count_reward": 0.7526041865348816,
       "step": 26
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 455.1614685058594,
-      "epoch": 0.08095952023988005,
-      "grad_norm": 6.823619814725695,
-      "kl": 1.068359375,
-      "learning_rate": 1.5882352941176473e-05,
-      "loss": 0.4192,
-      "reward": 1.9341484606266022,
-      "reward_std": 0.9543884545564651,
-      "rewards/accuracy_reward": 0.4635416716337204,
-      "rewards/reasoning_steps_reward": 0.8663194328546524,
-      "rewards/repetition_penalty_reward": -0.06888982094824314,
-      "rewards/tag_count_reward": 0.673177108168602,
+      "completion_length": 271.35939025878906,
+      "epoch": 0.04049493813273341,
+      "grad_norm": 6.973505953549008,
+      "kl": 0.4765625,
+      "learning_rate": 8.059701492537314e-06,
+      "loss": 0.0103,
+      "reward": 2.150063157081604,
+      "reward_std": 0.5991375297307968,
+      "rewards/accuracy_reward": 0.5260416865348816,
+      "rewards/reasoning_steps_reward": 0.8958333730697632,
+      "rewards/repetition_penalty_reward": -0.10123893804848194,
+      "rewards/tag_count_reward": 0.829427108168602,
       "step": 27
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 369.0052185058594,
-      "epoch": 0.08395802098950525,
-      "grad_norm": 5.114037986110308,
-      "kl": 0.4541015625,
-      "learning_rate": 1.647058823529412e-05,
-      "loss": 0.3594,
-      "reward": 2.284946322441101,
-      "reward_std": 0.8626722097396851,
-      "rewards/accuracy_reward": 0.6562500149011612,
-      "rewards/reasoning_steps_reward": 0.890625,
-      "rewards/repetition_penalty_reward": -0.0783350057899952,
-      "rewards/tag_count_reward": 0.8164062798023224,
+      "completion_length": 261.28125762939453,
+      "epoch": 0.04199475065616798,
+      "grad_norm": 3.850773043534155,
+      "kl": 0.293212890625,
+      "learning_rate": 8.35820895522388e-06,
+      "loss": -0.0304,
+      "reward": 2.154005229473114,
+      "reward_std": 0.621285080909729,
+      "rewards/accuracy_reward": 0.5312500298023224,
+      "rewards/reasoning_steps_reward": 0.859375,
+      "rewards/repetition_penalty_reward": -0.1012031976133585,
+      "rewards/tag_count_reward": 0.8645833730697632,
       "step": 28
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 344.8229217529297,
-      "epoch": 0.08695652173913043,
-      "grad_norm": 1009.5340319486562,
-      "kl": 16.3125,
-      "learning_rate": 1.7058823529411767e-05,
-      "loss": 1.4554,
-      "reward": 2.209560751914978,
-      "reward_std": 0.7678176611661911,
-      "rewards/accuracy_reward": 0.5781250149011612,
-      "rewards/reasoning_steps_reward": 0.897569477558136,
-      "rewards/repetition_penalty_reward": -0.06561294477432966,
-      "rewards/tag_count_reward": 0.7994791865348816,
+      "completion_length": 229.0260467529297,
+      "epoch": 0.04349456317960255,
+      "grad_norm": 1.4645034098283287,
+      "kl": 0.23095703125,
+      "learning_rate": 8.656716417910447e-06,
+      "loss": 0.0797,
+      "reward": 2.0209468007087708,
+      "reward_std": 0.6779871582984924,
+      "rewards/accuracy_reward": 0.5885416865348816,
+      "rewards/reasoning_steps_reward": 0.7534722685813904,
+      "rewards/repetition_penalty_reward": -0.10622344352304935,
+      "rewards/tag_count_reward": 0.7851562798023224,
       "step": 29
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 356.0572967529297,
-      "epoch": 0.08995502248875563,
-      "grad_norm": 23.197023290991723,
-      "kl": 0.953125,
-      "learning_rate": 1.7647058823529414e-05,
-      "loss": 0.2222,
-      "reward": 1.9122081696987152,
-      "reward_std": 0.8025388270616531,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.8281250298023224,
-      "rewards/repetition_penalty_reward": -0.06175030395388603,
-      "rewards/tag_count_reward": 0.7083333432674408,
+      "completion_length": 232.93229293823242,
+      "epoch": 0.04499437570303712,
+      "grad_norm": 1.247184264877055,
+      "kl": 0.162841796875,
+      "learning_rate": 8.955223880597016e-06,
+      "loss": -0.0625,
+      "reward": 2.040347546339035,
+      "reward_std": 0.6728687584400177,
+      "rewards/accuracy_reward": 0.5729166865348816,
+      "rewards/reasoning_steps_reward": 0.7795138955116272,
+      "rewards/repetition_penalty_reward": -0.10374973341822624,
+      "rewards/tag_count_reward": 0.7916666865348816,
       "step": 30
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 326.0104217529297,
-      "epoch": 0.09295352323838081,
-      "grad_norm": 8.878208799628512,
-      "kl": 0.9375,
-      "learning_rate": 1.823529411764706e-05,
-      "loss": 0.1336,
-      "reward": 1.718822568655014,
-      "reward_std": 0.798391655087471,
-      "rewards/accuracy_reward": 0.3125000149011612,
-      "rewards/reasoning_steps_reward": 0.8263888955116272,
-      "rewards/repetition_penalty_reward": -0.058087206445634365,
-      "rewards/tag_count_reward": 0.638020858168602,
+      "completion_length": 248.26562881469727,
+      "epoch": 0.046494188226471694,
+      "grad_norm": 0.8484087335607122,
+      "kl": 0.172607421875,
+      "learning_rate": 9.253731343283582e-06,
+      "loss": 0.0142,
+      "reward": 2.083979547023773,
+      "reward_std": 0.5710583031177521,
+      "rewards/accuracy_reward": 0.4687500074505806,
+      "rewards/reasoning_steps_reward": 0.9479167014360428,
+      "rewards/repetition_penalty_reward": -0.10221851244568825,
+      "rewards/tag_count_reward": 0.7695312649011612,
       "step": 31
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 302.17189025878906,
-      "epoch": 0.095952023988006,
-      "grad_norm": 5.191420623812987,
-      "kl": 2.138671875,
-      "learning_rate": 1.8823529411764708e-05,
-      "loss": -0.0731,
-      "reward": 1.7180909216403961,
-      "reward_std": 0.795787900686264,
-      "rewards/accuracy_reward": 0.2604166753590107,
-      "rewards/reasoning_steps_reward": 0.8732639402151108,
-      "rewards/repetition_penalty_reward": -0.04579801578074694,
-      "rewards/tag_count_reward": 0.630208358168602,
+      "completion_length": 252.73959350585938,
+      "epoch": 0.047994000749906264,
+      "grad_norm": 1.1744348492144583,
+      "kl": 0.1875,
+      "learning_rate": 9.552238805970149e-06,
+      "loss": 0.0776,
+      "reward": 2.3965033292770386,
+      "reward_std": 0.5822078287601471,
+      "rewards/accuracy_reward": 0.6718750149011612,
+      "rewards/reasoning_steps_reward": 0.9930555671453476,
+      "rewards/repetition_penalty_reward": -0.09785447083413601,
+      "rewards/tag_count_reward": 0.8294270932674408,
       "step": 32
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 329.6666793823242,
-      "epoch": 0.09895052473763119,
-      "grad_norm": 20.842747943141333,
-      "kl": 4.875,
-      "learning_rate": 1.9411764705882355e-05,
-      "loss": -0.1732,
-      "reward": 1.663686603307724,
-      "reward_std": 0.7371216714382172,
-      "rewards/accuracy_reward": 0.2239583432674408,
-      "rewards/reasoning_steps_reward": 0.8593750298023224,
-      "rewards/repetition_penalty_reward": -0.06417796947062016,
-      "rewards/tag_count_reward": 0.6445312649011612,
+      "completion_length": 250.71875762939453,
+      "epoch": 0.049493813273340834,
+      "grad_norm": 0.9115161284883908,
+      "kl": 0.181884765625,
+      "learning_rate": 9.850746268656717e-06,
+      "loss": 0.038,
+      "reward": 2.35109943151474,
+      "reward_std": 0.5347126573324203,
+      "rewards/accuracy_reward": 0.5625000223517418,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.09681737795472145,
+      "rewards/tag_count_reward": 0.8906250298023224,
       "step": 33
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 326.96875762939453,
-      "epoch": 0.10194902548725637,
-      "grad_norm": 3.7039312187876177,
-      "kl": 4.06640625,
-      "learning_rate": 2e-05,
-      "loss": -0.3241,
-      "reward": 1.5329826474189758,
-      "reward_std": 0.797045961022377,
-      "rewards/accuracy_reward": 0.1770833358168602,
-      "rewards/reasoning_steps_reward": 0.819444477558136,
-      "rewards/repetition_penalty_reward": -0.058597257360816,
-      "rewards/tag_count_reward": 0.595052108168602,
+      "completion_length": 256.7604217529297,
+      "epoch": 0.0509936257967754,
+      "grad_norm": 0.9316669826597189,
+      "kl": 0.2060546875,
+      "learning_rate": 1.0149253731343284e-05,
+      "loss": 0.1298,
+      "reward": 2.2372400164604187,
+      "reward_std": 0.5037485882639885,
+      "rewards/accuracy_reward": 0.4427083432674408,
+      "rewards/reasoning_steps_reward": 0.989583358168602,
+      "rewards/repetition_penalty_reward": -0.09348933398723602,
+      "rewards/tag_count_reward": 0.8984375149011612,
       "step": 34
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 469.40625762939453,
-      "epoch": 0.10494752623688156,
-      "grad_norm": 19.322333158500665,
-      "kl": 2.0654296875,
-      "learning_rate": 1.9999448019954837e-05,
-      "loss": -0.1809,
-      "reward": 1.9347608387470245,
-      "reward_std": 0.8773187696933746,
-      "rewards/accuracy_reward": 0.4270833507180214,
-      "rewards/reasoning_steps_reward": 0.8923610895872116,
-      "rewards/repetition_penalty_reward": -0.11254816874861717,
-      "rewards/tag_count_reward": 0.7278645932674408,
+      "completion_length": 231.1458396911621,
+      "epoch": 0.05249343832020997,
+      "grad_norm": 0.978595163631655,
+      "kl": 0.194580078125,
+      "learning_rate": 1.0447761194029851e-05,
+      "loss": 0.0872,
+      "reward": 2.379105567932129,
+      "reward_std": 0.48672880232334137,
+      "rewards/accuracy_reward": 0.5781250223517418,
+      "rewards/reasoning_steps_reward": 0.9704861491918564,
+      "rewards/repetition_penalty_reward": -0.07315144501626492,
+      "rewards/tag_count_reward": 0.9036458432674408,
       "step": 35
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 522.7396087646484,
-      "epoch": 0.10794602698650675,
-      "grad_norm": 3.6713399077901303,
-      "kl": 0.58203125,
-      "learning_rate": 1.9997792140755746e-05,
-      "loss": -0.0915,
-      "reward": 2.4146523475646973,
-      "reward_std": 0.6919413357973099,
-      "rewards/accuracy_reward": 0.7500000149011612,
-      "rewards/reasoning_steps_reward": 0.9340278208255768,
-      "rewards/repetition_penalty_reward": -0.15739644691348076,
-      "rewards/tag_count_reward": 0.888020858168602,
+      "completion_length": 206.8645896911621,
+      "epoch": 0.05399325084364454,
+      "grad_norm": 1.001370277577607,
+      "kl": 0.2109375,
+      "learning_rate": 1.074626865671642e-05,
+      "loss": -0.0238,
+      "reward": 2.4131481647491455,
+      "reward_std": 0.5277487263083458,
+      "rewards/accuracy_reward": 0.5937500149011612,
+      "rewards/reasoning_steps_reward": 0.9531250298023224,
+      "rewards/repetition_penalty_reward": -0.07122688181698322,
+      "rewards/tag_count_reward": 0.9375000149011612,
       "step": 36
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 433.7916793823242,
-      "epoch": 0.11094452773613193,
-      "grad_norm": 2.113817065645351,
-      "kl": 0.7890625,
-      "learning_rate": 1.999503254520518e-05,
-      "loss": -0.1104,
-      "reward": 2.4119739532470703,
-      "reward_std": 0.7027525901794434,
-      "rewards/accuracy_reward": 0.7656250298023224,
-      "rewards/reasoning_steps_reward": 0.9305555820465088,
-      "rewards/repetition_penalty_reward": -0.18524829670786858,
-      "rewards/tag_count_reward": 0.9010416865348816,
+      "completion_length": 204.64583587646484,
+      "epoch": 0.05549306336707911,
+      "grad_norm": 0.9668083665156728,
+      "kl": 0.22216796875,
+      "learning_rate": 1.1044776119402986e-05,
+      "loss": -0.017,
+      "reward": 2.486513674259186,
+      "reward_std": 0.5187593251466751,
+      "rewards/accuracy_reward": 0.7031250298023224,
+      "rewards/reasoning_steps_reward": 0.9166666865348816,
+      "rewards/repetition_penalty_reward": -0.06166350655257702,
+      "rewards/tag_count_reward": 0.9283854514360428,
       "step": 37
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 304.2708435058594,
-      "epoch": 0.11394302848575712,
-      "grad_norm": 3.3062068674445064,
-      "kl": 2.06640625,
-      "learning_rate": 1.999116953795147e-05,
-      "loss": -0.1854,
-      "reward": 2.173910915851593,
-      "reward_std": 0.8623018711805344,
-      "rewards/accuracy_reward": 0.6614583507180214,
-      "rewards/reasoning_steps_reward": 0.845486119389534,
-      "rewards/repetition_penalty_reward": -0.18329406157135963,
-      "rewards/tag_count_reward": 0.8502604365348816,
+      "completion_length": 221.57812881469727,
+      "epoch": 0.05699287589051369,
+      "grad_norm": 0.9933532091220145,
+      "kl": 0.27978515625,
+      "learning_rate": 1.1343283582089553e-05,
+      "loss": 0.0533,
+      "reward": 2.315297782421112,
+      "reward_std": 0.467474602162838,
+      "rewards/accuracy_reward": 0.5156250111758709,
+      "rewards/reasoning_steps_reward": 0.9461806118488312,
+      "rewards/repetition_penalty_reward": -0.05796616990119219,
+      "rewards/tag_count_reward": 0.911458358168602,
       "step": 38
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 262.3020935058594,
-      "epoch": 0.11694152923538231,
-      "grad_norm": 2.487067537690022,
-      "kl": 1.455078125,
-      "learning_rate": 1.9986203545455205e-05,
-      "loss": -0.1053,
-      "reward": 2.481999635696411,
-      "reward_std": 0.6076074242591858,
-      "rewards/accuracy_reward": 0.8906250298023224,
-      "rewards/reasoning_steps_reward": 0.8750000447034836,
-      "rewards/repetition_penalty_reward": -0.23544833436608315,
-      "rewards/tag_count_reward": 0.9518229514360428,
+      "completion_length": 211.67708587646484,
+      "epoch": 0.05849268841394826,
+      "grad_norm": 1.1014497191190673,
+      "kl": 0.24658203125,
+      "learning_rate": 1.1641791044776121e-05,
+      "loss": -0.0132,
+      "reward": 2.4762988686561584,
+      "reward_std": 0.4729772359132767,
+      "rewards/accuracy_reward": 0.7343750298023224,
+      "rewards/reasoning_steps_reward": 0.871527835726738,
+      "rewards/repetition_penalty_reward": -0.07231233641505241,
+      "rewards/tag_count_reward": 0.9427083432674408,
       "step": 39
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 254.07813262939453,
-      "epoch": 0.1199400299850075,
-      "grad_norm": 12.184820883902836,
-      "kl": 3.974609375,
-      "learning_rate": 1.9980135115942135e-05,
-      "loss": -0.0478,
-      "reward": 2.3865994215011597,
-      "reward_std": 0.5614848285913467,
-      "rewards/accuracy_reward": 0.770833358168602,
-      "rewards/reasoning_steps_reward": 0.876736119389534,
-      "rewards/repetition_penalty_reward": -0.19586599990725517,
-      "rewards/tag_count_reward": 0.934895858168602,
+      "completion_length": 255.4895896911621,
+      "epoch": 0.05999250093738283,
+      "grad_norm": 0.876579700760602,
+      "kl": 0.27001953125,
+      "learning_rate": 1.1940298507462686e-05,
+      "loss": 0.05,
+      "reward": 2.4912882447242737,
+      "reward_std": 0.5008950978517532,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9253472834825516,
+      "rewards/repetition_penalty_reward": -0.07598620746284723,
+      "rewards/tag_count_reward": 0.9752604365348816,
       "step": 40
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 237.33334350585938,
-      "epoch": 0.12293853073463268,
-      "grad_norm": 6.144455968970083,
-      "kl": 3.984375,
-      "learning_rate": 1.9972964919342664e-05,
-      "loss": -0.0385,
-      "reward": 2.3252296447753906,
-      "reward_std": 0.7344638109207153,
-      "rewards/accuracy_reward": 0.7500000298023224,
-      "rewards/reasoning_steps_reward": 0.8819445073604584,
-      "rewards/repetition_penalty_reward": -0.1699962317943573,
-      "rewards/tag_count_reward": 0.8632812649011612,
+      "completion_length": 302.0885543823242,
+      "epoch": 0.0614923134608174,
+      "grad_norm": 1.5357965494048256,
+      "kl": 0.3916015625,
+      "learning_rate": 1.2238805970149255e-05,
+      "loss": 0.1043,
+      "reward": 2.368771195411682,
+      "reward_std": 0.578594297170639,
+      "rewards/accuracy_reward": 0.5833333507180214,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.08912823535501957,
+      "rewards/tag_count_reward": 0.9231770932674408,
       "step": 41
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 291.4791717529297,
-      "epoch": 0.12593703148425786,
-      "grad_norm": 4998.4510260355355,
-      "kl": 91.7841796875,
-      "learning_rate": 1.9964693747217873e-05,
-      "loss": 4.7855,
-      "reward": 2.487729072570801,
-      "reward_std": 0.477145679295063,
-      "rewards/accuracy_reward": 0.8697916865348816,
-      "rewards/reasoning_steps_reward": 0.9340278208255768,
-      "rewards/repetition_penalty_reward": -0.18067368865013123,
-      "rewards/tag_count_reward": 0.864583358168602,
+      "completion_length": 320.3645935058594,
+      "epoch": 0.06299212598425197,
+      "grad_norm": 1.015169753494364,
+      "kl": 0.4111328125,
+      "learning_rate": 1.2537313432835823e-05,
+      "loss": 0.0663,
+      "reward": 2.4001752734184265,
+      "reward_std": 0.6187782883644104,
+      "rewards/accuracy_reward": 0.661458358168602,
+      "rewards/reasoning_steps_reward": 0.9392361640930176,
+      "rewards/repetition_penalty_reward": -0.10286298580467701,
+      "rewards/tag_count_reward": 0.9023437649011612,
       "step": 42
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 284.6145935058594,
-      "epoch": 0.12893553223388307,
-      "grad_norm": 5.6325713646688405,
-      "kl": 1.0654296875,
-      "learning_rate": 1.9955322512672162e-05,
-      "loss": -0.064,
-      "reward": 2.384717285633087,
-      "reward_std": 0.6374871581792831,
-      "rewards/accuracy_reward": 0.8385416865348816,
-      "rewards/reasoning_steps_reward": 0.901041716337204,
-      "rewards/repetition_penalty_reward": -0.1426265835762024,
-      "rewards/tag_count_reward": 0.7877604216337204,
+      "completion_length": 327.3854293823242,
+      "epoch": 0.06449193850768654,
+      "grad_norm": 1.6703329299260001,
+      "kl": 0.4345703125,
+      "learning_rate": 1.2835820895522388e-05,
+      "loss": 0.1761,
+      "reward": 2.34128737449646,
+      "reward_std": 0.6598222628235817,
+      "rewards/accuracy_reward": 0.5937500298023224,
+      "rewards/reasoning_steps_reward": 0.9392361044883728,
+      "rewards/repetition_penalty_reward": -0.09274039790034294,
+      "rewards/tag_count_reward": 0.9010416716337204,
       "step": 43
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 353.10939025878906,
-      "epoch": 0.13193403298350825,
-      "grad_norm": 1.0900138361408922,
-      "kl": 0.6494140625,
-      "learning_rate": 1.9944852250252416e-05,
-      "loss": -0.0128,
-      "reward": 2.404056489467621,
-      "reward_std": 0.5440548211336136,
-      "rewards/accuracy_reward": 0.7447916865348816,
-      "rewards/reasoning_steps_reward": 0.8906250298023224,
-      "rewards/repetition_penalty_reward": -0.13500603288412094,
-      "rewards/tag_count_reward": 0.9036458432674408,
+      "completion_length": 308.81250762939453,
+      "epoch": 0.06599175103112111,
+      "grad_norm": 1.8118029753094556,
+      "kl": 0.40234375,
+      "learning_rate": 1.3134328358208957e-05,
+      "loss": 0.2046,
+      "reward": 2.240744113922119,
+      "reward_std": 0.6155901998281479,
+      "rewards/accuracy_reward": 0.5520833432674408,
+      "rewards/reasoning_steps_reward": 0.8923611044883728,
+      "rewards/repetition_penalty_reward": -0.08260660991072655,
+      "rewards/tag_count_reward": 0.8789062649011612,
       "step": 44
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 435.9635543823242,
-      "epoch": 0.13493253373313344,
-      "grad_norm": 447.68957414171246,
-      "kl": 15.2900390625,
-      "learning_rate": 1.993328411583383e-05,
-      "loss": 1.0339,
-      "reward": 2.6905240416526794,
-      "reward_std": 0.36615417525172234,
-      "rewards/accuracy_reward": 0.9375000298023224,
-      "rewards/reasoning_steps_reward": 0.9184028059244156,
-      "rewards/repetition_penalty_reward": -0.13022266328334808,
-      "rewards/tag_count_reward": 0.9648437649011612,
+      "completion_length": 364.7135467529297,
+      "epoch": 0.06749156355455568,
+      "grad_norm": 5.936213660745324,
+      "kl": 0.47021484375,
+      "learning_rate": 1.3432835820895525e-05,
+      "loss": 0.425,
+      "reward": 1.964487075805664,
+      "reward_std": 0.7735159993171692,
+      "rewards/accuracy_reward": 0.4583333507180214,
+      "rewards/reasoning_steps_reward": 0.8750000149011612,
+      "rewards/repetition_penalty_reward": -0.08629421889781952,
+      "rewards/tag_count_reward": 0.7174479365348816,
       "step": 45
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 583.7187805175781,
-      "epoch": 0.13793103448275862,
-      "grad_norm": 1.610843891213588,
-      "kl": 0.5166015625,
-      "learning_rate": 1.992061938649227e-05,
-      "loss": 0.0019,
-      "reward": 2.6286553740501404,
-      "reward_std": 0.3213835656642914,
-      "rewards/accuracy_reward": 0.770833358168602,
-      "rewards/reasoning_steps_reward": 0.9774305522441864,
-      "rewards/repetition_penalty_reward": -0.10788994282484055,
-      "rewards/tag_count_reward": 0.9882812649011612,
+      "completion_length": 377.81251525878906,
+      "epoch": 0.06899137607799025,
+      "grad_norm": 68.04443222745681,
+      "kl": 3.87890625,
+      "learning_rate": 1.373134328358209e-05,
+      "loss": 0.5361,
+      "reward": 1.82472363114357,
+      "reward_std": 0.8007150739431381,
+      "rewards/accuracy_reward": 0.4479166716337204,
+      "rewards/reasoning_steps_reward": 0.8715278059244156,
+      "rewards/repetition_penalty_reward": -0.07545002363622189,
+      "rewards/tag_count_reward": 0.5807291716337204,
       "step": 46
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 726.1250152587891,
-      "epoch": 0.1409295352323838,
-      "grad_norm": 7.039405614303865,
-      "kl": 0.6123046875,
-      "learning_rate": 1.9906859460363307e-05,
-      "loss": -0.0022,
-      "reward": 2.555482268333435,
-      "reward_std": 0.44716860353946686,
-      "rewards/accuracy_reward": 0.7187500074505806,
-      "rewards/reasoning_steps_reward": 0.9600694179534912,
-      "rewards/repetition_penalty_reward": -0.06474354676902294,
-      "rewards/tag_count_reward": 0.9414062649011612,
+      "completion_length": 373.68750762939453,
+      "epoch": 0.07049118860142482,
+      "grad_norm": 383.1306772268743,
+      "kl": 10.515625,
+      "learning_rate": 1.4029850746268658e-05,
+      "loss": 0.602,
+      "reward": 1.422120749950409,
+      "reward_std": 0.7384027689695358,
+      "rewards/accuracy_reward": 0.2135416716337204,
+      "rewards/reasoning_steps_reward": 0.777777835726738,
+      "rewards/repetition_penalty_reward": -0.049667539075016975,
+      "rewards/tag_count_reward": 0.4804687649011612,
       "step": 47
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 823.7031402587891,
-      "epoch": 0.14392803598200898,
-      "grad_norm": 7.8638528575359885,
-      "kl": 0.5986328125,
-      "learning_rate": 1.989200585648788e-05,
-      "loss": 0.0195,
-      "reward": 2.400337427854538,
-      "reward_std": 0.5809096917510033,
-      "rewards/accuracy_reward": 0.7031250223517418,
-      "rewards/reasoning_steps_reward": 0.9097222238779068,
-      "rewards/repetition_penalty_reward": -0.05625995807349682,
-      "rewards/tag_count_reward": 0.8437500149011612,
+      "completion_length": 383.4479293823242,
+      "epoch": 0.0719910011248594,
+      "grad_norm": 127.08044008975428,
+      "kl": 78.771484375,
+      "learning_rate": 1.4328358208955224e-05,
+      "loss": 0.5014,
+      "reward": 1.5611045956611633,
+      "reward_std": 0.8897948116064072,
+      "rewards/accuracy_reward": 0.3072916716337204,
+      "rewards/reasoning_steps_reward": 0.751736119389534,
+      "rewards/repetition_penalty_reward": -0.05781908147037029,
+      "rewards/tag_count_reward": 0.559895858168602,
       "step": 48
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 958.1198120117188,
-      "epoch": 0.1469265367316342,
-      "grad_norm": 3.4063309297902227,
-      "kl": 0.56640625,
-      "learning_rate": 1.9876060214644568e-05,
-      "loss": 0.0231,
-      "reward": 1.8466069400310516,
-      "reward_std": 0.7590629458427429,
-      "rewards/accuracy_reward": 0.432291679084301,
-      "rewards/reasoning_steps_reward": 0.8385417014360428,
-      "rewards/repetition_penalty_reward": -0.034903590101748705,
-      "rewards/tag_count_reward": 0.6106770932674408,
+      "completion_length": 343.1770935058594,
+      "epoch": 0.07349081364829396,
+      "grad_norm": 8.004059356580026,
+      "kl": 1.6875,
+      "learning_rate": 1.4626865671641792e-05,
+      "loss": 0.2767,
+      "reward": 1.384813278913498,
+      "reward_std": 0.8601708263158798,
+      "rewards/accuracy_reward": 0.2187500037252903,
+      "rewards/reasoning_steps_reward": 0.7118055671453476,
+      "rewards/repetition_penalty_reward": -0.06787777040153742,
+      "rewards/tag_count_reward": 0.5221354439854622,
       "step": 49
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 996.1250305175781,
-      "epoch": 0.14992503748125938,
-      "grad_norm": 3.75487802054811,
-      "kl": 0.9189453125,
-      "learning_rate": 1.9859024295168593e-05,
-      "loss": 0.0443,
-      "reward": 1.636008232831955,
-      "reward_std": 0.7502989023923874,
-      "rewards/accuracy_reward": 0.3906250149011612,
-      "rewards/reasoning_steps_reward": 0.788194477558136,
-      "rewards/repetition_penalty_reward": -0.0336966784670949,
-      "rewards/tag_count_reward": 0.4908854439854622,
+      "completion_length": 351.7135543823242,
+      "epoch": 0.07499062617172854,
+      "grad_norm": 1.7830772174928224,
+      "kl": 1.0224609375,
+      "learning_rate": 1.492537313432836e-05,
+      "loss": 0.1119,
+      "reward": 1.2045287191867828,
+      "reward_std": 0.8166554719209671,
+      "rewards/accuracy_reward": 0.1666666716337204,
+      "rewards/reasoning_steps_reward": 0.6302083432674408,
+      "rewards/repetition_penalty_reward": -0.06630465760827065,
+      "rewards/tag_count_reward": 0.4739583432674408,
       "step": 50
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 1002.0833435058594,
-      "epoch": 0.15292353823088456,
-      "grad_norm": 26.36570320352729,
-      "kl": 2.140625,
-      "learning_rate": 1.9840899978757483e-05,
-      "loss": 0.0805,
-      "reward": 1.6898488700389862,
-      "reward_std": 0.7471358329057693,
-      "rewards/accuracy_reward": 0.3489583432674408,
-      "rewards/reasoning_steps_reward": 0.8315972834825516,
-      "rewards/repetition_penalty_reward": -0.046696340665221214,
-      "rewards/tag_count_reward": 0.5559896007180214,
+      "completion_length": 410.0677185058594,
+      "epoch": 0.0764904386951631,
+      "grad_norm": 1.780458421762445,
+      "kl": 1.1845703125,
+      "learning_rate": 1.5223880597014925e-05,
+      "loss": -0.0193,
+      "reward": 1.3819158375263214,
+      "reward_std": 0.8455617725849152,
+      "rewards/accuracy_reward": 0.2708333395421505,
+      "rewards/reasoning_steps_reward": 0.7777777910232544,
+      "rewards/repetition_penalty_reward": -0.13153912127017975,
+      "rewards/tag_count_reward": 0.4648437649011612,
       "step": 51
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 990.2968902587891,
-      "epoch": 0.15592203898050974,
-      "grad_norm": 29.24774607573346,
-      "kl": 6.390625,
-      "learning_rate": 1.9821689266263425e-05,
-      "loss": 0.1745,
-      "reward": 1.9025542736053467,
-      "reward_std": 0.8440426588058472,
-      "rewards/accuracy_reward": 0.5052083507180214,
-      "rewards/reasoning_steps_reward": 0.7934027910232544,
-      "rewards/repetition_penalty_reward": -0.06272357050329447,
-      "rewards/tag_count_reward": 0.6666666716337204,
+      "completion_length": 583.6302185058594,
+      "epoch": 0.07799025121859768,
+      "grad_norm": 0.8595799417367336,
+      "kl": 0.94970703125,
+      "learning_rate": 1.5522388059701494e-05,
+      "loss": -0.1912,
+      "reward": 1.392962396144867,
+      "reward_std": 0.8328078836202621,
+      "rewards/accuracy_reward": 0.31250000558793545,
+      "rewards/reasoning_steps_reward": 0.8055555820465088,
+      "rewards/repetition_penalty_reward": -0.19254108518362045,
+      "rewards/tag_count_reward": 0.4674479365348816,
       "step": 52
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 564.7864761352539,
-      "epoch": 0.15892053973013492,
-      "grad_norm": 21.634969699176207,
-      "kl": 1.97265625,
-      "learning_rate": 1.980139427847242e-05,
-      "loss": 0.6096,
-      "reward": 2.039624661207199,
-      "reward_std": 0.7389360666275024,
-      "rewards/accuracy_reward": 0.5468750074505806,
-      "rewards/reasoning_steps_reward": 0.8038195073604584,
-      "rewards/repetition_penalty_reward": -0.05976778268814087,
-      "rewards/tag_count_reward": 0.7486979365348816,
+      "completion_length": 761.7812805175781,
+      "epoch": 0.07949006374203224,
+      "grad_norm": 1.9426702160291227,
+      "kl": 0.73828125,
+      "learning_rate": 1.582089552238806e-05,
+      "loss": -0.0632,
+      "reward": 1.2566802203655243,
+      "reward_std": 0.7984166294336319,
+      "rewards/accuracy_reward": 0.2864583395421505,
+      "rewards/reasoning_steps_reward": 0.8125000149011612,
+      "rewards/repetition_penalty_reward": -0.2914969325065613,
+      "rewards/tag_count_reward": 0.4492187574505806,
       "step": 53
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 626.5000152587891,
-      "epoch": 0.1619190404797601,
-      "grad_norm": 20.599440681644708,
-      "kl": 1.03125,
-      "learning_rate": 1.9780017255870114e-05,
-      "loss": 0.5382,
-      "reward": 1.7361580431461334,
-      "reward_std": 0.7883689254522324,
-      "rewards/accuracy_reward": 0.3593750074505806,
-      "rewards/reasoning_steps_reward": 0.7899305671453476,
-      "rewards/repetition_penalty_reward": -0.047262136824429035,
-      "rewards/tag_count_reward": 0.634114608168602,
+      "completion_length": 769.3333587646484,
+      "epoch": 0.08098987626546682,
+      "grad_norm": 0.606126420750735,
+      "kl": 0.58154296875,
+      "learning_rate": 1.6119402985074627e-05,
+      "loss": -0.0516,
+      "reward": 1.2068730890750885,
+      "reward_std": 0.5995621234178543,
+      "rewards/accuracy_reward": 0.17708333861082792,
+      "rewards/reasoning_steps_reward": 0.8819444924592972,
+      "rewards/repetition_penalty_reward": -0.34304021298885345,
+      "rewards/tag_count_reward": 0.490885429084301,
       "step": 54
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 813.6041717529297,
-      "epoch": 0.16491754122938532,
-      "grad_norm": 14.859517097725304,
-      "kl": 2.419921875,
-      "learning_rate": 1.9757560558394493e-05,
-      "loss": 0.3132,
-      "reward": 1.3911511600017548,
-      "reward_std": 0.7518036961555481,
-      "rewards/accuracy_reward": 0.2187500074505806,
-      "rewards/reasoning_steps_reward": 0.6961806118488312,
-      "rewards/repetition_penalty_reward": -0.026383677031844854,
-      "rewards/tag_count_reward": 0.5026041865348816,
+      "completion_length": 632.9479370117188,
+      "epoch": 0.0824896887889014,
+      "grad_norm": 141.4492030747302,
+      "kl": 1.0986328125,
+      "learning_rate": 1.6417910447761197e-05,
+      "loss": -0.0583,
+      "reward": 1.5605631172657013,
+      "reward_std": 0.6802570223808289,
+      "rewards/accuracy_reward": 0.4010416716337204,
+      "rewards/reasoning_steps_reward": 0.88368059694767,
+      "rewards/repetition_penalty_reward": -0.2957737147808075,
+      "rewards/tag_count_reward": 0.5716145932674408,
       "step": 55
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 910.8177337646484,
-      "epoch": 0.1679160419790105,
-      "grad_norm": 6.499919813214984,
-      "kl": 0.9853515625,
-      "learning_rate": 1.9734026665175335e-05,
-      "loss": 0.0658,
-      "reward": 1.1412139385938644,
-      "reward_std": 0.6940838098526001,
-      "rewards/accuracy_reward": 0.1197916679084301,
-      "rewards/reasoning_steps_reward": 0.6545139253139496,
-      "rewards/repetition_penalty_reward": -0.0172062402125448,
-      "rewards/tag_count_reward": 0.3841145858168602,
+      "completion_length": 510.9791793823242,
+      "epoch": 0.08398950131233596,
+      "grad_norm": 0.9471168937555826,
+      "kl": 0.603515625,
+      "learning_rate": 1.671641791044776e-05,
+      "loss": -0.0603,
+      "reward": 1.6753197610378265,
+      "reward_std": 0.5637442171573639,
+      "rewards/accuracy_reward": 0.3958333507180214,
+      "rewards/reasoning_steps_reward": 0.9253472238779068,
+      "rewards/repetition_penalty_reward": -0.22789209336042404,
+      "rewards/tag_count_reward": 0.5820312798023224,
       "step": 56
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 949.7083587646484,
-      "epoch": 0.17091454272863568,
-      "grad_norm": 1.0108146959506905,
-      "kl": 0.8154296875,
-      "learning_rate": 1.9709418174260523e-05,
-      "loss": 0.0444,
-      "reward": 1.0093085765838623,
-      "reward_std": 0.6726640909910202,
-      "rewards/accuracy_reward": 0.07812500186264515,
-      "rewards/reasoning_steps_reward": 0.5972222685813904,
-      "rewards/repetition_penalty_reward": -0.014997066929936409,
-      "rewards/tag_count_reward": 0.3489583432674408,
+      "completion_length": 433.4479293823242,
+      "epoch": 0.08548931383577053,
+      "grad_norm": 12471.246658659455,
+      "kl": 123.4453125,
+      "learning_rate": 1.701492537313433e-05,
+      "loss": 7.2604,
+      "reward": 1.5392451286315918,
+      "reward_std": 0.656680166721344,
+      "rewards/accuracy_reward": 0.3645833469927311,
+      "rewards/reasoning_steps_reward": 0.84375,
+      "rewards/repetition_penalty_reward": -0.2185674048960209,
+      "rewards/tag_count_reward": 0.5494791865348816,
       "step": 57
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 923.5104370117188,
-      "epoch": 0.17391304347826086,
-      "grad_norm": 2.2827458536380156,
-      "kl": 0.9248046875,
-      "learning_rate": 1.9683737802329242e-05,
-      "loss": 0.0701,
-      "reward": 1.2827005088329315,
-      "reward_std": 0.6604696810245514,
-      "rewards/accuracy_reward": 0.11458333721384406,
-      "rewards/reasoning_steps_reward": 0.7690972834825516,
-      "rewards/repetition_penalty_reward": -0.021553035592660308,
-      "rewards/tag_count_reward": 0.4205729216337204,
+      "completion_length": 334.6198043823242,
+      "epoch": 0.0869891263592051,
+      "grad_norm": 169.80984697224983,
+      "kl": 123.8564453125,
+      "learning_rate": 1.7313432835820894e-05,
+      "loss": 0.1593,
+      "reward": 1.3166283071041107,
+      "reward_std": 0.6126263588666916,
+      "rewards/accuracy_reward": 0.1979166716337204,
+      "rewards/reasoning_steps_reward": 0.819444477558136,
+      "rewards/repetition_penalty_reward": -0.1720869541168213,
+      "rewards/tag_count_reward": 0.471354179084301,
       "step": 58
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 832.3229370117188,
-      "epoch": 0.17691154422788605,
-      "grad_norm": 11.547641471792147,
-      "kl": 1.482421875,
-      "learning_rate": 1.9656988384392075e-05,
-      "loss": 0.2098,
-      "reward": 1.5504461526870728,
-      "reward_std": 0.5935858860611916,
-      "rewards/accuracy_reward": 0.16666667256504297,
-      "rewards/reasoning_steps_reward": 0.855902835726738,
-      "rewards/repetition_penalty_reward": -0.029415032360702753,
-      "rewards/tag_count_reward": 0.5572916865348816,
+      "completion_length": 297.93750762939453,
+      "epoch": 0.08848893888263967,
+      "grad_norm": 9.047081897906422,
+      "kl": 2.00390625,
+      "learning_rate": 1.7611940298507464e-05,
+      "loss": -0.1778,
+      "reward": 1.1133750975131989,
+      "reward_std": 0.5235892608761787,
+      "rewards/accuracy_reward": 0.06770833488553762,
+      "rewards/reasoning_steps_reward": 0.8072916865348816,
+      "rewards/repetition_penalty_reward": -0.13141660019755363,
+      "rewards/tag_count_reward": 0.369791679084301,
       "step": 59
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 811.9479217529297,
-      "epoch": 0.17991004497751126,
-      "grad_norm": 7.13025211784621,
-      "kl": 2.056640625,
-      "learning_rate": 1.9629172873477995e-05,
-      "loss": 0.1859,
-      "reward": 1.8182637393474579,
-      "reward_std": 0.7368517369031906,
-      "rewards/accuracy_reward": 0.3593750111758709,
-      "rewards/reasoning_steps_reward": 0.8611111044883728,
-      "rewards/repetition_penalty_reward": -0.0441494369879365,
-      "rewards/tag_count_reward": 0.641927108168602,
+      "completion_length": 355.6614761352539,
+      "epoch": 0.08998875140607424,
+      "grad_norm": 8.682581929106801,
+      "kl": 1.0380859375,
+      "learning_rate": 1.791044776119403e-05,
+      "loss": -0.1428,
+      "reward": 1.3278673589229584,
+      "reward_std": 0.5868176072835922,
+      "rewards/accuracy_reward": 0.2031250074505806,
+      "rewards/reasoning_steps_reward": 0.824652835726738,
+      "rewards/repetition_penalty_reward": -0.19079595804214478,
+      "rewards/tag_count_reward": 0.490885429084301,
       "step": 60
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 754.9479370117188,
-      "epoch": 0.18290854572713644,
-      "grad_norm": 8.766884713406547,
-      "kl": 1.634765625,
-      "learning_rate": 1.96002943403084e-05,
-      "loss": 0.2092,
-      "reward": 2.166029632091522,
-      "reward_std": 0.7226460427045822,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.9079861789941788,
-      "rewards/repetition_penalty_reward": -0.06487324088811874,
-      "rewards/tag_count_reward": 0.6770833730697632,
+      "completion_length": 389.0104217529297,
+      "epoch": 0.09148856392950881,
+      "grad_norm": 2.638836671205147,
+      "kl": 0.8349609375,
+      "learning_rate": 1.8208955223880598e-05,
+      "loss": -0.0578,
+      "reward": 1.3582229912281036,
+      "reward_std": 0.5021841153502464,
+      "rewards/accuracy_reward": 0.2187500074505806,
+      "rewards/reasoning_steps_reward": 0.862847238779068,
+      "rewards/repetition_penalty_reward": -0.26764511317014694,
+      "rewards/tag_count_reward": 0.5442708432674408,
       "step": 61
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 641.5833587646484,
-      "epoch": 0.18590704647676162,
-      "grad_norm": 5.967296694311895,
-      "kl": 1.23828125,
-      "learning_rate": 1.9570355972958098e-05,
-      "loss": 0.2609,
-      "reward": 1.9569042026996613,
-      "reward_std": 0.6188908591866493,
-      "rewards/accuracy_reward": 0.4114583432674408,
-      "rewards/reasoning_steps_reward": 0.9444444626569748,
-      "rewards/repetition_penalty_reward": -0.07347787916660309,
-      "rewards/tag_count_reward": 0.6744792014360428,
+      "completion_length": 338.09376525878906,
+      "epoch": 0.09298837645294339,
+      "grad_norm": 3.079165996915571,
+      "kl": 0.822265625,
+      "learning_rate": 1.8507462686567165e-05,
+      "loss": -0.0992,
+      "reward": 1.3493062555789948,
+      "reward_std": 0.4447471499443054,
+      "rewards/accuracy_reward": 0.15104167303070426,
+      "rewards/reasoning_steps_reward": 0.8923611491918564,
+      "rewards/repetition_penalty_reward": -0.22144031897187233,
+      "rewards/tag_count_reward": 0.5273437649011612,
       "step": 62
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 655.6041870117188,
-      "epoch": 0.1889055472263868,
-      "grad_norm": 8.21428152429746,
-      "kl": 1.9296875,
-      "learning_rate": 1.953936107650336e-05,
-      "loss": 0.2559,
-      "reward": 1.7195740938186646,
-      "reward_std": 0.5279941856861115,
-      "rewards/accuracy_reward": 0.2187500037252903,
-      "rewards/reasoning_steps_reward": 0.9461805820465088,
-      "rewards/repetition_penalty_reward": -0.10290857404470444,
-      "rewards/tag_count_reward": 0.657552108168602,
+      "completion_length": 354.5,
+      "epoch": 0.09448818897637795,
+      "grad_norm": 4.4024040515363145,
+      "kl": 0.748046875,
+      "learning_rate": 1.8805970149253735e-05,
+      "loss": -0.0358,
+      "reward": 1.5263647735118866,
+      "reward_std": 0.6498845219612122,
+      "rewards/accuracy_reward": 0.3229166753590107,
+      "rewards/reasoning_steps_reward": 0.888888955116272,
+      "rewards/repetition_penalty_reward": -0.2323157861828804,
+      "rewards/tag_count_reward": 0.5468750149011612,
       "step": 63
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 768.5729217529297,
-      "epoch": 0.191904047976012,
-      "grad_norm": 4.853080558533273,
-      "kl": 3.99609375,
-      "learning_rate": 1.9507313072657057e-05,
-      "loss": 0.2929,
-      "reward": 1.5277346670627594,
-      "reward_std": 0.40065091848373413,
-      "rewards/accuracy_reward": 0.0677083358168602,
-      "rewards/reasoning_steps_reward": 0.9513889253139496,
-      "rewards/repetition_penalty_reward": -0.15542514994740486,
-      "rewards/tag_count_reward": 0.6640625149011612,
+      "completion_length": 343.4166717529297,
+      "epoch": 0.09598800149981253,
+      "grad_norm": 14.853099160994677,
+      "kl": 0.8349609375,
+      "learning_rate": 1.9104477611940298e-05,
+      "loss": -0.1063,
+      "reward": 1.2628207504749298,
+      "reward_std": 0.58592738956213,
+      "rewards/accuracy_reward": 0.17187500651925802,
+      "rewards/reasoning_steps_reward": 0.8315972536802292,
+      "rewards/repetition_penalty_reward": -0.20809948071837425,
+      "rewards/tag_count_reward": 0.4674479216337204,
       "step": 64
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 745.9583587646484,
-      "epoch": 0.19490254872563717,
-      "grad_norm": 2.314098183884576,
-      "kl": 3.66015625,
-      "learning_rate": 1.9474215499390912e-05,
-      "loss": 0.3051,
-      "reward": 1.5480964481830597,
-      "reward_std": 0.35690316557884216,
-      "rewards/accuracy_reward": 0.015625000465661287,
-      "rewards/reasoning_steps_reward": 0.9253472238779068,
-      "rewards/repetition_penalty_reward": -0.1090216189622879,
-      "rewards/tag_count_reward": 0.716145858168602,
+      "completion_length": 316.5364685058594,
+      "epoch": 0.09748781402324709,
+      "grad_norm": 2.3731241908019913,
+      "kl": 0.974609375,
+      "learning_rate": 1.9402985074626868e-05,
+      "loss": -0.1689,
+      "reward": 1.2171970903873444,
+      "reward_std": 0.6599665582180023,
+      "rewards/accuracy_reward": 0.18750000931322575,
+      "rewards/reasoning_steps_reward": 0.8020833730697632,
+      "rewards/repetition_penalty_reward": -0.21509456261992455,
+      "rewards/tag_count_reward": 0.4427083432674408,
       "step": 65
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 594.9427261352539,
-      "epoch": 0.19790104947526238,
-      "grad_norm": 2.348236466880056,
-      "kl": 2.451171875,
-      "learning_rate": 1.944007201054492e-05,
-      "loss": 0.3924,
-      "reward": 1.5447391867637634,
-      "reward_std": 0.509847991168499,
-      "rewards/accuracy_reward": 0.03645833441987634,
-      "rewards/reasoning_steps_reward": 0.8750000149011612,
-      "rewards/repetition_penalty_reward": -0.10630252212285995,
-      "rewards/tag_count_reward": 0.739583358168602,
+      "completion_length": 365.17708587646484,
+      "epoch": 0.09898762654668167,
+      "grad_norm": 1.1098635397425487,
+      "kl": 0.8447265625,
+      "learning_rate": 1.9701492537313435e-05,
+      "loss": -0.0102,
+      "reward": 1.2291882634162903,
+      "reward_std": 0.40925049781799316,
+      "rewards/accuracy_reward": 0.06770833395421505,
+      "rewards/reasoning_steps_reward": 0.942708358168602,
+      "rewards/repetition_penalty_reward": -0.2408638447523117,
+      "rewards/tag_count_reward": 0.459635429084301,
       "step": 66
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 620.7760620117188,
-      "epoch": 0.20089955022488756,
-      "grad_norm": 0.6607962588027059,
-      "kl": 4.9921875,
-      "learning_rate": 1.9404886375423982e-05,
-      "loss": 0.3226,
-      "reward": 1.5875399112701416,
-      "reward_std": 0.30489787086844444,
-      "rewards/accuracy_reward": 0.015625000465661287,
-      "rewards/reasoning_steps_reward": 0.9479166865348816,
-      "rewards/repetition_penalty_reward": -0.11298095621168613,
-      "rewards/tag_count_reward": 0.7369792014360428,
+      "completion_length": 352.6770935058594,
+      "epoch": 0.10048743907011623,
+      "grad_norm": 0.9724287484197509,
+      "kl": 0.814453125,
+      "learning_rate": 2e-05,
+      "loss": -0.068,
+      "reward": 1.5917953252792358,
+      "reward_std": 0.6212150007486343,
+      "rewards/accuracy_reward": 0.3645833469927311,
+      "rewards/reasoning_steps_reward": 0.9583333432674408,
+      "rewards/repetition_penalty_reward": -0.24153802171349525,
+      "rewards/tag_count_reward": 0.5104166865348816,
       "step": 67
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 544.3229293823242,
-      "epoch": 0.20389805097451275,
-      "grad_norm": 1.2843597560384168,
-      "kl": 4.6953125,
-      "learning_rate": 1.93686624783818e-05,
-      "loss": 0.2727,
-      "reward": 1.6841119825839996,
-      "reward_std": 0.3194814845919609,
-      "rewards/accuracy_reward": 0.03125000046566129,
-      "rewards/reasoning_steps_reward": 0.9565972238779068,
-      "rewards/repetition_penalty_reward": -0.11493314802646637,
-      "rewards/tag_count_reward": 0.8111979514360428,
+      "completion_length": 379.45314025878906,
+      "epoch": 0.1019872515935508,
+      "grad_norm": 0.8395794709698925,
+      "kl": 0.4814453125,
+      "learning_rate": 1.9999862464405377e-05,
+      "loss": -0.0332,
+      "reward": 1.6506008505821228,
+      "reward_std": 0.4853805750608444,
+      "rewards/accuracy_reward": 0.3489583358168602,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.29254162311553955,
+      "rewards/tag_count_reward": 0.6080729365348816,
       "step": 68
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 547.8958435058594,
-      "epoch": 0.20689655172413793,
-      "grad_norm": 0.709620329405288,
-      "kl": 4.9921875,
-      "learning_rate": 1.9331404318392028e-05,
-      "loss": 0.2179,
-      "reward": 1.8003253638744354,
-      "reward_std": 0.37343159317970276,
-      "rewards/accuracy_reward": 0.08854166837409139,
-      "rewards/reasoning_steps_reward": 0.9565972685813904,
-      "rewards/repetition_penalty_reward": -0.1015844214707613,
-      "rewards/tag_count_reward": 0.856770858168602,
+      "completion_length": 406.4166793823242,
+      "epoch": 0.10348706411698538,
+      "grad_norm": 1.5362392749572744,
+      "kl": 0.56005859375,
+      "learning_rate": 1.9999449861404716e-05,
+      "loss": -0.0482,
+      "reward": 1.6696705520153046,
+      "reward_std": 0.5340171456336975,
+      "rewards/accuracy_reward": 0.416666679084301,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.3342357203364372,
+      "rewards/tag_count_reward": 0.602864608168602,
       "step": 69
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 495.00000762939453,
-      "epoch": 0.2098950524737631,
-      "grad_norm": 0.7619813319396578,
-      "kl": 4.671875,
-      "learning_rate": 1.9293116008606838e-05,
-      "loss": 0.188,
-      "reward": 1.8160299956798553,
-      "reward_std": 0.27780742943286896,
-      "rewards/accuracy_reward": 0.046875000931322575,
-      "rewards/reasoning_steps_reward": 0.9739583432674408,
-      "rewards/repetition_penalty_reward": -0.1188659518957138,
-      "rewards/tag_count_reward": 0.9140625298023224,
+      "completion_length": 373.52083587646484,
+      "epoch": 0.10498687664041995,
+      "grad_norm": 1.5205558305374207,
+      "kl": 0.63134765625,
+      "learning_rate": 1.999876220234753e-05,
+      "loss": -0.0912,
+      "reward": 1.6751901507377625,
+      "reward_std": 0.4232308343052864,
+      "rewards/accuracy_reward": 0.36458334140479565,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.26100781559944153,
+      "rewards/tag_count_reward": 0.5872395932674408,
       "step": 70
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 646.7083511352539,
-      "epoch": 0.2128935532233883,
-      "grad_norm": 0.7272889048910283,
-      "kl": 5.125,
-      "learning_rate": 1.925380177590282e-05,
-      "loss": 0.212,
-      "reward": 1.8391913771629333,
-      "reward_std": 0.2926352843642235,
-      "rewards/accuracy_reward": 0.03125000046566129,
-      "rewards/reasoning_steps_reward": 0.9635417014360428,
-      "rewards/repetition_penalty_reward": -0.09830864518880844,
-      "rewards/tag_count_reward": 0.9427083730697632,
+      "completion_length": 632.9062805175781,
+      "epoch": 0.10648668916385452,
+      "grad_norm": 0.41857006484500325,
+      "kl": 0.51123046875,
+      "learning_rate": 1.999779950614934e-05,
+      "loss": -0.0632,
+      "reward": 1.9554068744182587,
+      "reward_std": 0.36940491758286953,
+      "rewards/accuracy_reward": 0.6510416865348816,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.3766243904829025,
+      "rewards/tag_count_reward": 0.696614608168602,
       "step": 71
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 691.5677108764648,
-      "epoch": 0.2158920539730135,
-      "grad_norm": 0.5588946822791705,
-      "kl": 5.1640625,
-      "learning_rate": 1.921346596041437e-05,
-      "loss": 0.2451,
-      "reward": 1.953030288219452,
-      "reward_std": 0.3250604011118412,
-      "rewards/accuracy_reward": 0.09895833348855376,
-      "rewards/reasoning_steps_reward": 0.9583333879709244,
-      "rewards/repetition_penalty_reward": -0.061292664147913456,
-      "rewards/tag_count_reward": 0.9570312798023224,
+      "completion_length": 740.0573120117188,
+      "epoch": 0.10798650168728909,
+      "grad_norm": 0.9159052896835216,
+      "kl": 1.2783203125,
+      "learning_rate": 1.999656179929115e-05,
+      "loss": -0.0142,
+      "reward": 1.6608175337314606,
+      "reward_std": 0.33733372390270233,
+      "rewards/accuracy_reward": 0.338541679084301,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.3639221265912056,
+      "rewards/tag_count_reward": 0.696614608168602,
       "step": 72
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 690.0364837646484,
-      "epoch": 0.21889055472263869,
-      "grad_norm": 0.5154663707678571,
-      "kl": 4.48828125,
-      "learning_rate": 1.917211301505453e-05,
-      "loss": 0.2857,
-      "reward": 1.984506070613861,
-      "reward_std": 0.4579034373164177,
-      "rewards/accuracy_reward": 0.2083333395421505,
-      "rewards/reasoning_steps_reward": 0.8940972983837128,
-      "rewards/repetition_penalty_reward": -0.043705823831260204,
-      "rewards/tag_count_reward": 0.9257812798023224,
+      "completion_length": 640.7708435058594,
+      "epoch": 0.10948631421072366,
+      "grad_norm": 0.6371368439400827,
+      "kl": 0.52587890625,
+      "learning_rate": 1.9995049115818706e-05,
+      "loss": -0.0003,
+      "reward": 2.032588928937912,
+      "reward_std": 0.3413342162966728,
+      "rewards/accuracy_reward": 0.5989583432674408,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.2942340075969696,
+      "rewards/tag_count_reward": 0.7330729365348816,
       "step": 73
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 421.1666793823242,
-      "epoch": 0.22188905547226387,
-      "grad_norm": 0.7720729416463754,
-      "kl": 2.33984375,
-      "learning_rate": 1.9129747505023438e-05,
-      "loss": 0.2752,
-      "reward": 2.0761736631393433,
-      "reward_std": 0.5404656231403351,
-      "rewards/accuracy_reward": 0.27604167349636555,
-      "rewards/reasoning_steps_reward": 0.8906250894069672,
-      "rewards/repetition_penalty_reward": -0.041013902984559536,
-      "rewards/tag_count_reward": 0.950520858168602,
+      "completion_length": 540.7708435058594,
+      "epoch": 0.11098612673415822,
+      "grad_norm": 0.48623664555771756,
+      "kl": 0.60498046875,
+      "learning_rate": 1.9993261497341575e-05,
+      "loss": -0.0128,
+      "reward": 2.286080002784729,
+      "reward_std": 0.3919960707426071,
+      "rewards/accuracy_reward": 0.8020833730697632,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.2516804449260235,
+      "rewards/tag_count_reward": 0.7408854365348816,
       "step": 74
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 286.5260429382324,
-      "epoch": 0.22488755622188905,
-      "grad_norm": 0.8315806740092944,
-      "kl": 1.33984375,
-      "learning_rate": 1.9086374107304312e-05,
-      "loss": 0.2576,
-      "reward": 2.031211197376251,
-      "reward_std": 0.42878295481204987,
-      "rewards/accuracy_reward": 0.18229167303070426,
-      "rewards/reasoning_steps_reward": 0.9322916269302368,
-      "rewards/repetition_penalty_reward": -0.03389316704124212,
-      "rewards/tag_count_reward": 0.950520858168602,
+      "completion_length": 480.89064025878906,
+      "epoch": 0.1124859392575928,
+      "grad_norm": 0.596999402158007,
+      "kl": 0.3017578125,
+      "learning_rate": 1.9991198993031992e-05,
+      "loss": -0.0083,
+      "reward": 2.0890829265117645,
+      "reward_std": 0.5263374149799347,
+      "rewards/accuracy_reward": 0.5781250074505806,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.20649003237485886,
+      "rewards/tag_count_reward": 0.7330729216337204,
       "step": 75
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 235.10938262939453,
-      "epoch": 0.22788605697151423,
-      "grad_norm": 0.9080987838709209,
-      "kl": 0.7822265625,
-      "learning_rate": 1.9041997610147166e-05,
-      "loss": 0.1866,
-      "reward": 2.127915918827057,
-      "reward_std": 0.4003848433494568,
-      "rewards/accuracy_reward": 0.244791679084301,
-      "rewards/reasoning_steps_reward": 0.9739583879709244,
-      "rewards/repetition_penalty_reward": -0.0491675129160285,
-      "rewards/tag_count_reward": 0.9583333432674408,
+      "completion_length": 388.0833511352539,
+      "epoch": 0.11398575178102738,
+      "grad_norm": 7.4143713553392665,
+      "kl": 1.03271484375,
+      "learning_rate": 1.99888616596235e-05,
+      "loss": -0.019,
+      "reward": 2.0290270149707794,
+      "reward_std": 0.5157302021980286,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.16845569387078285,
+      "rewards/tag_count_reward": 0.7096354365348816,
       "step": 76
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 257.5625114440918,
-      "epoch": 0.23088455772113944,
-      "grad_norm": 1.013137501122327,
-      "kl": 0.517578125,
-      "learning_rate": 1.8996622912540182e-05,
-      "loss": 0.1469,
-      "reward": 2.171510338783264,
-      "reward_std": 0.43282945454120636,
-      "rewards/accuracy_reward": 0.260416679084301,
-      "rewards/reasoning_steps_reward": 0.9947916567325592,
-      "rewards/repetition_penalty_reward": -0.048541837371885777,
-      "rewards/tag_count_reward": 0.9648437798023224,
+      "completion_length": 372.8698043823242,
+      "epoch": 0.11548556430446194,
+      "grad_norm": 1.6227089625901896,
+      "kl": 0.443359375,
+      "learning_rate": 1.9986249561409415e-05,
+      "loss": -0.1403,
+      "reward": 2.0930177569389343,
+      "reward_std": 0.5763500481843948,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9843750298023224,
+      "rewards/repetition_penalty_reward": -0.14135734736919403,
+      "rewards/tag_count_reward": 0.6666666716337204,
       "step": 77
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 375.3698043823242,
-      "epoch": 0.23388305847076463,
-      "grad_norm": 0.8031277081942563,
-      "kl": 0.859375,
-      "learning_rate": 1.8950255023668876e-05,
-      "loss": 0.4293,
-      "reward": 2.1424886882305145,
-      "reward_std": 0.5950686857104301,
-      "rewards/accuracy_reward": 0.354166679084301,
-      "rewards/reasoning_steps_reward": 0.9635417014360428,
-      "rewards/repetition_penalty_reward": -0.046313464641571045,
-      "rewards/tag_count_reward": 0.8710937649011612,
+      "completion_length": 461.5208435058594,
+      "epoch": 0.11698537682789652,
+      "grad_norm": 0.510885474337652,
+      "kl": 0.268310546875,
+      "learning_rate": 1.998336277024103e-05,
+      "loss": -0.0721,
+      "reward": 2.0055829286575317,
+      "reward_std": 0.4140588045120239,
+      "rewards/accuracy_reward": 0.5260416828095913,
+      "rewards/reasoning_steps_reward": 0.9809027910232544,
+      "rewards/repetition_penalty_reward": -0.22531990334391594,
+      "rewards/tag_count_reward": 0.7239583432674408,
       "step": 78
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 491.32814025878906,
-      "epoch": 0.2368815592203898,
-      "grad_norm": 1.7806633395846703,
-      "kl": 0.951171875,
-      "learning_rate": 1.8902899062363142e-05,
-      "loss": 0.6398,
-      "reward": 1.9999366104602814,
-      "reward_std": 0.7513662576675415,
-      "rewards/accuracy_reward": 0.3437500074505806,
-      "rewards/reasoning_steps_reward": 0.9409722983837128,
-      "rewards/repetition_penalty_reward": -0.032181489281356335,
-      "rewards/tag_count_reward": 0.747395858168602,
+      "completion_length": 501.5573043823242,
+      "epoch": 0.11848518935133108,
+      "grad_norm": 0.47404049136724163,
+      "kl": 0.241943359375,
+      "learning_rate": 1.998020136552566e-05,
+      "loss": 0.001,
+      "reward": 2.024559497833252,
+      "reward_std": 0.4507194012403488,
+      "rewards/accuracy_reward": 0.5572916716337204,
+      "rewards/reasoning_steps_reward": 0.9809028059244156,
+      "rewards/repetition_penalty_reward": -0.2545204684138298,
+      "rewards/tag_count_reward": 0.7408854365348816,
       "step": 79
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 629.9583587646484,
-      "epoch": 0.239880059970015,
-      "grad_norm": 0.894829793618965,
-      "kl": 0.8115234375,
-      "learning_rate": 1.8854560256532098e-05,
-      "loss": 0.5427,
-      "reward": 1.8730204403400421,
-      "reward_std": 0.8077119290828705,
-      "rewards/accuracy_reward": 0.354166679084301,
-      "rewards/reasoning_steps_reward": 0.9253472834825516,
-      "rewards/repetition_penalty_reward": -0.02628516126424074,
-      "rewards/tag_count_reward": 0.6197916865348816,
+      "completion_length": 470.04688262939453,
+      "epoch": 0.11998500187476566,
+      "grad_norm": 0.6494805682984701,
+      "kl": 0.2578125,
+      "learning_rate": 1.9976765434224426e-05,
+      "loss": 0.0393,
+      "reward": 2.306118667125702,
+      "reward_std": 0.3345252051949501,
+      "rewards/accuracy_reward": 0.7812500149011612,
+      "rewards/reasoning_steps_reward": 0.9826389253139496,
+      "rewards/repetition_penalty_reward": -0.2064681462943554,
+      "rewards/tag_count_reward": 0.7486979365348816,
       "step": 80
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 446.9895935058594,
-      "epoch": 0.24287856071964017,
-      "grad_norm": 1.5621375006184983,
-      "kl": 0.9130859375,
-      "learning_rate": 1.8805243942587e-05,
-      "loss": 0.7422,
-      "reward": 2.229804575443268,
-      "reward_std": 0.8122627884149551,
-      "rewards/accuracy_reward": 0.5520833358168602,
-      "rewards/reasoning_steps_reward": 0.9565972536802292,
-      "rewards/repetition_penalty_reward": -0.04970931261777878,
-      "rewards/tag_count_reward": 0.7708333432674408,
+      "completion_length": 371.0416717529297,
+      "epoch": 0.12148481439820022,
+      "grad_norm": 0.6158896116207796,
+      "kl": 0.27783203125,
+      "learning_rate": 1.9973055070849912e-05,
+      "loss": -0.0077,
+      "reward": 2.2738314270973206,
+      "reward_std": 0.3718957081437111,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.9652778059244156,
+      "rewards/repetition_penalty_reward": -0.1354569736868143,
+      "rewards/tag_count_reward": 0.7513020932674408,
       "step": 81
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 326.78126525878906,
-      "epoch": 0.24587706146926536,
-      "grad_norm": 0.734038354451238,
-      "kl": 0.5283203125,
-      "learning_rate": 1.8754955564852082e-05,
-      "loss": 0.4824,
-      "reward": 2.3457452058792114,
-      "reward_std": 0.41162994503974915,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.9930555671453476,
-      "rewards/repetition_penalty_reward": -0.08611250855028629,
-      "rewards/tag_count_reward": 0.8971354216337204,
+      "completion_length": 331.3958435058594,
+      "epoch": 0.1229846269216348,
+      "grad_norm": 0.681584270450499,
+      "kl": 0.27490234375,
+      "learning_rate": 1.996907037746352e-05,
+      "loss": -0.0158,
+      "reward": 2.2777538299560547,
+      "reward_std": 0.45179731398820877,
+      "rewards/accuracy_reward": 0.5989583432674408,
+      "rewards/reasoning_steps_reward": 0.9565972685813904,
+      "rewards/repetition_penalty_reward": -0.07597886584699154,
+      "rewards/tag_count_reward": 0.7981770932674408,
       "step": 82
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 268.6093864440918,
-      "epoch": 0.24887556221889057,
-      "grad_norm": 0.7617786864490775,
-      "kl": 0.4599609375,
-      "learning_rate": 1.870370067496355e-05,
-      "loss": 0.4915,
-      "reward": 2.696290135383606,
-      "reward_std": 0.42367345839738846,
-      "rewards/accuracy_reward": 0.8697916716337204,
-      "rewards/reasoning_steps_reward": 0.9756944477558136,
-      "rewards/repetition_penalty_reward": -0.09581072442233562,
-      "rewards/tag_count_reward": 0.9466145932674408,
+      "completion_length": 326.64063262939453,
+      "epoch": 0.12448443944506937,
+      "grad_norm": 0.6980728390227965,
+      "kl": 0.236083984375,
+      "learning_rate": 1.9964811463672685e-05,
+      "loss": -0.0154,
+      "reward": 2.332102954387665,
+      "reward_std": 0.472077339887619,
+      "rewards/accuracy_reward": 0.5468750149011612,
+      "rewards/reasoning_steps_reward": 0.9427083730697632,
+      "rewards/repetition_penalty_reward": -0.07024090643972158,
+      "rewards/tag_count_reward": 0.9127604365348816,
       "step": 83
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 251.5833396911621,
-      "epoch": 0.2518740629685157,
-      "grad_norm": 0.8354783555661154,
-      "kl": 0.4912109375,
-      "learning_rate": 1.8651484931256685e-05,
-      "loss": 0.3193,
-      "reward": 2.5308874249458313,
-      "reward_std": 0.34281000867486,
-      "rewards/accuracy_reward": 0.6770833432674408,
-      "rewards/reasoning_steps_reward": 0.9843750298023224,
-      "rewards/repetition_penalty_reward": -0.09281065501272678,
-      "rewards/tag_count_reward": 0.9622395932674408,
+      "completion_length": 345.2291793823242,
+      "epoch": 0.12598425196850394,
+      "grad_norm": 0.6918666028590195,
+      "kl": 0.29150390625,
+      "learning_rate": 1.996027844662785e-05,
+      "loss": -0.0238,
+      "reward": 2.347783863544464,
+      "reward_std": 0.497940331697464,
+      "rewards/accuracy_reward": 0.479166679084301,
+      "rewards/reasoning_steps_reward": 0.9878472536802292,
+      "rewards/repetition_penalty_reward": -0.06454263348132372,
+      "rewards/tag_count_reward": 0.9453125149011612,
       "step": 84
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 217.48438262939453,
-      "epoch": 0.25487256371814093,
-      "grad_norm": 0.8219462123752692,
-      "kl": 0.4619140625,
-      "learning_rate": 1.8598314098141208e-05,
-      "loss": 0.3273,
-      "reward": 2.6751975417137146,
-      "reward_std": 0.2594960853457451,
-      "rewards/accuracy_reward": 0.8072916865348816,
-      "rewards/reasoning_steps_reward": 0.970486119389534,
-      "rewards/repetition_penalty_reward": -0.0817470382899046,
-      "rewards/tag_count_reward": 0.9791667014360428,
+      "completion_length": 352.5729293823242,
+      "epoch": 0.1274840644919385,
+      "grad_norm": 0.662339973250043,
+      "kl": 0.207763671875,
+      "learning_rate": 1.9955471451019264e-05,
+      "loss": 0.0105,
+      "reward": 2.5184255242347717,
+      "reward_std": 0.423542745411396,
+      "rewards/accuracy_reward": 0.645833358168602,
+      "rewards/reasoning_steps_reward": 0.9756944477558136,
+      "rewards/repetition_penalty_reward": -0.06534186284989119,
+      "rewards/tag_count_reward": 0.9622395932674408,
       "step": 85
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 238.35937881469727,
-      "epoch": 0.25787106446776614,
-      "grad_norm": 0.8013857932799926,
-      "kl": 0.5302734375,
-      "learning_rate": 1.8544194045464888e-05,
-      "loss": 0.5448,
-      "reward": 2.324418604373932,
-      "reward_std": 0.5969930738210678,
-      "rewards/accuracy_reward": 0.5625000223517418,
-      "rewards/reasoning_steps_reward": 0.8784722536802292,
-      "rewards/repetition_penalty_reward": -0.06837661191821098,
-      "rewards/tag_count_reward": 0.9518229365348816,
+      "completion_length": 373.8958435058594,
+      "epoch": 0.1289838770153731,
+      "grad_norm": 1.1704532474263813,
+      "kl": 0.46435546875,
+      "learning_rate": 1.995039060907352e-05,
+      "loss": -0.0028,
+      "reward": 2.421631157398224,
+      "reward_std": 0.5489099845290184,
+      "rewards/accuracy_reward": 0.5989583432674408,
+      "rewards/reasoning_steps_reward": 0.96875,
+      "rewards/repetition_penalty_reward": -0.05753568932414055,
+      "rewards/tag_count_reward": 0.911458358168602,
       "step": 86
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 253.87500762939453,
-      "epoch": 0.2608695652173913,
-      "grad_norm": 1.2404216424832606,
-      "kl": 0.623046875,
-      "learning_rate": 1.848913074786555e-05,
-      "loss": 0.3312,
-      "reward": 2.4448606371879578,
-      "reward_std": 0.5359718501567841,
-      "rewards/accuracy_reward": 0.6927083507180214,
-      "rewards/reasoning_steps_reward": 0.8732638955116272,
-      "rewards/repetition_penalty_reward": -0.07163260504603386,
-      "rewards/tag_count_reward": 0.950520858168602,
+      "completion_length": 353.6979217529297,
+      "epoch": 0.13048368953880765,
+      "grad_norm": 3.3058755481715214,
+      "kl": 1.44921875,
+      "learning_rate": 1.994503606054994e-05,
+      "loss": -0.0166,
+      "reward": 2.3031901717185974,
+      "reward_std": 0.5215350016951561,
+      "rewards/accuracy_reward": 0.510416679084301,
+      "rewards/reasoning_steps_reward": 0.9618055820465088,
+      "rewards/repetition_penalty_reward": -0.0635634008795023,
+      "rewards/tag_count_reward": 0.8945312798023224,
       "step": 87
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 266.10939025878906,
-      "epoch": 0.2638680659670165,
-      "grad_norm": 2.28058295704287,
-      "kl": 0.61572265625,
-      "learning_rate": 1.843313028411149e-05,
-      "loss": 0.5956,
-      "reward": 2.511428415775299,
-      "reward_std": 0.48727843910455704,
-      "rewards/accuracy_reward": 0.708333358168602,
-      "rewards/reasoning_steps_reward": 0.9253472536802292,
-      "rewards/repetition_penalty_reward": -0.05845023598521948,
-      "rewards/tag_count_reward": 0.9361979365348816,
+      "completion_length": 340.1927185058594,
+      "epoch": 0.13198350206224221,
+      "grad_norm": 1.3397859193343156,
+      "kl": 0.47265625,
+      "learning_rate": 1.9939407952736737e-05,
+      "loss": -0.0518,
+      "reward": 2.3962929248809814,
+      "reward_std": 0.5894087105989456,
+      "rewards/accuracy_reward": 0.598958358168602,
+      "rewards/reasoning_steps_reward": 0.9809027910232544,
+      "rewards/repetition_penalty_reward": -0.07028705440461636,
+      "rewards/tag_count_reward": 0.8867187649011612,
       "step": 88
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 290.9114685058594,
-      "epoch": 0.26686656671664166,
-      "grad_norm": 259.99242737782754,
-      "kl": 3.0830078125,
-      "learning_rate": 1.8376198836430415e-05,
-      "loss": 0.6364,
-      "reward": 2.46868097782135,
-      "reward_std": 0.635415643453598,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.9461806118488312,
-      "rewards/repetition_penalty_reward": -0.06083299312740564,
-      "rewards/tag_count_reward": 0.9166667014360428,
+      "completion_length": 319.2916793823242,
+      "epoch": 0.13348331458567678,
+      "grad_norm": 1.6274329590710197,
+      "kl": 0.83447265625,
+      "learning_rate": 1.9933506440446932e-05,
+      "loss": -0.0982,
+      "reward": 2.4367164373397827,
+      "reward_std": 0.6716105788946152,
+      "rewards/accuracy_reward": 0.6458333507180214,
+      "rewards/reasoning_steps_reward": 0.9548611789941788,
+      "rewards/repetition_penalty_reward": -0.051998937502503395,
+      "rewards/tag_count_reward": 0.888020858168602,
       "step": 89
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 379.1302185058594,
-      "epoch": 0.2698650674662669,
-      "grad_norm": 1.9382632692598936,
-      "kl": 0.8193359375,
-      "learning_rate": 1.8318342689826938e-05,
-      "loss": 0.7494,
-      "reward": 2.3082011342048645,
-      "reward_std": 0.7488896995782852,
-      "rewards/accuracy_reward": 0.5572916716337204,
-      "rewards/reasoning_steps_reward": 0.9392361342906952,
-      "rewards/repetition_penalty_reward": -0.04249336663633585,
-      "rewards/tag_count_reward": 0.8541667014360428,
+      "completion_length": 329.4791717529297,
+      "epoch": 0.13498312710911137,
+      "grad_norm": 0.7057852059401287,
+      "kl": 0.335693359375,
+      "learning_rate": 1.992733168601413e-05,
+      "loss": -0.069,
+      "reward": 2.477515757083893,
+      "reward_std": 0.4706941097974777,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.9704861491918564,
+      "rewards/repetition_penalty_reward": -0.07369956281036139,
+      "rewards/tag_count_reward": 0.9244791865348816,
       "step": 90
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 344.7395935058594,
-      "epoch": 0.272863568215892,
-      "grad_norm": 1.4041925703207834,
-      "kl": 0.7041015625,
-      "learning_rate": 1.8259568231388737e-05,
-      "loss": 0.7354,
-      "reward": 2.346260666847229,
-      "reward_std": 0.6538338512182236,
-      "rewards/accuracy_reward": 0.552083358168602,
-      "rewards/reasoning_steps_reward": 0.9739583730697632,
-      "rewards/repetition_penalty_reward": -0.04696857463568449,
-      "rewards/tag_count_reward": 0.8671875149011612,
+      "completion_length": 321.5677185058594,
+      "epoch": 0.13648293963254593,
+      "grad_norm": 0.7018868149602894,
+      "kl": 0.37060546875,
+      "learning_rate": 1.9920883859288035e-05,
+      "loss": -0.0407,
+      "reward": 2.4766435027122498,
+      "reward_std": 0.5056409984827042,
+      "rewards/accuracy_reward": 0.6510416865348816,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.07326984778046608,
+      "rewards/tag_count_reward": 0.9231770932674408,
       "step": 91
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 329.8073043823242,
-      "epoch": 0.27586206896551724,
-      "grad_norm": 0.9388360459997904,
-      "kl": 0.697265625,
-      "learning_rate": 1.819988194958146e-05,
-      "loss": 0.7864,
-      "reward": 2.366865336894989,
-      "reward_std": 0.7087779939174652,
-      "rewards/accuracy_reward": 0.5937500149011612,
-      "rewards/reasoning_steps_reward": 0.947916716337204,
-      "rewards/repetition_penalty_reward": -0.053707641549408436,
-      "rewards/tag_count_reward": 0.8789062798023224,
+      "completion_length": 309.5208435058594,
+      "epoch": 0.1379827521559805,
+      "grad_norm": 1.1625720671949755,
+      "kl": 0.3447265625,
+      "learning_rate": 1.991416313762978e-05,
+      "loss": -0.0604,
+      "reward": 2.4806445837020874,
+      "reward_std": 0.46803582459688187,
+      "rewards/accuracy_reward": 0.677083358168602,
+      "rewards/reasoning_steps_reward": 0.9531250447034836,
+      "rewards/repetition_penalty_reward": -0.08445966802537441,
+      "rewards/tag_count_reward": 0.9348958432674408,
       "step": 92
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 323.0104293823242,
-      "epoch": 0.27886056971514245,
-      "grad_norm": 1.0904070944040622,
-      "kl": 0.67578125,
-      "learning_rate": 1.8139290433532415e-05,
-      "loss": 0.7764,
-      "reward": 2.373181402683258,
-      "reward_std": 0.7174654752016068,
-      "rewards/accuracy_reward": 0.6093750298023224,
-      "rewards/reasoning_steps_reward": 0.9392361044883728,
-      "rewards/repetition_penalty_reward": -0.05433602724224329,
-      "rewards/tag_count_reward": 0.8789062798023224,
+      "completion_length": 333.87500762939453,
+      "epoch": 0.13948256467941508,
+      "grad_norm": 0.7197532890419,
+      "kl": 0.40478515625,
+      "learning_rate": 1.990716970590706e-05,
+      "loss": -0.0581,
+      "reward": 2.5884640216827393,
+      "reward_std": 0.3948858380317688,
+      "rewards/accuracy_reward": 0.739583358168602,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.08688333258032799,
+      "rewards/tag_count_reward": 0.9531250149011612,
       "step": 93
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 286.31250762939453,
-      "epoch": 0.2818590704647676,
-      "grad_norm": 1.345138349799746,
-      "kl": 0.6591796875,
-      "learning_rate": 1.807780037230315e-05,
-      "loss": 0.9459,
-      "reward": 2.401767611503601,
-      "reward_std": 0.7818585783243179,
-      "rewards/accuracy_reward": 0.6718750149011612,
-      "rewards/reasoning_steps_reward": 0.9218750447034836,
-      "rewards/repetition_penalty_reward": -0.052659488283097744,
-      "rewards/tag_count_reward": 0.8606770932674408,
+      "completion_length": 331.9166793823242,
+      "epoch": 0.14098237720284965,
+      "grad_norm": 3.2153625010582005,
+      "kl": 5.23779296875,
+      "learning_rate": 1.989990375648903e-05,
+      "loss": -0.0484,
+      "reward": 2.4388928413391113,
+      "reward_std": 0.33317605406045914,
+      "rewards/accuracy_reward": 0.6093750149011612,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.09322533197700977,
+      "rewards/tag_count_reward": 0.950520858168602,
       "step": 94
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 320.9583435058594,
-      "epoch": 0.2848575712143928,
-      "grad_norm": 1.6639693291434057,
-      "kl": 0.837890625,
-      "learning_rate": 1.8015418554151024e-05,
-      "loss": 0.7482,
-      "reward": 2.097085416316986,
-      "reward_std": 0.9631160348653793,
-      "rewards/accuracy_reward": 0.536458358168602,
-      "rewards/reasoning_steps_reward": 0.8246527761220932,
-      "rewards/repetition_penalty_reward": -0.04397374298423529,
-      "rewards/tag_count_reward": 0.7799479216337204,
+      "completion_length": 354.98439025878906,
+      "epoch": 0.1424821897262842,
+      "grad_norm": 1.7303354539555662,
+      "kl": 0.7705078125,
+      "learning_rate": 1.9892365489241023e-05,
+      "loss": -0.0393,
+      "reward": 2.3029019832611084,
+      "reward_std": 0.4447196274995804,
+      "rewards/accuracy_reward": 0.4843750149011612,
+      "rewards/reasoning_steps_reward": 0.9791667014360428,
+      "rewards/repetition_penalty_reward": -0.10855649225413799,
+      "rewards/tag_count_reward": 0.9479166716337204,
       "step": 95
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 107.17187881469727,
-      "epoch": 0.28785607196401797,
-      "grad_norm": 2.6809348147477285,
-      "kl": 1.669921875,
-      "learning_rate": 1.7952151865779792e-05,
-      "loss": -0.1937,
-      "reward": 1.646718680858612,
-      "reward_std": 1.0644067823886871,
-      "rewards/accuracy_reward": 0.3437500037252903,
-      "rewards/reasoning_steps_reward": 0.7274306118488312,
-      "rewards/repetition_penalty_reward": -0.03253482934087515,
-      "rewards/tag_count_reward": 0.6080729514360428,
+      "completion_length": 365.9635467529297,
+      "epoch": 0.1439820022497188,
+      "grad_norm": 0.6730986080559331,
+      "kl": 0.39892578125,
+      "learning_rate": 1.988455511151906e-05,
+      "loss": -0.0247,
+      "reward": 2.3117390871047974,
+      "reward_std": 0.43109431117773056,
+      "rewards/accuracy_reward": 0.4687500111758709,
+      "rewards/reasoning_steps_reward": 0.9809028059244156,
+      "rewards/repetition_penalty_reward": -0.11968471482396126,
+      "rewards/tag_count_reward": 0.9817708432674408,
       "step": 96
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 137.05208587646484,
-      "epoch": 0.2908545727136432,
-      "grad_norm": 1.661973040336361,
-      "kl": 0.6298828125,
-      "learning_rate": 1.7888007291579357e-05,
-      "loss": -0.0513,
-      "reward": 2.4063791632652283,
-      "reward_std": 0.647190622985363,
-      "rewards/accuracy_reward": 0.6197916865348816,
-      "rewards/reasoning_steps_reward": 0.9531250596046448,
-      "rewards/repetition_penalty_reward": -0.0636730408295989,
-      "rewards/tag_count_reward": 0.8971354365348816,
+      "completion_length": 367.5104293823242,
+      "epoch": 0.14548181477315336,
+      "grad_norm": 0.6453267932676297,
+      "kl": 0.283203125,
+      "learning_rate": 1.987647283816412e-05,
+      "loss": -0.016,
+      "reward": 2.5260453820228577,
+      "reward_std": 0.3044511452317238,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.11284366995096207,
+      "rewards/tag_count_reward": 0.9947916716337204,
       "step": 97
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 151.68750381469727,
-      "epoch": 0.2938530734632684,
-      "grad_norm": 2.7127088778027684,
-      "kl": 0.8984375,
-      "learning_rate": 1.7822991912854716e-05,
-      "loss": 0.195,
-      "reward": 2.5627546310424805,
-      "reward_std": 0.4789634570479393,
-      "rewards/accuracy_reward": 0.7135416865348816,
-      "rewards/reasoning_steps_reward": 0.9826389104127884,
-      "rewards/repetition_penalty_reward": -0.060509427450597286,
-      "rewards/tag_count_reward": 0.9270833432674408,
+      "completion_length": 338.60938262939453,
+      "epoch": 0.14698162729658792,
+      "grad_norm": 0.702216348224918,
+      "kl": 0.31689453125,
+      "learning_rate": 1.9868118891496268e-05,
+      "loss": 0.013,
+      "reward": 2.7034881114959717,
+      "reward_std": 0.254051860421896,
+      "rewards/accuracy_reward": 0.817708358168602,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.09599110670387745,
+      "rewards/tag_count_reward": 0.9973958432674408,
       "step": 98
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 211.39062881469727,
-      "epoch": 0.29685157421289354,
-      "grad_norm": 18.504039023965966,
-      "kl": 1.9638671875,
-      "learning_rate": 1.77571129070442e-05,
-      "loss": 0.3326,
-      "reward": 2.2352594137191772,
-      "reward_std": 0.7260878309607506,
-      "rewards/accuracy_reward": 0.583333358168602,
-      "rewards/reasoning_steps_reward": 0.890625,
-      "rewards/repetition_penalty_reward": -0.06291783228516579,
-      "rewards/tag_count_reward": 0.82421875,
+      "completion_length": 336.6822967529297,
+      "epoch": 0.1484814398200225,
+      "grad_norm": 0.6294487017254564,
+      "kl": 0.28369140625,
+      "learning_rate": 1.98594935013085e-05,
+      "loss": -0.0063,
+      "reward": 2.4194209575653076,
+      "reward_std": 0.31747131049633026,
+      "rewards/accuracy_reward": 0.5312500074505806,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.09012776426970959,
+      "rewards/tag_count_reward": 0.9921875149011612,
       "step": 99
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 410.6458435058594,
-      "epoch": 0.29985007496251875,
-      "grad_norm": 7.369461971351607,
-      "kl": 1.18359375,
-      "learning_rate": 1.7690377546927134e-05,
-      "loss": 0.182,
-      "reward": 2.507060259580612,
-      "reward_std": 0.49995335936546326,
-      "rewards/accuracy_reward": 0.6458333488553762,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.055439687334001064,
-      "rewards/tag_count_reward": 0.9375000298023224,
+      "completion_length": 336.93750762939453,
+      "epoch": 0.14998125234345708,
+      "grad_norm": 0.7622200900254438,
+      "kl": 0.2880859375,
+      "learning_rate": 1.985059690486045e-05,
+      "loss": 0.0147,
+      "reward": 2.602359890937805,
+      "reward_std": 0.4236067906022072,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9878472536802292,
+      "rewards/repetition_penalty_reward": -0.07949773035943508,
+      "rewards/tag_count_reward": 0.985677108168602,
       "step": 100
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 614.0625152587891,
-      "epoch": 0.3028485757121439,
-      "grad_norm": 4.425874093822116,
-      "kl": 1.77734375,
-      "learning_rate": 1.7622793199820935e-05,
-      "loss": 0.2128,
-      "reward": 2.6125723719596863,
-      "reward_std": 0.4479042589664459,
-      "rewards/accuracy_reward": 0.7239583432674408,
-      "rewards/reasoning_steps_reward": 0.987847238779068,
-      "rewards/repetition_penalty_reward": -0.0510562164708972,
-      "rewards/tag_count_reward": 0.9518229216337204,
+      "completion_length": 311.07814025878906,
+      "epoch": 0.15148106486689164,
+      "grad_norm": 0.820138039709133,
+      "kl": 0.572998046875,
+      "learning_rate": 1.9841429346871863e-05,
+      "loss": -0.023,
+      "reward": 2.39324152469635,
+      "reward_std": 0.5194854661822319,
+      "rewards/accuracy_reward": 0.5104166865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0703001944348216,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 101
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 817.4895935058594,
-      "epoch": 0.3058470764617691,
-      "grad_norm": 4.565018933297784,
-      "kl": 3.43359375,
-      "learning_rate": 1.7554367326767793e-05,
-      "loss": 0.1219,
-      "reward": 2.5981725454330444,
-      "reward_std": 0.4292430207133293,
-      "rewards/accuracy_reward": 0.6875000223517418,
-      "rewards/reasoning_steps_reward": 0.9930555820465088,
-      "rewards/repetition_penalty_reward": -0.043320574797689915,
-      "rewards/tag_count_reward": 0.9609375298023224,
+      "completion_length": 294.9427185058594,
+      "epoch": 0.1529808773903262,
+      "grad_norm": 0.6430598505979571,
+      "kl": 0.318359375,
+      "learning_rate": 1.9831991079515836e-05,
+      "loss": -0.0646,
+      "reward": 2.543990731239319,
+      "reward_std": 0.49773871898651123,
+      "rewards/accuracy_reward": 0.7239583432674408,
+      "rewards/reasoning_steps_reward": 0.91493059694767,
+      "rewards/repetition_penalty_reward": -0.07536709867417812,
+      "rewards/tag_count_reward": 0.98046875,
       "step": 102
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 140.87500381469727,
-      "epoch": 0.30884557721139433,
-      "grad_norm": 14.749084867590199,
-      "kl": 1.5126953125,
-      "learning_rate": 1.7485107481711014e-05,
-      "loss": 0.3689,
-      "reward": 2.522024154663086,
-      "reward_std": 0.3764337971806526,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.9878472238779068,
-      "rewards/repetition_penalty_reward": -0.07910444028675556,
-      "rewards/tag_count_reward": 0.9674479365348816,
+      "completion_length": 309.2604293823242,
+      "epoch": 0.1544806899137608,
+      "grad_norm": 0.9425563472766899,
+      "kl": 0.55615234375,
+      "learning_rate": 1.982228236241192e-05,
+      "loss": -0.063,
+      "reward": 2.535941779613495,
+      "reward_std": 0.4504154324531555,
+      "rewards/accuracy_reward": 0.6822916865348816,
+      "rewards/reasoning_steps_reward": 0.94618059694767,
+      "rewards/repetition_penalty_reward": -0.07039502263069153,
+      "rewards/tag_count_reward": 0.977864608168602,
       "step": 103
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 140.06771278381348,
-      "epoch": 0.3118440779610195,
-      "grad_norm": 1.5854423021252229,
-      "kl": 0.75390625,
-      "learning_rate": 1.7415021310661073e-05,
-      "loss": 0.2421,
-      "reward": 2.5152770280838013,
-      "reward_std": 0.5358111336827278,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 0.989583358168602,
-      "rewards/repetition_penalty_reward": -0.08628544956445694,
-      "rewards/tag_count_reward": 0.9453125149011612,
+      "completion_length": 309.23439025878906,
+      "epoch": 0.15598050243719536,
+      "grad_norm": 0.8058787849857703,
+      "kl": 0.38916015625,
+      "learning_rate": 1.9812303462618945e-05,
+      "loss": -0.0416,
+      "reward": 2.438301682472229,
+      "reward_std": 0.46069950610399246,
+      "rewards/accuracy_reward": 0.5729166939854622,
+      "rewards/reasoning_steps_reward": 0.9652778059244156,
+      "rewards/repetition_penalty_reward": -0.05692401435226202,
+      "rewards/tag_count_reward": 0.9570312649011612,
       "step": 104
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 167.8229217529297,
-      "epoch": 0.3148425787106447,
-      "grad_norm": 2.0606435436365245,
-      "kl": 0.578125,
-      "learning_rate": 1.7344116550851546e-05,
-      "loss": 0.477,
-      "reward": 2.623923897743225,
-      "reward_std": 0.48354143649339676,
-      "rewards/accuracy_reward": 0.7343750149011612,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.07659695856273174,
-      "rewards/tag_count_reward": 0.9713541865348816,
+      "completion_length": 318.7916717529297,
+      "epoch": 0.15748031496062992,
+      "grad_norm": 0.8045229402138098,
+      "kl": 0.427734375,
+      "learning_rate": 1.9802054654627694e-05,
+      "loss": -0.0673,
+      "reward": 2.3380337357521057,
+      "reward_std": 0.5075135007500648,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/reasoning_steps_reward": 0.9670139402151108,
+      "rewards/repetition_penalty_reward": -0.06778228841722012,
+      "rewards/tag_count_reward": 0.938802108168602,
       "step": 105
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 142.77083587646484,
-      "epoch": 0.31784107946026985,
-      "grad_norm": 10.522238174054582,
-      "kl": 1.28515625,
-      "learning_rate": 1.7272401029884932e-05,
-      "loss": 0.2673,
-      "reward": 2.4795719981193542,
-      "reward_std": 0.5011198297142982,
-      "rewards/accuracy_reward": 0.5937500149011612,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.08379613049328327,
-      "rewards/tag_count_reward": 0.9765625149011612,
+      "completion_length": 338.3854217529297,
+      "epoch": 0.15898012748406448,
+      "grad_norm": 0.7209652038046875,
+      "kl": 0.27587890625,
+      "learning_rate": 1.9791536220353355e-05,
+      "loss": -0.0348,
+      "reward": 2.4589534401893616,
+      "reward_std": 0.45359161496162415,
+      "rewards/accuracy_reward": 0.578125,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.06665424816310406,
+      "rewards/tag_count_reward": 0.9648437649011612,
       "step": 106
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 134.3802146911621,
-      "epoch": 0.32083958020989506,
-      "grad_norm": 3.3181439930079146,
-      "kl": 0.7978515625,
-      "learning_rate": 1.719988266486854e-05,
-      "loss": 0.1636,
-      "reward": 2.5852218866348267,
-      "reward_std": 0.4785446897149086,
-      "rewards/accuracy_reward": 0.6875000149011612,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.08665317296981812,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 311.0208511352539,
+      "epoch": 0.16047994000749907,
+      "grad_norm": 0.6649826820375959,
+      "kl": 0.36474609375,
+      "learning_rate": 1.9780748449127745e-05,
+      "loss": -0.061,
+      "reward": 2.545065224170685,
+      "reward_std": 0.5411981120705605,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.9670138955116272,
+      "rewards/repetition_penalty_reward": -0.07950077578425407,
+      "rewards/tag_count_reward": 0.9648437649011612,
       "step": 107
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 152.4479217529297,
-      "epoch": 0.3238380809595202,
-      "grad_norm": 3.2954339112996336,
-      "kl": 0.8662109375,
-      "learning_rate": 1.7126569461540445e-05,
-      "loss": 0.1118,
-      "reward": 2.564746856689453,
-      "reward_std": 0.34381402283906937,
-      "rewards/accuracy_reward": 0.661458358168602,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.08759699948132038,
-      "rewards/tag_count_reward": 0.9908854365348816,
+      "completion_length": 315.9896011352539,
+      "epoch": 0.16197975253093364,
+      "grad_norm": 0.8786847698949797,
+      "kl": 0.3779296875,
+      "learning_rate": 1.976969163769137e-05,
+      "loss": -0.083,
+      "reward": 2.484778046607971,
+      "reward_std": 0.5565851628780365,
+      "rewards/accuracy_reward": 0.6354166865348816,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.06513523031026125,
+      "rewards/tag_count_reward": 0.9440104365348816,
       "step": 108
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 170.9166717529297,
-      "epoch": 0.3268365817091454,
-      "grad_norm": 5.1235193996536745,
-      "kl": 1.318359375,
-      "learning_rate": 1.70524695133857e-05,
-      "loss": 0.2634,
-      "reward": 2.745190441608429,
-      "reward_std": 0.3354305140674114,
-      "rewards/accuracy_reward": 0.8697916865348816,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.09465333260595798,
-      "rewards/tag_count_reward": 0.9752604365348816,
+      "completion_length": 327.48438262939453,
+      "epoch": 0.1634795650543682,
+      "grad_norm": 0.8374810187766324,
+      "kl": 0.47119140625,
+      "learning_rate": 1.9758366090185255e-05,
+      "loss": -0.0833,
+      "reward": 2.4872137904167175,
+      "reward_std": 0.546069398522377,
+      "rewards/accuracy_reward": 0.677083358168602,
+      "rewards/reasoning_steps_reward": 0.9600694626569748,
+      "rewards/repetition_penalty_reward": -0.07962668687105179,
+      "rewards/tag_count_reward": 0.9296875298023224,
       "step": 109
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 157.3489646911621,
-      "epoch": 0.32983508245877063,
-      "grad_norm": 6.042829717081394,
-      "kl": 0.63818359375,
-      "learning_rate": 1.6977591000742855e-05,
-      "loss": 0.1922,
-      "reward": 2.687567353248596,
-      "reward_std": 0.2190675288438797,
-      "rewards/accuracy_reward": 0.7916667014360428,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.08326606638729572,
-      "rewards/tag_count_reward": 0.9843750149011612,
+      "completion_length": 266.3177185058594,
+      "epoch": 0.1649793775778028,
+      "grad_norm": 8.874417749876018,
+      "kl": 1.828125,
+      "learning_rate": 1.974677211814259e-05,
+      "loss": -0.1294,
+      "reward": 1.922803819179535,
+      "reward_std": 0.5993900671601295,
+      "rewards/accuracy_reward": 0.2812500111758709,
+      "rewards/reasoning_steps_reward": 0.9583333283662796,
+      "rewards/repetition_penalty_reward": -0.0719878925010562,
+      "rewards/tag_count_reward": 0.7552083432674408,
       "step": 110
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 149.07813262939453,
-      "epoch": 0.3328335832083958,
-      "grad_norm": 27.456070537172923,
-      "kl": 1.2197265625,
-      "learning_rate": 1.6901942189900867e-05,
-      "loss": 0.0855,
-      "reward": 2.70491623878479,
-      "reward_std": 0.4226943477988243,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.07372974790632725,
-      "rewards/tag_count_reward": 0.981770858168602,
+      "completion_length": 282.6145935058594,
+      "epoch": 0.16647919010123735,
+      "grad_norm": 3.0395634773886893,
+      "kl": 1.529296875,
+      "learning_rate": 1.973491004048014e-05,
+      "loss": -0.1461,
+      "reward": 2.2567337453365326,
+      "reward_std": 0.6766453832387924,
+      "rewards/accuracy_reward": 0.5364583507180214,
+      "rewards/reasoning_steps_reward": 0.9548611491918564,
+      "rewards/repetition_penalty_reward": -0.07963782362639904,
+      "rewards/tag_count_reward": 0.845052108168602,
       "step": 111
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 152.4583396911621,
-      "epoch": 0.335832083958021,
-      "grad_norm": 1317.4357368028043,
-      "kl": 24.67529296875,
-      "learning_rate": 1.6825531432186545e-05,
-      "loss": 1.4049,
-      "reward": 2.6002301573753357,
-      "reward_std": 0.3956380560994148,
-      "rewards/accuracy_reward": 0.6979166716337204,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.0807595532387495,
-      "rewards/tag_count_reward": 0.9830729365348816,
+      "completion_length": 294.3958435058594,
+      "epoch": 0.1679790026246719,
+      "grad_norm": 0.7041219895358697,
+      "kl": 0.29052734375,
+      "learning_rate": 1.9722780183489477e-05,
+      "loss": -0.0136,
+      "reward": 2.4400432407855988,
+      "reward_std": 0.19339026510715485,
+      "rewards/accuracy_reward": 0.546875013038516,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.08990472927689552,
+      "rewards/tag_count_reward": 0.9882812649011612,
       "step": 112
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 151.70313262939453,
-      "epoch": 0.33883058470764615,
-      "grad_norm": 186.82025106193726,
-      "kl": 5.892578125,
-      "learning_rate": 1.6748367163042577e-05,
-      "loss": 0.1938,
-      "reward": 2.6722583174705505,
-      "reward_std": 0.47787418961524963,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.07123132981359959,
-      "rewards/tag_count_reward": 0.9518229216337204,
+      "completion_length": 285.1770935058594,
+      "epoch": 0.16947881514810648,
+      "grad_norm": 0.8587640795737388,
+      "kl": 0.6201171875,
+      "learning_rate": 1.9710382880828028e-05,
+      "loss": -0.0167,
+      "reward": 2.425099015235901,
+      "reward_std": 0.20590725913643837,
+      "rewards/accuracy_reward": 0.5677083507180214,
+      "rewards/reasoning_steps_reward": 0.9687500298023224,
+      "rewards/repetition_penalty_reward": -0.09443240240216255,
+      "rewards/tag_count_reward": 0.9830729365348816,
       "step": 113
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 151.92188262939453,
-      "epoch": 0.34182908545727136,
-      "grad_norm": 4.951400788715416,
-      "kl": 0.75830078125,
-      "learning_rate": 1.6670457901096328e-05,
-      "loss": 0.0216,
-      "reward": 2.5308876037597656,
-      "reward_std": 0.4789142981171608,
-      "rewards/accuracy_reward": 0.6666666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.07197706028819084,
-      "rewards/tag_count_reward": 0.9361979514360428,
+      "completion_length": 303.3541793823242,
+      "epoch": 0.17097862767154107,
+      "grad_norm": 0.6234593444589465,
+      "kl": 0.29541015625,
+      "learning_rate": 1.969771847350987e-05,
+      "loss": 0.0059,
+      "reward": 2.361898422241211,
+      "reward_std": 0.21231437101960182,
+      "rewards/accuracy_reward": 0.44791667722165585,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.07560167275369167,
+      "rewards/tag_count_reward": 0.9947916716337204,
       "step": 114
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 154.68750381469727,
-      "epoch": 0.3448275862068966,
-      "grad_norm": 1.1162974968117996,
-      "kl": 0.4677734375,
-      "learning_rate": 1.659181224721938e-05,
-      "loss": 0.021,
-      "reward": 2.730695605278015,
-      "reward_std": 0.3208945095539093,
-      "rewards/accuracy_reward": 0.8385416865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.07008570153266191,
-      "rewards/tag_count_reward": 0.9622395932674408,
+      "completion_length": 261.2395935058594,
+      "epoch": 0.17247844019497563,
+      "grad_norm": 0.7078512660636744,
+      "kl": 0.2978515625,
+      "learning_rate": 1.968478730989636e-05,
+      "loss": -0.0062,
+      "reward": 2.1943055987358093,
+      "reward_std": 0.16293694078922272,
+      "rewards/accuracy_reward": 0.2760416716337204,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.05569446366280317,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 115
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 166.71875381469727,
-      "epoch": 0.34782608695652173,
-      "grad_norm": 2.052554622098234,
-      "kl": 0.48193359375,
-      "learning_rate": 1.6512438883578047e-05,
-      "loss": 0.0718,
-      "reward": 2.7236560583114624,
-      "reward_std": 0.2550913393497467,
-      "rewards/accuracy_reward": 0.8437500298023224,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.0706147812306881,
-      "rewards/tag_count_reward": 0.9505208432674408,
+      "completion_length": 253.6718864440918,
+      "epoch": 0.1739782527184102,
+      "grad_norm": 0.7740364590084335,
+      "kl": 0.3125,
+      "learning_rate": 1.9671589745686563e-05,
+      "loss": 0.0356,
+      "reward": 2.4594351053237915,
+      "reward_std": 0.15120337810367346,
+      "rewards/accuracy_reward": 0.5312500102445483,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.052283622324466705,
+      "rewards/tag_count_reward": 0.9960937649011612,
       "step": 116
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 150.95833587646484,
-      "epoch": 0.35082458770614694,
-      "grad_norm": 7.239442859707766,
-      "kl": 1.1904296875,
-      "learning_rate": 1.6432346572674897e-05,
-      "loss": 0.0808,
-      "reward": 2.7389387488365173,
-      "reward_std": 0.36363864317536354,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.06835305411368608,
-      "rewards/tag_count_reward": 0.9531250149011612,
+      "completion_length": 240.8385467529297,
+      "epoch": 0.17547806524184478,
+      "grad_norm": 0.6815939277095767,
+      "kl": 0.5048828125,
+      "learning_rate": 1.965812614390743e-05,
+      "loss": 0.014,
+      "reward": 2.592436134815216,
+      "reward_std": 0.151106015779078,
+      "rewards/accuracy_reward": 0.6614583507180214,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.0533972904086113,
+      "rewards/tag_count_reward": 0.989583358168602,
       "step": 117
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 159.48958587646484,
-      "epoch": 0.3538230884557721,
-      "grad_norm": 2.1682395821362865,
-      "kl": 0.53564453125,
-      "learning_rate": 1.6351544156381413e-05,
-      "loss": 0.0727,
-      "reward": 2.877605438232422,
-      "reward_std": 0.2183693777769804,
-      "rewards/accuracy_reward": 0.9687500149011612,
-      "rewards/reasoning_steps_reward": 0.9965277761220932,
-      "rewards/repetition_penalty_reward": -0.055120449513196945,
-      "rewards/tag_count_reward": 0.9674479365348816,
+      "completion_length": 229.6510467529297,
+      "epoch": 0.17697787776527935,
+      "grad_norm": 0.8414344996361253,
+      "kl": 0.3369140625,
+      "learning_rate": 1.9644396874903865e-05,
+      "loss": 0.0194,
+      "reward": 2.4946166276931763,
+      "reward_std": 0.23435868322849274,
+      "rewards/accuracy_reward": 0.5416666865348816,
+      "rewards/reasoning_steps_reward": 0.998263880610466,
+      "rewards/repetition_penalty_reward": -0.04401196166872978,
+      "rewards/tag_count_reward": 0.9986979216337204,
       "step": 118
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 157.2395896911621,
-      "epoch": 0.3568215892053973,
-      "grad_norm": 38.71837348464348,
-      "kl": 2.7021484375,
-      "learning_rate": 1.6270040554961866e-05,
-      "loss": 0.2602,
-      "reward": 2.6591410040855408,
-      "reward_std": 0.34208307787775993,
-      "rewards/accuracy_reward": 0.739583358168602,
+      "completion_length": 213.03125762939453,
+      "epoch": 0.1784776902887139,
+      "grad_norm": 0.7595325300854338,
+      "kl": 0.33984375,
+      "learning_rate": 1.9630402316328506e-05,
+      "loss": 0.0338,
+      "reward": 2.496553957462311,
+      "reward_std": 0.13166913157328963,
+      "rewards/accuracy_reward": 0.526041679084301,
       "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.05440086964517832,
-      "rewards/tag_count_reward": 0.973958358168602,
+      "rewards/repetition_penalty_reward": -0.022977421525865793,
+      "rewards/tag_count_reward": 0.993489608168602,
       "step": 119
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 145.41666793823242,
-      "epoch": 0.3598200899550225,
-      "grad_norm": 4.316214986822223,
-      "kl": 0.68115234375,
-      "learning_rate": 1.6187844766088586e-05,
-      "loss": 0.2187,
-      "reward": 2.7976441979408264,
-      "reward_std": 0.3110164441168308,
-      "rewards/accuracy_reward": 0.8802083432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.043501587584614754,
-      "rewards/tag_count_reward": 0.9609375149011612,
+      "completion_length": 192.5364646911621,
+      "epoch": 0.17997750281214847,
+      "grad_norm": 0.8887334027661878,
+      "kl": 0.33837890625,
+      "learning_rate": 1.9616142853131342e-05,
+      "loss": 0.0245,
+      "reward": 2.5528977513313293,
+      "reward_std": 0.1918736957013607,
+      "rewards/accuracy_reward": 0.5781250176951289,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.018716785591095686,
+      "rewards/tag_count_reward": 0.9986979216337204,
       "step": 120
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 130.14062690734863,
-      "epoch": 0.36281859070464767,
-      "grad_norm": 1.648081178543267,
-      "kl": 0.45947265625,
-      "learning_rate": 1.6104965863848615e-05,
-      "loss": -0.0071,
-      "reward": 2.7758957743644714,
-      "reward_std": 0.303048393689096,
-      "rewards/accuracy_reward": 0.8593750149011612,
-      "rewards/reasoning_steps_reward": 0.9930555671453476,
-      "rewards/repetition_penalty_reward": -0.027055577840656042,
-      "rewards/tag_count_reward": 0.9505208432674408,
+      "completion_length": 184.10416793823242,
+      "epoch": 0.18147731533558306,
+      "grad_norm": 0.945237692955735,
+      "kl": 0.3544921875,
+      "learning_rate": 1.9601618877549113e-05,
+      "loss": 0.0562,
+      "reward": 2.4059919714927673,
+      "reward_std": 0.2110268771648407,
+      "rewards/accuracy_reward": 0.43229167722165585,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.01718513877131045,
+      "rewards/tag_count_reward": 0.99609375,
       "step": 121
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 125.65104675292969,
-      "epoch": 0.3658170914542729,
-      "grad_norm": 1.2095122817688286,
-      "kl": 0.59716796875,
-      "learning_rate": 1.6021412997741994e-05,
-      "loss": -0.0131,
-      "reward": 2.820314347743988,
-      "reward_std": 0.2720828726887703,
-      "rewards/accuracy_reward": 0.8854166865348816,
-      "rewards/reasoning_steps_reward": 0.9843750149011612,
-      "rewards/repetition_penalty_reward": -0.027342125307768583,
-      "rewards/tag_count_reward": 0.977864608168602,
+      "completion_length": 196.5833396911621,
+      "epoch": 0.18297712785901762,
+      "grad_norm": 0.8305782528862219,
+      "kl": 0.35009765625,
+      "learning_rate": 1.9586830789094548e-05,
+      "loss": 0.0135,
+      "reward": 2.6578049659729004,
+      "reward_std": 0.27065123803913593,
+      "rewards/accuracy_reward": 0.6927083432674408,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.023184617515653372,
+      "rewards/tag_count_reward": 0.9934895932674408,
       "step": 122
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 149.05729293823242,
-      "epoch": 0.36881559220389803,
-      "grad_norm": 58.69921900530141,
-      "kl": 5.037109375,
-      "learning_rate": 1.593719539167169e-05,
-      "loss": 0.309,
-      "reward": 2.668303608894348,
-      "reward_std": 0.38671524077653885,
-      "rewards/accuracy_reward": 0.7343750149011612,
-      "rewards/reasoning_steps_reward": 0.986111119389534,
-      "rewards/repetition_penalty_reward": -0.01963039650581777,
-      "rewards/tag_count_reward": 0.9674479365348816,
+      "completion_length": 190.0885467529297,
+      "epoch": 0.1844769403824522,
+      "grad_norm": 0.8502534823244641,
+      "kl": 0.3662109375,
+      "learning_rate": 1.9571778994545356e-05,
+      "loss": -0.0024,
+      "reward": 2.5200935006141663,
+      "reward_std": 0.20754290046170354,
+      "rewards/accuracy_reward": 0.541666679084301,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.012458672048524022,
+      "rewards/tag_count_reward": 0.99609375,
       "step": 123
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 124.10937881469727,
-      "epoch": 0.37181409295352325,
-      "grad_norm": 13.742825745337187,
-      "kl": 0.82275390625,
-      "learning_rate": 1.5852322342925294e-05,
-      "loss": 0.0778,
-      "reward": 2.820654571056366,
-      "reward_std": 0.292279414832592,
-      "rewards/accuracy_reward": 0.8750000149011612,
+      "completion_length": 193.5520896911621,
+      "epoch": 0.18597675290588678,
+      "grad_norm": 0.9748005418103824,
+      "kl": 0.376953125,
+      "learning_rate": 1.9556463907933038e-05,
+      "loss": 0.0298,
+      "reward": 2.628620147705078,
+      "reward_std": 0.28626001439988613,
+      "rewards/accuracy_reward": 0.6562500298023224,
       "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.021793504944071174,
-      "rewards/tag_count_reward": 0.9726562649011612,
+      "rewards/repetition_penalty_reward": -0.01460911170579493,
+      "rewards/tag_count_reward": 0.9921875149011612,
       "step": 124
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 139.92187881469727,
-      "epoch": 0.3748125937031484,
-      "grad_norm": 2.232033331871603,
-      "kl": 0.484375,
-      "learning_rate": 1.5766803221148676e-05,
-      "loss": 0.105,
-      "reward": 2.7213205695152283,
-      "reward_std": 0.32770272716879845,
-      "rewards/accuracy_reward": 0.7760416865348816,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.0208670892752707,
-      "rewards/tag_count_reward": 0.9713541865348816,
+      "completion_length": 192.9114646911621,
+      "epoch": 0.18747656542932134,
+      "grad_norm": 1.0358678308364058,
+      "kl": 0.47607421875,
+      "learning_rate": 1.9540885950531507e-05,
+      "loss": 0.01,
+      "reward": 2.6429734230041504,
+      "reward_std": 0.3361320048570633,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.0232592502143234,
+      "rewards/tag_count_reward": 0.9856770932674408,
       "step": 125
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 123.58854293823242,
-      "epoch": 0.3778110944527736,
-      "grad_norm": 1.0178638016852966,
-      "kl": 0.49951171875,
-      "learning_rate": 1.568064746731156e-05,
-      "loss": 0.011,
-      "reward": 2.6907403469085693,
-      "reward_std": 0.35912006720900536,
-      "rewards/accuracy_reward": 0.7447916865348816,
-      "rewards/reasoning_steps_reward": 0.9843750149011612,
-      "rewards/repetition_penalty_reward": -0.01498887687921524,
-      "rewards/tag_count_reward": 0.9765625149011612,
+      "completion_length": 201.3958396911621,
+      "epoch": 0.1889763779527559,
+      "grad_norm": 1.1639527894884278,
+      "kl": 0.4892578125,
+      "learning_rate": 1.9525045550845482e-05,
+      "loss": 0.0132,
+      "reward": 2.66941100358963,
+      "reward_std": 0.25102632492780685,
+      "rewards/accuracy_reward": 0.7343750149011612,
+      "rewards/reasoning_steps_reward": 0.989583358168602,
+      "rewards/repetition_penalty_reward": -0.03241182304918766,
+      "rewards/tag_count_reward": 0.977864608168602,
       "step": 126
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 119.31250190734863,
-      "epoch": 0.3808095952023988,
-      "grad_norm": 1.1770522092684346,
-      "kl": 0.888671875,
-      "learning_rate": 1.5593864592665333e-05,
-      "loss": -0.0463,
-      "reward": 2.7778323888778687,
-      "reward_std": 0.38024942576885223,
-      "rewards/accuracy_reward": 0.848958358168602,
-      "rewards/reasoning_steps_reward": 0.9756944626569748,
-      "rewards/repetition_penalty_reward": -0.014268482336774468,
-      "rewards/tag_count_reward": 0.9674479365348816,
+      "completion_length": 204.21875762939453,
+      "epoch": 0.19047619047619047,
+      "grad_norm": 2.368193480039129,
+      "kl": 1.4140625,
+      "learning_rate": 1.9508943144598726e-05,
+      "loss": 0.0439,
+      "reward": 2.4418792724609375,
+      "reward_std": 0.3257411792874336,
+      "rewards/accuracy_reward": 0.5260416716337204,
+      "rewards/reasoning_steps_reward": 0.9809028059244156,
+      "rewards/repetition_penalty_reward": -0.020794388838112354,
+      "rewards/tag_count_reward": 0.9557291865348816,
       "step": 127
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 121.88021087646484,
-      "epoch": 0.383808095952024,
-      "grad_norm": 1.1780540061687166,
-      "kl": 0.45703125,
-      "learning_rate": 1.550646417769301e-05,
-      "loss": -0.0279,
-      "reward": 2.692775547504425,
-      "reward_std": 0.4350905865430832,
-      "rewards/accuracy_reward": 0.755208358168602,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.012953592464327812,
-      "rewards/tag_count_reward": 0.9713541865348816,
+      "completion_length": 183.81250762939453,
+      "epoch": 0.19197600299962506,
+      "grad_norm": 11.16928341041206,
+      "kl": 3.96875,
+      "learning_rate": 1.9492579174722043e-05,
+      "loss": 0.0888,
+      "reward": 2.5279927849769592,
+      "reward_std": 0.21637535840272903,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.021920496597886086,
+      "rewards/tag_count_reward": 0.9804687649011612,
       "step": 128
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 113.96354484558105,
-      "epoch": 0.3868065967016492,
-      "grad_norm": 1.695442486056536,
-      "kl": 0.6533203125,
-      "learning_rate": 1.541845587105159e-05,
-      "loss": -0.0497,
-      "reward": 2.6655062437057495,
-      "reward_std": 0.5277450531721115,
-      "rewards/accuracy_reward": 0.7656250298023224,
-      "rewards/reasoning_steps_reward": 0.965277835726738,
-      "rewards/repetition_penalty_reward": -0.014615388121455908,
-      "rewards/tag_count_reward": 0.9492187649011612,
+      "completion_length": 203.21875381469727,
+      "epoch": 0.19347581552305962,
+      "grad_norm": 84.32921564323993,
+      "kl": 12.12060546875,
+      "learning_rate": 1.9475954091341098e-05,
+      "loss": 0.4114,
+      "reward": 2.7423276901245117,
+      "reward_std": 0.2567145712673664,
+      "rewards/accuracy_reward": 0.8125000149011612,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.02937377756461501,
+      "rewards/tag_count_reward": 0.9817708432674408,
       "step": 129
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 119.38020896911621,
-      "epoch": 0.38980509745127434,
-      "grad_norm": 1.53256816336331,
-      "kl": 0.7158203125,
-      "learning_rate": 1.532984938850689e-05,
-      "loss": -0.0148,
-      "reward": 2.6938071250915527,
-      "reward_std": 0.4943315237760544,
-      "rewards/accuracy_reward": 0.7760416865348816,
-      "rewards/reasoning_steps_reward": 0.9756944924592972,
-      "rewards/repetition_penalty_reward": -0.009752006619237363,
-      "rewards/tag_count_reward": 0.9518229365348816,
+      "completion_length": 215.53125762939453,
+      "epoch": 0.19497562804649418,
+      "grad_norm": 0.8646472823122007,
+      "kl": 0.4267578125,
+      "learning_rate": 1.9459068351764032e-05,
+      "loss": 0.0533,
+      "reward": 2.3658406734466553,
+      "reward_std": 0.14664249867200851,
+      "rewards/accuracy_reward": 0.4114583432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.029992765747010708,
+      "rewards/tag_count_reward": 0.9843750149011612,
       "step": 130
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 120.53646087646484,
-      "epoch": 0.39280359820089955,
-      "grad_norm": 2.1945480451348214,
-      "kl": 0.9951171875,
-      "learning_rate": 1.524065451186095e-05,
-      "loss": -0.0313,
-      "reward": 2.531009793281555,
-      "reward_std": 0.5116675943136215,
-      "rewards/accuracy_reward": 0.6302083432674408,
-      "rewards/reasoning_steps_reward": 0.9670139253139496,
-      "rewards/repetition_penalty_reward": -0.016733432421460748,
-      "rewards/tag_count_reward": 0.9505208432674408,
+      "completion_length": 222.41666793823242,
+      "epoch": 0.19647544056992877,
+      "grad_norm": 0.8326546294074292,
+      "kl": 0.48095703125,
+      "learning_rate": 1.94419224204689e-05,
+      "loss": 0.0102,
+      "reward": 2.5197873711586,
+      "reward_std": 0.3018568679690361,
+      "rewards/accuracy_reward": 0.5885416828095913,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.03056000219658017,
+      "rewards/tag_count_reward": 0.973958358168602,
       "step": 131
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 123.78125381469727,
-      "epoch": 0.39580209895052476,
-      "grad_norm": 4.5129778094997155,
-      "kl": 1.5009765625,
-      "learning_rate": 1.5150881087872184e-05,
-      "loss": 0.0509,
-      "reward": 2.7841535210609436,
-      "reward_std": 0.3951072469353676,
-      "rewards/accuracy_reward": 0.8541666865348816,
-      "rewards/reasoning_steps_reward": 0.9774305820465088,
-      "rewards/repetition_penalty_reward": -0.013589567271992564,
-      "rewards/tag_count_reward": 0.966145858168602,
+      "completion_length": 221.04687881469727,
+      "epoch": 0.19797525309336333,
+      "grad_norm": 0.8968322010706142,
+      "kl": 0.5380859375,
+      "learning_rate": 1.9424516769090863e-05,
+      "loss": -0.0605,
+      "reward": 2.4514987468719482,
+      "reward_std": 0.38813910633325577,
+      "rewards/accuracy_reward": 0.546875,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.03070627013221383,
+      "rewards/tag_count_reward": 0.970052108168602,
       "step": 132
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 127.78646087646484,
-      "epoch": 0.3988005997001499,
-      "grad_norm": 38.945437112042384,
-      "kl": 6.130859375,
-      "learning_rate": 1.5060539027168317e-05,
-      "loss": 0.2991,
-      "reward": 2.6191622018814087,
-      "reward_std": 0.470632191747427,
-      "rewards/accuracy_reward": 0.7031250149011612,
-      "rewards/reasoning_steps_reward": 0.9774305671453476,
-      "rewards/repetition_penalty_reward": -0.013216342777013779,
-      "rewards/tag_count_reward": 0.9518229365348816,
+      "completion_length": 216.1197967529297,
+      "epoch": 0.1994750656167979,
+      "grad_norm": 0.720816161798331,
+      "kl": 0.40869140625,
+      "learning_rate": 1.9406851876409254e-05,
+      "loss": -0.0202,
+      "reward": 2.6190152168273926,
+      "reward_std": 0.29873134195804596,
+      "rewards/accuracy_reward": 0.7031250298023224,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.0354987857863307,
+      "rewards/tag_count_reward": 0.9843750149011612,
       "step": 133
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 122.30208587646484,
-      "epoch": 0.4017991004497751,
-      "grad_norm": 11.506961978299184,
-      "kl": 2.6162109375,
-      "learning_rate": 1.4969638303152296e-05,
-      "loss": 0.0406,
-      "reward": 2.60586279630661,
-      "reward_std": 0.49396244436502457,
-      "rewards/accuracy_reward": 0.692708358168602,
-      "rewards/reasoning_steps_reward": 0.9687500596046448,
-      "rewards/repetition_penalty_reward": -0.019137236289680004,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 218.9479217529297,
+      "epoch": 0.20097487814023246,
+      "grad_norm": 0.8292765779680333,
+      "kl": 0.4521484375,
+      "learning_rate": 1.938892822833437e-05,
+      "loss": -0.0563,
+      "reward": 2.720021903514862,
+      "reward_std": 0.4130494073033333,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.02650598995387554,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 134
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 119.50521278381348,
-      "epoch": 0.4047976011994003,
-      "grad_norm": 2.0647869792670006,
-      "kl": 0.6298828125,
-      "learning_rate": 1.4878188950901275e-05,
-      "loss": 0.0324,
-      "reward": 2.8466954231262207,
-      "reward_std": 0.35728102922439575,
-      "rewards/accuracy_reward": 0.9010416865348816,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.011811425443738699,
-      "rewards/tag_count_reward": 0.9713541865348816,
+      "completion_length": 248.96875381469727,
+      "epoch": 0.20247469066366705,
+      "grad_norm": 0.751372103948477,
+      "kl": 0.578125,
+      "learning_rate": 1.9370746317894135e-05,
+      "loss": -0.0156,
+      "reward": 2.5825566053390503,
+      "reward_std": 0.4179469347000122,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.9791666567325592,
+      "rewards/repetition_penalty_reward": -0.03332882048562169,
+      "rewards/tag_count_reward": 0.9804687649011612,
       "step": 135
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 120.55729293823242,
-      "epoch": 0.4077961019490255,
-      "grad_norm": 1.2964235131560873,
-      "kl": 0.46630859375,
-      "learning_rate": 1.4786201066058767e-05,
-      "loss": 0.023,
-      "reward": 2.9288823008537292,
-      "reward_std": 0.20188137842342257,
-      "rewards/accuracy_reward": 0.9531250149011612,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.011221994878724217,
-      "rewards/tag_count_reward": 0.9921875149011612,
+      "completion_length": 247.97917556762695,
+      "epoch": 0.2039745031871016,
+      "grad_norm": 2.8917497533599517,
+      "kl": 0.42822265625,
+      "learning_rate": 1.9352306645220518e-05,
+      "loss": -0.0077,
+      "reward": 2.4527002573013306,
+      "reward_std": 0.44423961639404297,
+      "rewards/accuracy_reward": 0.5312500223517418,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.029070657677948475,
+      "rewards/tag_count_reward": 0.9661458432674408,
       "step": 136
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 121.47917175292969,
-      "epoch": 0.4107946026986507,
-      "grad_norm": 4.427417919453853,
-      "kl": 0.8115234375,
-      "learning_rate": 1.4693684803720139e-05,
-      "loss": 0.0257,
-      "reward": 2.752627432346344,
-      "reward_std": 0.28887180984020233,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 0.9878472238779068,
-      "rewards/repetition_penalty_reward": -0.017771947663277388,
-      "rewards/tag_count_reward": 0.9856770932674408,
+      "completion_length": 244.0677146911621,
+      "epoch": 0.20547431571053618,
+      "grad_norm": 0.7664671252381398,
+      "kl": 0.3759765625,
+      "learning_rate": 1.9333609717535788e-05,
+      "loss": 0.015,
+      "reward": 2.645112633705139,
+      "reward_std": 0.43032628297805786,
+      "rewards/accuracy_reward": 0.723958358168602,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.03500907029956579,
+      "rewards/tag_count_reward": 0.98046875,
       "step": 137
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 116.52604484558105,
-      "epoch": 0.41379310344827586,
-      "grad_norm": 3.187742864763733,
-      "kl": 0.6494140625,
-      "learning_rate": 1.4600650377311523e-05,
-      "loss": 0.0343,
-      "reward": 2.9099594950675964,
-      "reward_std": 0.214247893425636,
-      "rewards/accuracy_reward": 0.9375000149011612,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.011915652547031641,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 236.29687881469727,
+      "epoch": 0.20697412823397077,
+      "grad_norm": 0.9030694840003886,
+      "kl": 0.48681640625,
+      "learning_rate": 1.931465604913856e-05,
+      "loss": -0.0409,
+      "reward": 2.5029749870300293,
+      "reward_std": 0.49775829538702965,
+      "rewards/accuracy_reward": 0.6197916865348816,
+      "rewards/reasoning_steps_reward": 0.9565972685813904,
+      "rewards/repetition_penalty_reward": -0.03565355762839317,
+      "rewards/tag_count_reward": 0.962239608168602,
       "step": 138
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 116.72396087646484,
-      "epoch": 0.41679160419790107,
-      "grad_norm": 1.8697959272406102,
-      "kl": 0.705078125,
-      "learning_rate": 1.4507108057462297e-05,
-      "loss": -0.0029,
-      "reward": 2.6899372935295105,
-      "reward_std": 0.2231074832379818,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9878472238779068,
-      "rewards/repetition_penalty_reward": -0.010149642825126648,
-      "rewards/tag_count_reward": 0.9830729365348816,
+      "completion_length": 248.0729217529297,
+      "epoch": 0.20847394075740533,
+      "grad_norm": 2.8562687232727018,
+      "kl": 3.19384765625,
+      "learning_rate": 1.9295446161389644e-05,
+      "loss": -0.0617,
+      "reward": 2.431870937347412,
+      "reward_std": 0.6133489608764648,
+      "rewards/accuracy_reward": 0.5729166865348816,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.030802744440734386,
+      "rewards/tag_count_reward": 0.9453125149011612,
       "step": 139
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 111.42708778381348,
-      "epoch": 0.4197901049475262,
-      "grad_norm": 23.21366760717317,
-      "kl": 1.552734375,
-      "learning_rate": 1.4413068170871252e-05,
-      "loss": 0.0225,
-      "reward": 2.7771247029304504,
-      "reward_std": 0.2754965058993548,
-      "rewards/accuracy_reward": 0.8229167014360428,
-      "rewards/reasoning_steps_reward": 0.9861111342906952,
-      "rewards/repetition_penalty_reward": -0.008465770864859223,
-      "rewards/tag_count_reward": 0.9765625149011612,
+      "completion_length": 227.87500762939453,
+      "epoch": 0.2099737532808399,
+      "grad_norm": 11.214953837004993,
+      "kl": 1.908203125,
+      "learning_rate": 1.9275980582697707e-05,
+      "loss": -0.079,
+      "reward": 2.5050920844078064,
+      "reward_std": 0.5771966800093651,
+      "rewards/accuracy_reward": 0.6770833432674408,
+      "rewards/reasoning_steps_reward": 0.925347313284874,
+      "rewards/repetition_penalty_reward": -0.032234320882707834,
+      "rewards/tag_count_reward": 0.934895858168602,
       "step": 140
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 116.60937881469727,
-      "epoch": 0.42278860569715143,
-      "grad_norm": 1.2588277588625658,
-      "kl": 0.47802734375,
-      "learning_rate": 1.4318541099166556e-05,
-      "loss": 0.03,
-      "reward": 2.9449267387390137,
-      "reward_std": 0.10798337496817112,
-      "rewards/accuracy_reward": 0.9635416716337204,
-      "rewards/reasoning_steps_reward": 0.9965277761220932,
-      "rewards/repetition_penalty_reward": -0.009934437868651003,
-      "rewards/tag_count_reward": 0.9947916716337204,
+      "completion_length": 244.82813262939453,
+      "epoch": 0.21147356580427445,
+      "grad_norm": 0.8549661150364268,
+      "kl": 0.64990234375,
+      "learning_rate": 1.9256259848504737e-05,
+      "loss": -0.0524,
+      "reward": 2.30450901389122,
+      "reward_std": 0.440264068543911,
+      "rewards/accuracy_reward": 0.432291679084301,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.03359871078282595,
+      "rewards/tag_count_reward": 0.954427108168602,
       "step": 141
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 107.51041793823242,
-      "epoch": 0.4257871064467766,
-      "grad_norm": 1.6180185537864251,
-      "kl": 0.7041015625,
-      "learning_rate": 1.4223537277759667e-05,
-      "loss": -0.0281,
-      "reward": 2.9196689128875732,
-      "reward_std": 0.2676460010698065,
-      "rewards/accuracy_reward": 0.9635416865348816,
-      "rewards/reasoning_steps_reward": 0.9809027910232544,
-      "rewards/repetition_penalty_reward": -0.005244338244665414,
-      "rewards/tag_count_reward": 0.98046875,
+      "completion_length": 231.68750381469727,
+      "epoch": 0.21297337832770905,
+      "grad_norm": 1.9836172250089459,
+      "kl": 1.09423828125,
+      "learning_rate": 1.9236284501271317e-05,
+      "loss": -0.0897,
+      "reward": 2.265152394771576,
+      "reward_std": 0.5349541902542114,
+      "rewards/accuracy_reward": 0.4427083358168602,
+      "rewards/reasoning_steps_reward": 0.9288195073604584,
+      "rewards/repetition_penalty_reward": -0.03215676499530673,
+      "rewards/tag_count_reward": 0.9257812649011612,
       "step": 142
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 106.96354293823242,
-      "epoch": 0.4287856071964018,
-      "grad_norm": 3.173672388607253,
-      "kl": 0.861328125,
-      "learning_rate": 1.4128067194693316e-05,
-      "loss": 0.0193,
-      "reward": 2.866965413093567,
-      "reward_std": 0.36002135276794434,
-      "rewards/accuracy_reward": 0.911458358168602,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.006298626307398081,
-      "rewards/tag_count_reward": 0.9687500149011612,
+      "completion_length": 229.2760467529297,
+      "epoch": 0.2144731908511436,
+      "grad_norm": 2.657790488829322,
+      "kl": 2.5712890625,
+      "learning_rate": 1.9216055090461693e-05,
+      "loss": -0.1423,
+      "reward": 2.298185646533966,
+      "reward_std": 0.6972773224115372,
+      "rewards/accuracy_reward": 0.526041679084301,
+      "rewards/reasoning_steps_reward": 0.911458358168602,
+      "rewards/repetition_penalty_reward": -0.033845747821033,
+      "rewards/tag_count_reward": 0.8945312798023224,
       "step": 143
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 110.04166984558105,
-      "epoch": 0.431784107946027,
-      "grad_norm": 3.72381753576245,
-      "kl": 1.1474609375,
-      "learning_rate": 1.4032141389483648e-05,
-      "loss": 0.0526,
-      "reward": 2.784775972366333,
-      "reward_std": 0.2770255096256733,
-      "rewards/accuracy_reward": 0.817708358168602,
-      "rewards/reasoning_steps_reward": 0.9913194328546524,
-      "rewards/repetition_penalty_reward": -0.007324688020162284,
-      "rewards/tag_count_reward": 0.9830729365348816,
+      "completion_length": 219.67187881469727,
+      "epoch": 0.21597300337457817,
+      "grad_norm": 0.8727878748564548,
+      "kl": 0.5927734375,
+      "learning_rate": 1.9195572172528678e-05,
+      "loss": -0.097,
+      "reward": 2.390109956264496,
+      "reward_std": 0.429065003991127,
+      "rewards/accuracy_reward": 0.557291679084301,
+      "rewards/reasoning_steps_reward": 0.9357638955116272,
+      "rewards/repetition_penalty_reward": -0.03914357628673315,
+      "rewards/tag_count_reward": 0.9361979365348816,
       "step": 144
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 100.05208396911621,
-      "epoch": 0.43478260869565216,
-      "grad_norm": 98.84817232879341,
-      "kl": 7.1748046875,
-      "learning_rate": 1.3935770451956732e-05,
-      "loss": 0.2566,
-      "reward": 2.7451387643814087,
-      "reward_std": 0.23619456216692924,
-      "rewards/accuracy_reward": 0.770833358168602,
-      "rewards/reasoning_steps_reward": 0.9947916865348816,
-      "rewards/repetition_penalty_reward": -0.006163446931168437,
-      "rewards/tag_count_reward": 0.9856770932674408,
+      "completion_length": 231.10937881469727,
+      "epoch": 0.21747281589801276,
+      "grad_norm": 0.7543285200149539,
+      "kl": 0.5830078125,
+      "learning_rate": 1.9174836310898334e-05,
+      "loss": -0.0843,
+      "reward": 2.436844766139984,
+      "reward_std": 0.527190275490284,
+      "rewards/accuracy_reward": 0.604166679084301,
+      "rewards/reasoning_steps_reward": 0.9618056118488312,
+      "rewards/repetition_penalty_reward": -0.04188787518069148,
+      "rewards/tag_count_reward": 0.9127604216337204,
       "step": 145
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 95.88541984558105,
-      "epoch": 0.43778110944527737,
-      "grad_norm": 1.6790256708529574,
-      "kl": 0.71337890625,
-      "learning_rate": 1.3838965021079447e-05,
-      "loss": 0.022,
-      "reward": 2.582995355129242,
-      "reward_std": 0.33666881918907166,
-      "rewards/accuracy_reward": 0.630208358168602,
-      "rewards/reasoning_steps_reward": 0.9756944477558136,
-      "rewards/repetition_penalty_reward": -0.0033762191596906632,
-      "rewards/tag_count_reward": 0.9804687649011612,
+      "completion_length": 229.2083396911621,
+      "epoch": 0.21897262842144732,
+      "grad_norm": 0.639298152517526,
+      "kl": 0.673828125,
+      "learning_rate": 1.9153848075954465e-05,
+      "loss": -0.1119,
+      "reward": 2.3503913283348083,
+      "reward_std": 0.4797208532691002,
+      "rewards/accuracy_reward": 0.5416666679084301,
+      "rewards/reasoning_steps_reward": 0.954861119389534,
+      "rewards/repetition_penalty_reward": -0.06150110438466072,
+      "rewards/tag_count_reward": 0.915364608168602,
       "step": 146
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 99.12500190734863,
-      "epoch": 0.4407796101949025,
-      "grad_norm": 1.000815515911012,
-      "kl": 0.4951171875,
-      "learning_rate": 1.3741735783785022e-05,
-      "loss": 0.014,
-      "reward": 2.7670981884002686,
-      "reward_std": 0.24415395595133305,
-      "rewards/accuracy_reward": 0.7864583432674408,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.0037352032377384603,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 237.23958587646484,
+      "epoch": 0.2204724409448819,
+      "grad_norm": 3.076436673284199,
+      "kl": 0.697265625,
+      "learning_rate": 1.9132608045022954e-05,
+      "loss": -0.0704,
+      "reward": 2.4564713835716248,
+      "reward_std": 0.4616401568055153,
+      "rewards/accuracy_reward": 0.6197916716337204,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.0834592841565609,
+      "rewards/tag_count_reward": 0.9375000149011612,
       "step": 147
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 90.77604293823242,
-      "epoch": 0.44377811094452774,
-      "grad_norm": 10.482881482497314,
-      "kl": 1.677734375,
-      "learning_rate": 1.3644093473793213e-05,
-      "loss": 0.0242,
-      "reward": 2.8607059121131897,
-      "reward_std": 0.39441923797130585,
-      "rewards/accuracy_reward": 0.9166667014360428,
-      "rewards/reasoning_steps_reward": 0.987847238779068,
-      "rewards/repetition_penalty_reward": -0.004745513964735437,
-      "rewards/tag_count_reward": 0.9609375149011612,
+      "completion_length": 217.3697967529297,
+      "epoch": 0.22197225346831645,
+      "grad_norm": 0.7063771066104143,
+      "kl": 0.52490234375,
+      "learning_rate": 1.9111116802355853e-05,
+      "loss": -0.0912,
+      "reward": 2.410871744155884,
+      "reward_std": 0.5071974396705627,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.060916513204574585,
+      "rewards/tag_count_reward": 0.923177108168602,
       "step": 148
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 98.33333587646484,
-      "epoch": 0.44677661169415295,
-      "grad_norm": 3.0619014190201397,
-      "kl": 0.70263671875,
-      "learning_rate": 1.3546048870425356e-05,
-      "loss": -0.0034,
-      "reward": 2.7546103596687317,
-      "reward_std": 0.3782291766256094,
-      "rewards/accuracy_reward": 0.802083358168602,
-      "rewards/reasoning_steps_reward": 0.9826389104127884,
-      "rewards/repetition_penalty_reward": -0.007976488093845546,
-      "rewards/tag_count_reward": 0.9778645932674408,
+      "completion_length": 220.9895896911621,
+      "epoch": 0.22347206599175104,
+      "grad_norm": 0.759921740363777,
+      "kl": 0.7099609375,
+      "learning_rate": 1.9089374939115335e-05,
+      "loss": -0.1306,
+      "reward": 2.3488033413887024,
+      "reward_std": 0.6183192133903503,
+      "rewards/accuracy_reward": 0.5677083507180214,
+      "rewards/reasoning_steps_reward": 0.939236119389534,
+      "rewards/repetition_penalty_reward": -0.06569322943687439,
+      "rewards/tag_count_reward": 0.907552108168602,
       "step": 149
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 96.68750381469727,
-      "epoch": 0.4497751124437781,
-      "grad_norm": 1.757178598271745,
-      "kl": 0.6220703125,
-      "learning_rate": 1.3447612797414371e-05,
-      "loss": -0.0372,
-      "reward": 2.696329712867737,
-      "reward_std": 0.48958031833171844,
-      "rewards/accuracy_reward": 0.78125,
-      "rewards/reasoning_steps_reward": 0.9722222685813904,
-      "rewards/repetition_penalty_reward": -0.003757232567295432,
-      "rewards/tag_count_reward": 0.9466146230697632,
+      "completion_length": 217.5833396911621,
+      "epoch": 0.2249718785151856,
+      "grad_norm": 5.333380760593154,
+      "kl": 1.005859375,
+      "learning_rate": 1.906738305335741e-05,
+      "loss": -0.1023,
+      "reward": 2.319095730781555,
+      "reward_std": 0.6061732321977615,
+      "rewards/accuracy_reward": 0.5572916865348816,
+      "rewards/reasoning_steps_reward": 0.9461806118488312,
+      "rewards/repetition_penalty_reward": -0.09062658622860909,
+      "rewards/tag_count_reward": 0.9062500149011612,
       "step": 150
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 96.88021087646484,
-      "epoch": 0.4527736131934033,
-      "grad_norm": 1.0864550515764144,
-      "kl": 0.47802734375,
-      "learning_rate": 1.3348796121709862e-05,
-      "loss": -0.0041,
-      "reward": 2.914130389690399,
-      "reward_std": 0.26327061653137207,
-      "rewards/accuracy_reward": 0.942708358168602,
-      "rewards/reasoning_steps_reward": 0.991319477558136,
-      "rewards/repetition_penalty_reward": -0.004272434976883233,
-      "rewards/tag_count_reward": 0.9843750149011612,
+      "completion_length": 221.75000762939453,
+      "epoch": 0.22647169103862017,
+      "grad_norm": 1.2824333531797762,
+      "kl": 2.69970703125,
+      "learning_rate": 1.90451417500155e-05,
+      "loss": -0.091,
+      "reward": 2.622887670993805,
+      "reward_std": 0.5151065196841955,
+      "rewards/accuracy_reward": 0.8072916865348816,
+      "rewards/reasoning_steps_reward": 0.949652835726738,
+      "rewards/repetition_penalty_reward": -0.06895278673619032,
+      "rewards/tag_count_reward": 0.934895858168602,
       "step": 151
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 98.10416793823242,
-      "epoch": 0.45577211394302847,
-      "grad_norm": 1.1731563120826676,
-      "kl": 0.47802734375,
-      "learning_rate": 1.3249609752278454e-05,
-      "loss": 0.006,
-      "reward": 2.5750681161880493,
-      "reward_std": 0.31880153343081474,
-      "rewards/accuracy_reward": 0.6093750074505806,
-      "rewards/reasoning_steps_reward": 0.9930555671453476,
-      "rewards/repetition_penalty_reward": -0.006529179809149355,
-      "rewards/tag_count_reward": 0.9791666716337204,
+      "completion_length": 220.57292556762695,
+      "epoch": 0.22797150356205476,
+      "grad_norm": 0.7007261870507696,
+      "kl": 0.51171875,
+      "learning_rate": 1.902265164088378e-05,
+      "loss": -0.0408,
+      "reward": 2.4209659099578857,
+      "reward_std": 0.49653469771146774,
+      "rewards/accuracy_reward": 0.5729166865348816,
+      "rewards/reasoning_steps_reward": 0.97743059694767,
+      "rewards/repetition_penalty_reward": -0.08120433799922466,
+      "rewards/tag_count_reward": 0.9518229365348816,
       "step": 152
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 92.29166984558105,
-      "epoch": 0.4587706146926537,
-      "grad_norm": 1.1346690177946404,
-      "kl": 0.45361328125,
-      "learning_rate": 1.315006463889948e-05,
-      "loss": -0.0176,
-      "reward": 2.9095540046691895,
-      "reward_std": 0.2783464193344116,
-      "rewards/accuracy_reward": 0.9531250298023224,
-      "rewards/reasoning_steps_reward": 0.9826389104127884,
-      "rewards/repetition_penalty_reward": -0.0027724849642254412,
-      "rewards/tag_count_reward": 0.9765625149011612,
+      "completion_length": 232.13541793823242,
+      "epoch": 0.22947131608548932,
+      "grad_norm": 0.65651651084759,
+      "kl": 0.345703125,
+      "learning_rate": 1.899991334460036e-05,
+      "loss": 0.0085,
+      "reward": 2.626923680305481,
+      "reward_std": 0.23435234278440475,
+      "rewards/accuracy_reward": 0.7343750298023224,
+      "rewards/reasoning_steps_reward": 0.9913194626569748,
+      "rewards/repetition_penalty_reward": -0.08835414797067642,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 153
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 96.83333587646484,
-      "epoch": 0.4617691154422789,
-      "grad_norm": 1.1933914174647031,
-      "kl": 0.48486328125,
-      "learning_rate": 1.3050171770956176e-05,
-      "loss": 0.0015,
-      "reward": 2.72132408618927,
-      "reward_std": 0.32843077182769775,
-      "rewards/accuracy_reward": 0.7656250298023224,
-      "rewards/reasoning_steps_reward": 0.9895833432674408,
-      "rewards/repetition_penalty_reward": -0.009144885872956365,
-      "rewards/tag_count_reward": 0.9752604365348816,
+      "completion_length": 218.23437881469727,
+      "epoch": 0.23097112860892388,
+      "grad_norm": 0.7528729019247309,
+      "kl": 0.63623046875,
+      "learning_rate": 1.8976927486630252e-05,
+      "loss": 0.013,
+      "reward": 2.2777963876724243,
+      "reward_std": 0.3943594992160797,
+      "rewards/accuracy_reward": 0.4218750149011612,
+      "rewards/reasoning_steps_reward": 0.9704861044883728,
+      "rewards/repetition_penalty_reward": -0.08722100034356117,
+      "rewards/tag_count_reward": 0.97265625,
       "step": 154
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 99.83333587646484,
-      "epoch": 0.46476761619190404,
-      "grad_norm": 1.2021972493346804,
-      "kl": 0.47314453125,
-      "learning_rate": 1.2949942176222497e-05,
-      "loss": -0.0459,
-      "reward": 2.7491848468780518,
-      "reward_std": 0.4561289846897125,
-      "rewards/accuracy_reward": 0.8229167014360428,
-      "rewards/reasoning_steps_reward": 0.9791666716337204,
-      "rewards/repetition_penalty_reward": -0.0034193213214166462,
-      "rewards/tag_count_reward": 0.9505208432674408,
+      "completion_length": 210.3385467529297,
+      "epoch": 0.23247094113235844,
+      "grad_norm": 9.645470554535006,
+      "kl": 1.740234375,
+      "learning_rate": 1.8953694699248193e-05,
+      "loss": -0.0036,
+      "reward": 2.56173574924469,
+      "reward_std": 0.3356763105839491,
+      "rewards/accuracy_reward": 0.6927083507180214,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.09060808829963207,
+      "rewards/tag_count_reward": 0.9752604365348816,
       "step": 155
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 106.64062690734863,
-      "epoch": 0.46776611694152925,
-      "grad_norm": 1.3131511244620084,
-      "kl": 0.56982421875,
-      "learning_rate": 1.2849386919645686e-05,
-      "loss": -0.0388,
-      "reward": 2.743965268135071,
-      "reward_std": 0.39171791821718216,
-      "rewards/accuracy_reward": 0.8281250149011612,
-      "rewards/reasoning_steps_reward": 0.9826389104127884,
-      "rewards/repetition_penalty_reward": -0.010809163737576455,
-      "rewards/tag_count_reward": 0.9440104216337204,
+      "completion_length": 206.6614646911621,
+      "epoch": 0.23397075365579303,
+      "grad_norm": 0.7853069216863842,
+      "kl": 0.40625,
+      "learning_rate": 1.893021562152122e-05,
+      "loss": 0.0268,
+      "reward": 2.8448190689086914,
+      "reward_std": 0.22713683173060417,
+      "rewards/accuracy_reward": 0.9531250149011612,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.05405241996049881,
+      "rewards/tag_count_reward": 0.9596354365348816,
       "step": 156
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 103.53125190734863,
-      "epoch": 0.4707646176911544,
-      "grad_norm": 1.69232758602469,
-      "kl": 0.56396484375,
-      "learning_rate": 1.2748517102124755e-05,
-      "loss": -0.0245,
-      "reward": 2.7549294233322144,
-      "reward_std": 0.3383819945156574,
-      "rewards/accuracy_reward": 0.817708358168602,
-      "rewards/reasoning_steps_reward": 0.9826389402151108,
-      "rewards/repetition_penalty_reward": -0.006355439778417349,
-      "rewards/tag_count_reward": 0.9609375149011612,
+      "completion_length": 181.40625762939453,
+      "epoch": 0.2354705661792276,
+      "grad_norm": 2.706381634571779,
+      "kl": 0.9560546875,
+      "learning_rate": 1.8906490899291125e-05,
+      "loss": 0.0151,
+      "reward": 2.746555209159851,
+      "reward_std": 0.21183432638645172,
+      "rewards/accuracy_reward": 0.8125000149011612,
+      "rewards/reasoning_steps_reward": 0.9930555522441864,
+      "rewards/repetition_penalty_reward": -0.0498858280479908,
+      "rewards/tag_count_reward": 0.9908854365348816,
       "step": 157
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 102.05208587646484,
-      "epoch": 0.4737631184407796,
-      "grad_norm": 2.265769890602018,
-      "kl": 0.73681640625,
-      "learning_rate": 1.2647343859284997e-05,
-      "loss": -0.0689,
-      "reward": 2.662193477153778,
-      "reward_std": 0.4981583207845688,
-      "rewards/accuracy_reward": 0.786458358168602,
-      "rewards/reasoning_steps_reward": 0.9722222536802292,
-      "rewards/repetition_penalty_reward": -0.011851702001877129,
-      "rewards/tag_count_reward": 0.9153645932674408,
+      "completion_length": 176.8020896911621,
+      "epoch": 0.23697037870266216,
+      "grad_norm": 1.1526188899297036,
+      "kl": 0.47314453125,
+      "learning_rate": 1.888252118515666e-05,
+      "loss": -0.0084,
+      "reward": 2.248090386390686,
+      "reward_std": 0.29650406911969185,
+      "rewards/accuracy_reward": 0.33333334140479565,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.03576376847922802,
+      "rewards/tag_count_reward": 0.9817708432674408,
       "step": 158
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 103.22916984558105,
-      "epoch": 0.4767616191904048,
-      "grad_norm": 2.255759107812268,
-      "kl": 0.79541015625,
-      "learning_rate": 1.2545878360248633e-05,
-      "loss": 0.0144,
-      "reward": 2.8301729559898376,
-      "reward_std": 0.3695299196988344,
-      "rewards/accuracy_reward": 0.9010416865348816,
-      "rewards/reasoning_steps_reward": 0.9982638955116272,
-      "rewards/repetition_penalty_reward": -0.009236796642653644,
-      "rewards/tag_count_reward": 0.9401041865348816,
+      "completion_length": 167.71875381469727,
+      "epoch": 0.23847019122609675,
+      "grad_norm": 0.8634079765155277,
+      "kl": 0.38037109375,
+      "learning_rate": 1.88583071384556e-05,
+      "loss": 0.015,
+      "reward": 2.783277213573456,
+      "reward_std": 0.27950316295027733,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03746931627392769,
+      "rewards/tag_count_reward": 0.99609375,
       "step": 159
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 100.39062881469727,
-      "epoch": 0.47976011994003,
-      "grad_norm": 7.190371202010685,
-      "kl": 1.99609375,
-      "learning_rate": 1.2444131806401818e-05,
-      "loss": 0.0233,
-      "reward": 2.587311804294586,
-      "reward_std": 0.4335063770413399,
-      "rewards/accuracy_reward": 0.6875000298023224,
-      "rewards/reasoning_steps_reward": 0.9722222536802292,
-      "rewards/repetition_penalty_reward": -0.007306311279535294,
-      "rewards/tag_count_reward": 0.9348958432674408,
+      "completion_length": 159.5208396911621,
+      "epoch": 0.2399700037495313,
+      "grad_norm": 0.9199485039081772,
+      "kl": 0.43212890625,
+      "learning_rate": 1.883384942524661e-05,
+      "loss": 0.0147,
+      "reward": 2.60649836063385,
+      "reward_std": 0.3520687147974968,
+      "rewards/accuracy_reward": 0.6822916865348816,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.03456068178638816,
+      "rewards/tag_count_reward": 0.9830729365348816,
       "step": 160
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 106.15104484558105,
-      "epoch": 0.4827586206896552,
-      "grad_norm": 11.57931701299926,
-      "kl": 2.580078125,
-      "learning_rate": 1.2342115430158024e-05,
-      "loss": 0.0864,
-      "reward": 2.5142497420310974,
-      "reward_std": 0.592901311814785,
-      "rewards/accuracy_reward": 0.6510417014360428,
-      "rewards/reasoning_steps_reward": 0.9722222238779068,
-      "rewards/repetition_penalty_reward": -0.012660016654990613,
-      "rewards/tag_count_reward": 0.903645858168602,
+      "completion_length": 157.79687881469727,
+      "epoch": 0.24146981627296588,
+      "grad_norm": 0.9081570917337833,
+      "kl": 0.4970703125,
+      "learning_rate": 1.880914871829092e-05,
+      "loss": 0.003,
+      "reward": 2.5711691975593567,
+      "reward_std": 0.3647758923470974,
+      "rewards/accuracy_reward": 0.6510416865348816,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.034733695443719625,
+      "rewards/tag_count_reward": 0.9791666716337204,
       "step": 161
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 102.13541793823242,
-      "epoch": 0.48575712143928035,
-      "grad_norm": 4.161233572872993,
-      "kl": 1.421875,
-      "learning_rate": 1.223984049371805e-05,
-      "loss": 0.0328,
-      "reward": 2.5102869868278503,
-      "reward_std": 0.34895218163728714,
-      "rewards/accuracy_reward": 0.5937500223517418,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.0066401001531630754,
-      "rewards/tag_count_reward": 0.9440104365348816,
+      "completion_length": 160.578125,
+      "epoch": 0.24296962879640044,
+      "grad_norm": 0.7819084575897781,
+      "kl": 0.42578125,
+      "learning_rate": 1.8784205697033803e-05,
+      "loss": 0.0206,
+      "reward": 2.7083481550216675,
+      "reward_std": 0.23164555057883263,
+      "rewards/accuracy_reward": 0.770833358168602,
+      "rewards/reasoning_steps_reward": 0.9913194477558136,
+      "rewards/repetition_penalty_reward": -0.04208600614219904,
+      "rewards/tag_count_reward": 0.98828125,
       "step": 162
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 100.828125,
-      "epoch": 0.48875562218890556,
-      "grad_norm": 3.9809594705586053,
-      "kl": 1.162109375,
-      "learning_rate": 1.2137318287826699e-05,
-      "loss": -0.0078,
-      "reward": 2.5699517726898193,
-      "reward_std": 0.4455106034874916,
-      "rewards/accuracy_reward": 0.6770833507180214,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.015117747941985726,
-      "rewards/tag_count_reward": 0.942708358168602,
+      "completion_length": 152.55208587646484,
+      "epoch": 0.24446944131983503,
+      "grad_norm": 1.4454320643452467,
+      "kl": 1.4462890625,
+      "learning_rate": 1.875902104758592e-05,
+      "loss": -0.0319,
+      "reward": 2.8053765892982483,
+      "reward_std": 0.314240001142025,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.028824928449466825,
+      "rewards/tag_count_reward": 0.9817708432674408,
       "step": 163
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 102.94270896911621,
-      "epoch": 0.4917541229385307,
-      "grad_norm": 19.775528338909474,
-      "kl": 2.1142578125,
-      "learning_rate": 1.2034560130526341e-05,
-      "loss": 0.0267,
-      "reward": 2.7591124176979065,
-      "reward_std": 0.3192651905119419,
-      "rewards/accuracy_reward": 0.8281250298023224,
-      "rewards/reasoning_steps_reward": 0.9687500447034836,
-      "rewards/repetition_penalty_reward": -0.006512661639135331,
-      "rewards/tag_count_reward": 0.9687500149011612,
+      "completion_length": 141.1666717529297,
+      "epoch": 0.2459692538432696,
+      "grad_norm": 16.262069244582136,
+      "kl": 6.43603515625,
+      "learning_rate": 1.873359546270442e-05,
+      "loss": 0.0621,
+      "reward": 2.3994747400283813,
+      "reward_std": 0.2881123125553131,
+      "rewards/accuracy_reward": 0.4583333358168602,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.02630652580410242,
+      "rewards/tag_count_reward": 0.9830729365348816,
       "step": 164
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 106.59896278381348,
-      "epoch": 0.4947526236881559,
-      "grad_norm": 1.1276005185146887,
-      "kl": 0.5009765625,
-      "learning_rate": 1.1931577365907433e-05,
-      "loss": 0.0095,
-      "reward": 2.7071834206581116,
-      "reward_std": 0.24475602060556412,
-      "rewards/accuracy_reward": 0.7500000149011612,
-      "rewards/reasoning_steps_reward": 0.9826388955116272,
-      "rewards/repetition_penalty_reward": -0.009830599068664014,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 152.3072967529297,
+      "epoch": 0.24746906636670415,
+      "grad_norm": 29.040919544310924,
+      "kl": 5.2421875,
+      "learning_rate": 1.8707929641773876e-05,
+      "loss": 0.2231,
+      "reward": 2.7308311462402344,
+      "reward_std": 0.3806469663977623,
+      "rewards/accuracy_reward": 0.8125000149011612,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.02481132885441184,
+      "rewards/tag_count_reward": 0.9726562649011612,
       "step": 165
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 103.80729675292969,
-      "epoch": 0.49775112443778113,
-      "grad_norm": 1.4396819065845348,
-      "kl": 0.5517578125,
-      "learning_rate": 1.1828381362856195e-05,
-      "loss": 0.0061,
-      "reward": 2.7072086334228516,
-      "reward_std": 0.2500213086605072,
-      "rewards/accuracy_reward": 0.7552083432674408,
+      "completion_length": 137.9166717529297,
+      "epoch": 0.24896887889013875,
+      "grad_norm": 4.111363572085073,
+      "kl": 1.9384765625,
+      "learning_rate": 1.8682024290787092e-05,
+      "loss": -0.0177,
+      "reward": 2.639627754688263,
+      "reward_std": 0.49519332498311996,
+      "rewards/accuracy_reward": 0.723958358168602,
       "rewards/reasoning_steps_reward": 0.9809028059244156,
-      "rewards/repetition_penalty_reward": -0.008069265051744878,
-      "rewards/tag_count_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.022264669416472316,
+      "rewards/tag_count_reward": 0.9570312649011612,
       "step": 166
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 105.23437690734863,
-      "epoch": 0.5007496251874063,
-      "grad_norm": 1.4162738213578512,
-      "kl": 0.58447265625,
-      "learning_rate": 1.1724983513799505e-05,
-      "loss": -0.028,
-      "reward": 2.83200603723526,
-      "reward_std": 0.2547433339059353,
-      "rewards/accuracy_reward": 0.880208358168602,
-      "rewards/reasoning_steps_reward": 0.9774305820465088,
-      "rewards/repetition_penalty_reward": -0.006101653270889074,
-      "rewards/tag_count_reward": 0.98046875,
+      "completion_length": 143.21875762939453,
+      "epoch": 0.2504686914135733,
+      "grad_norm": 2.038792986981373,
+      "kl": 1.0732421875,
+      "learning_rate": 1.8655880122325633e-05,
+      "loss": -0.0366,
+      "reward": 2.5853430032730103,
+      "reward_std": 0.3307228982448578,
+      "rewards/accuracy_reward": 0.6770833432674408,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.020993857644498348,
+      "rewards/tag_count_reward": 0.9570312649011612,
       "step": 167
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 102.34896087646484,
-      "epoch": 0.5037481259370314,
-      "grad_norm": 1.5741520276879495,
-      "kl": 0.5859375,
-      "learning_rate": 1.1621395233447247e-05,
-      "loss": -0.0284,
-      "reward": 2.856131374835968,
-      "reward_std": 0.336722657084465,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.970486119389534,
-      "rewards/repetition_penalty_reward": -0.00628192734438926,
-      "rewards/tag_count_reward": 0.9752604216337204,
+      "completion_length": 137.98438262939453,
+      "epoch": 0.25196850393700787,
+      "grad_norm": 1.1481401231898494,
+      "kl": 1.109375,
+      "learning_rate": 1.862949785554025e-05,
+      "loss": -0.0102,
+      "reward": 2.74522465467453,
+      "reward_std": 0.24203419662080705,
+      "rewards/accuracy_reward": 0.7864583432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.013021895661950111,
+      "rewards/tag_count_reward": 0.9856770932674408,
       "step": 168
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 103.50521087646484,
-      "epoch": 0.5067466266866567,
-      "grad_norm": 3.143223791327549,
-      "kl": 0.8681640625,
-      "learning_rate": 1.1517627957532155e-05,
-      "loss": -0.0218,
-      "reward": 2.8190484642982483,
-      "reward_std": 0.3480563126504421,
-      "rewards/accuracy_reward": 0.927083358168602,
-      "rewards/reasoning_steps_reward": 0.9409722536802292,
-      "rewards/repetition_penalty_reward": -0.012548819300718606,
-      "rewards/tag_count_reward": 0.9635416865348816,
+      "completion_length": 139.8385467529297,
+      "epoch": 0.25346831646044243,
+      "grad_norm": 0.9880924032467467,
+      "kl": 0.47119140625,
+      "learning_rate": 1.8602878216131093e-05,
+      "loss": -0.0024,
+      "reward": 2.5487694144248962,
+      "reward_std": 0.3845925033092499,
+      "rewards/accuracy_reward": 0.5989583432674408,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.016334847081452608,
+      "rewards/tag_count_reward": 0.9765625,
       "step": 169
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 98.671875,
-      "epoch": 0.5097451274362819,
-      "grad_norm": 4.087516775956481,
-      "kl": 1.3857421875,
-      "learning_rate": 1.1413693141547354e-05,
-      "loss": -0.0334,
-      "reward": 2.7752469778060913,
-      "reward_std": 0.4824451133608818,
-      "rewards/accuracy_reward": 0.9010416865348816,
-      "rewards/reasoning_steps_reward": 0.9531250149011612,
-      "rewards/repetition_penalty_reward": -0.009909308631904423,
-      "rewards/tag_count_reward": 0.9309895932674408,
+      "completion_length": 144.5000057220459,
+      "epoch": 0.254968128983877,
+      "grad_norm": 1.1326948326278583,
+      "kl": 0.64599609375,
+      "learning_rate": 1.8576021936327747e-05,
+      "loss": -0.0373,
+      "reward": 2.54432612657547,
+      "reward_std": 0.4256473332643509,
+      "rewards/accuracy_reward": 0.598958358168602,
+      "rewards/reasoning_steps_reward": 0.9809028059244156,
+      "rewards/repetition_penalty_reward": -0.016003758762963116,
+      "rewards/tag_count_reward": 0.9804687649011612,
       "step": 170
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 108.43750190734863,
-      "epoch": 0.512743628185907,
-      "grad_norm": 1.808660970560849,
-      "kl": 1.2373046875,
-      "learning_rate": 1.1309602259481726e-05,
-      "loss": -0.0589,
-      "reward": 2.6380687952041626,
-      "reward_std": 0.5042757764458656,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.97743059694767,
-      "rewards/repetition_penalty_reward": -0.013841008301824331,
-      "rewards/tag_count_reward": 0.9453125,
+      "completion_length": 141.0885467529297,
+      "epoch": 0.25646794150731156,
+      "grad_norm": 0.907847690585872,
+      "kl": 0.45703125,
+      "learning_rate": 1.8548929754869095e-05,
+      "loss": -0.036,
+      "reward": 2.566925883293152,
+      "reward_std": 0.34289751946926117,
+      "rewards/accuracy_reward": 0.619791679084301,
+      "rewards/reasoning_steps_reward": 0.9809027910232544,
+      "rewards/repetition_penalty_reward": -0.022049905732274055,
+      "rewards/tag_count_reward": 0.9882812649011612,
       "step": 171
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 107.00000190734863,
-      "epoch": 0.5157421289355323,
-      "grad_norm": 3.2045053936261754,
-      "kl": 1.26953125,
-      "learning_rate": 1.1205366802553231e-05,
-      "loss": -0.0773,
-      "reward": 2.749893009662628,
-      "reward_std": 0.5450254082679749,
-      "rewards/accuracy_reward": 0.8437500298023224,
-      "rewards/reasoning_steps_reward": 0.9670138955116272,
-      "rewards/repetition_penalty_reward": -0.007485455018468201,
-      "rewards/tag_count_reward": 0.9466145932674408,
+      "completion_length": 145.02605056762695,
+      "epoch": 0.2579677540307462,
+      "grad_norm": 4.163866220713108,
+      "kl": 1.8134765625,
+      "learning_rate": 1.8521602416982998e-05,
+      "loss": -0.0077,
+      "reward": 2.6473976969718933,
+      "reward_std": 0.26540281693451107,
+      "rewards/accuracy_reward": 0.7135416716337204,
+      "rewards/reasoning_steps_reward": 0.9739583432674408,
+      "rewards/repetition_penalty_reward": -0.01926900539547205,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 172
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 104.08854484558105,
-      "epoch": 0.5187406296851574,
-      "grad_norm": 4.461811464537438,
-      "kl": 1.9453125,
-      "learning_rate": 1.1100998277940316e-05,
-      "loss": -0.0487,
-      "reward": 2.8390066623687744,
-      "reward_std": 0.4234722927212715,
-      "rewards/accuracy_reward": 0.927083358168602,
-      "rewards/reasoning_steps_reward": 0.9687500149011612,
-      "rewards/repetition_penalty_reward": -0.007347506354562938,
-      "rewards/tag_count_reward": 0.9505208432674408,
+      "completion_length": 151.84896087646484,
+      "epoch": 0.25946756655418074,
+      "grad_norm": 0.7222410832833253,
+      "kl": 0.3779296875,
+      "learning_rate": 1.8494040674365785e-05,
+      "loss": -0.0015,
+      "reward": 2.90548974275589,
+      "reward_std": 0.15629624016582966,
+      "rewards/accuracy_reward": 0.9375000149011612,
+      "rewards/reasoning_steps_reward": 0.9913194626569748,
+      "rewards/repetition_penalty_reward": -0.01812157710082829,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 173
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 110.59895896911621,
-      "epoch": 0.5217391304347826,
-      "grad_norm": 1.1431834037063382,
-      "kl": 0.70361328125,
-      "learning_rate": 1.0996508207511565e-05,
-      "loss": -0.0495,
-      "reward": 2.7981215715408325,
-      "reward_std": 0.32188424095511436,
-      "rewards/accuracy_reward": 0.880208358168602,
-      "rewards/reasoning_steps_reward": 0.9809027910232544,
-      "rewards/repetition_penalty_reward": -0.02262500289361924,
-      "rewards/tag_count_reward": 0.9596354216337204,
+      "completion_length": 167.0052146911621,
+      "epoch": 0.2609673790776153,
+      "grad_norm": 0.7895271947524055,
+      "kl": 0.3896484375,
+      "learning_rate": 1.8466245285161593e-05,
+      "loss": -0.0036,
+      "reward": 2.6482080221176147,
+      "reward_std": 0.2084397617727518,
+      "rewards/accuracy_reward": 0.6822916865348816,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.02236495795659721,
+      "rewards/tag_count_reward": 0.9934895932674408,
       "step": 174
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 116.19270896911621,
-      "epoch": 0.5247376311844077,
-      "grad_norm": 6.089832729503498,
-      "kl": 1.47265625,
-      "learning_rate": 1.089190812655374e-05,
-      "loss": 0.0148,
-      "reward": 2.733491837978363,
-      "reward_std": 0.37758616358041763,
-      "rewards/accuracy_reward": 0.8020833432674408,
-      "rewards/reasoning_steps_reward": 0.9843750298023224,
-      "rewards/repetition_penalty_reward": -0.019112566020339727,
-      "rewards/tag_count_reward": 0.9661458432674408,
+      "completion_length": 174.53125762939453,
+      "epoch": 0.26246719160104987,
+      "grad_norm": 0.8431262926543861,
+      "kl": 0.537109375,
+      "learning_rate": 1.8438217013941494e-05,
+      "loss": -0.0388,
+      "reward": 2.708277404308319,
+      "reward_std": 0.4050016924738884,
+      "rewards/accuracy_reward": 0.7812500149011612,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.03130600415170193,
+      "rewards/tag_count_reward": 0.9791666716337204,
       "step": 175
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 122.48437690734863,
-      "epoch": 0.527736131934033,
-      "grad_norm": 1.0305128960164511,
-      "kl": 0.55419921875,
-      "learning_rate": 1.0787209582498315e-05,
-      "loss": 0.0177,
-      "reward": 2.8952815532684326,
-      "reward_std": 0.21657648496329784,
-      "rewards/accuracy_reward": 0.9322916865348816,
-      "rewards/reasoning_steps_reward": 0.9947916865348816,
-      "rewards/repetition_penalty_reward": -0.0161769341211766,
-      "rewards/tag_count_reward": 0.9843750149011612,
+      "completion_length": 179.11979293823242,
+      "epoch": 0.26396700412448443,
+      "grad_norm": 0.8034495401351996,
+      "kl": 0.494140625,
+      "learning_rate": 1.8409956631682475e-05,
+      "loss": -0.0131,
+      "reward": 2.5661210417747498,
+      "reward_std": 0.3050253167748451,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9843750298023224,
+      "rewards/repetition_penalty_reward": -0.023722857236862183,
+      "rewards/tag_count_reward": 0.9804687649011612,
       "step": 176
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 122.35416984558105,
-      "epoch": 0.5307346326836582,
-      "grad_norm": 1.3250160837320035,
-      "kl": 0.54052734375,
-      "learning_rate": 1.0682424133646712e-05,
-      "loss": 0.0118,
-      "reward": 2.8939477801322937,
-      "reward_std": 0.1932453876361251,
-      "rewards/accuracy_reward": 0.9218750149011612,
-      "rewards/reasoning_steps_reward": 0.9947916716337204,
-      "rewards/repetition_penalty_reward": -0.012302407994866371,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 201.26562881469727,
+      "epoch": 0.265466816647919,
+      "grad_norm": 0.9139688015624915,
+      "kl": 0.6591796875,
+      "learning_rate": 1.838146491574624e-05,
+      "loss": -0.0279,
+      "reward": 2.6243019104003906,
+      "reward_std": 0.411740280687809,
+      "rewards/accuracy_reward": 0.6979166865348816,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.024569710716605186,
+      "rewards/tag_count_reward": 0.9752604216337204,
       "step": 177
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 116.97916984558105,
-      "epoch": 0.5337331334332833,
-      "grad_norm": 1.1202124937220295,
-      "kl": 0.50048828125,
-      "learning_rate": 1.0577563347894286e-05,
-      "loss": 0.0343,
-      "reward": 2.8213730454444885,
-      "reward_std": 0.12381599424406886,
-      "rewards/accuracy_reward": 0.8489583432674408,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.017168688587844372,
-      "rewards/tag_count_reward": 0.9895833432674408,
+      "completion_length": 203.03125762939453,
+      "epoch": 0.26696662917135355,
+      "grad_norm": 0.7129835639250315,
+      "kl": 0.44970703125,
+      "learning_rate": 1.83527426498578e-05,
+      "loss": 0.0649,
+      "reward": 2.529602825641632,
+      "reward_std": 0.2042294256389141,
+      "rewards/accuracy_reward": 0.5677083395421505,
+      "rewards/reasoning_steps_reward": 0.9965277910232544,
+      "rewards/repetition_penalty_reward": -0.030727183911949396,
+      "rewards/tag_count_reward": 0.9960937649011612,
       "step": 178
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 115.73958587646484,
-      "epoch": 0.5367316341829086,
-      "grad_norm": 0.9845571322277763,
-      "kl": 0.4287109375,
-      "learning_rate": 1.0472638801453287e-05,
-      "loss": 0.0274,
-      "reward": 2.886799395084381,
-      "reward_std": 0.19121930375695229,
-      "rewards/accuracy_reward": 0.9010417014360428,
-      "rewards/reasoning_steps_reward": 1.0,
-      "rewards/repetition_penalty_reward": -0.011638134252279997,
-      "rewards/tag_count_reward": 0.9973958432674408,
+      "completion_length": 228.640625,
+      "epoch": 0.26846644169478817,
+      "grad_norm": 1.3351814365191201,
+      "kl": 1.0810546875,
+      "learning_rate": 1.832379062408394e-05,
+      "loss": -0.035,
+      "reward": 2.6204107999801636,
+      "reward_std": 0.32189762964844704,
+      "rewards/accuracy_reward": 0.7187500149011612,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.031064926646649837,
+      "rewards/tag_count_reward": 0.9674479365348816,
       "step": 179
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 106.58333396911621,
-      "epoch": 0.5397301349325337,
-      "grad_norm": 1.3925075814052417,
-      "kl": 0.671875,
-      "learning_rate": 1.0367662077574898e-05,
-      "loss": 0.006,
-      "reward": 2.823054552078247,
-      "reward_std": 0.2618987523019314,
-      "rewards/accuracy_reward": 0.8593750149011612,
-      "rewards/reasoning_steps_reward": 0.998263880610466,
-      "rewards/repetition_penalty_reward": -0.009844740270636976,
-      "rewards/tag_count_reward": 0.9752604514360428,
+      "completion_length": 221.95833587646484,
+      "epoch": 0.26996625421822273,
+      "grad_norm": 0.6416068587211241,
+      "kl": 0.63525390625,
+      "learning_rate": 1.8294609634811482e-05,
+      "loss": -0.0238,
+      "reward": 2.7204131484031677,
+      "reward_std": 0.26637247391045094,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.02785079344175756,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 180
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 105.11979484558105,
-      "epoch": 0.5427286356821589,
-      "grad_norm": 2.280358299836913,
-      "kl": 0.8330078125,
-      "learning_rate": 1.0262644765270472e-05,
-      "loss": 0.0481,
-      "reward": 2.8652560710906982,
-      "reward_std": 0.35385340452194214,
-      "rewards/accuracy_reward": 0.9062500298023224,
-      "rewards/reasoning_steps_reward": 0.9930555522441864,
-      "rewards/repetition_penalty_reward": -0.008007904718397185,
-      "rewards/tag_count_reward": 0.973958358168602,
+      "completion_length": 228.32812881469727,
+      "epoch": 0.2714660667416573,
+      "grad_norm": 13.370041013613847,
+      "kl": 2.455078125,
+      "learning_rate": 1.8265200484725364e-05,
+      "loss": -0.0158,
+      "reward": 2.6316198110580444,
+      "reward_std": 0.4319092929363251,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.9687500298023224,
+      "rewards/repetition_penalty_reward": -0.031140658538788557,
+      "rewards/tag_count_reward": 0.9492187649011612,
       "step": 181
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 98.85416793823242,
-      "epoch": 0.545727136431784,
-      "grad_norm": 8.675677318683208,
-      "kl": 1.4765625,
-      "learning_rate": 1.0157598458032165e-05,
-      "loss": 0.056,
-      "reward": 2.7417566776275635,
-      "reward_std": 0.3702938035130501,
-      "rewards/accuracy_reward": 0.7812500149011612,
-      "rewards/reasoning_steps_reward": 0.9947916865348816,
-      "rewards/repetition_penalty_reward": -0.004337083431892097,
-      "rewards/tag_count_reward": 0.9700520932674408,
+      "completion_length": 219.8541717529297,
+      "epoch": 0.27296587926509186,
+      "grad_norm": 6.890078416473243,
+      "kl": 4.005859375,
+      "learning_rate": 1.823556398278657e-05,
+      "loss": -0.0641,
+      "reward": 2.506425619125366,
+      "reward_std": 0.6498684138059616,
+      "rewards/accuracy_reward": 0.6614583432674408,
+      "rewards/reasoning_steps_reward": 0.9340277761220932,
+      "rewards/repetition_penalty_reward": -0.026560600381344557,
+      "rewards/tag_count_reward": 0.9375000149011612,
       "step": 182
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 94.06771087646484,
-      "epoch": 0.5487256371814093,
-      "grad_norm": 9.74056119805903,
-      "kl": 1.44775390625,
-      "learning_rate": 1.0052534752553063e-05,
-      "loss": 0.0459,
-      "reward": 2.8742672204971313,
-      "reward_std": 0.2908545406535268,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9774305671453476,
-      "rewards/repetition_penalty_reward": -0.0042049614421557635,
-      "rewards/tag_count_reward": 0.984375,
+      "completion_length": 224.7083396911621,
+      "epoch": 0.2744656917885264,
+      "grad_norm": 0.8460594296171978,
+      "kl": 0.955078125,
+      "learning_rate": 1.820570094420989e-05,
+      "loss": -0.0508,
+      "reward": 2.5241443514823914,
+      "reward_std": 0.5859170779585838,
+      "rewards/accuracy_reward": 0.6458333507180214,
+      "rewards/reasoning_steps_reward": 0.9565972536802292,
+      "rewards/repetition_penalty_reward": -0.026202970184385777,
+      "rewards/tag_count_reward": 0.9479167014360428,
       "step": 183
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 94.85937690734863,
-      "epoch": 0.5517241379310345,
-      "grad_norm": 6.447606788332282,
-      "kl": 1.68603515625,
-      "learning_rate": 9.947465247446942e-06,
-      "loss": 0.0884,
-      "reward": 2.8351686000823975,
-      "reward_std": 0.41117405891418457,
-      "rewards/accuracy_reward": 0.9010416865348816,
-      "rewards/reasoning_steps_reward": 0.9791666865348816,
-      "rewards/repetition_penalty_reward": -0.007279429468326271,
-      "rewards/tag_count_reward": 0.9622395932674408,
+      "completion_length": 227.98438262939453,
+      "epoch": 0.275965504311961,
+      "grad_norm": 2.8022393456487182,
+      "kl": 3.2421875,
+      "learning_rate": 1.817561219044148e-05,
+      "loss": -0.1265,
+      "reward": 2.574957013130188,
+      "reward_std": 0.7494899779558182,
+      "rewards/accuracy_reward": 0.770833358168602,
+      "rewards/reasoning_steps_reward": 0.91493059694767,
+      "rewards/repetition_penalty_reward": -0.030077794566750526,
+      "rewards/tag_count_reward": 0.9192708432674408,
       "step": 184
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 92.75000381469727,
-      "epoch": 0.5547226386806596,
-      "grad_norm": 38.497156677104286,
-      "kl": 2.2099609375,
-      "learning_rate": 9.842401541967838e-06,
-      "loss": 0.1526,
-      "reward": 2.821943938732147,
-      "reward_std": 0.33023548871278763,
-      "rewards/accuracy_reward": 0.8593750298023224,
-      "rewards/reasoning_steps_reward": 0.989583358168602,
-      "rewards/repetition_penalty_reward": -0.0035769873938988894,
-      "rewards/tag_count_reward": 0.9765625,
+      "completion_length": 212.43750381469727,
+      "epoch": 0.27746531683539555,
+      "grad_norm": 1.6154380778220612,
+      "kl": 1.8671875,
+      "learning_rate": 1.814529854913626e-05,
+      "loss": -0.1303,
+      "reward": 2.467576324939728,
+      "reward_std": 0.7319738119840622,
+      "rewards/accuracy_reward": 0.661458358168602,
+      "rewards/reasoning_steps_reward": 0.9184027910232544,
+      "rewards/repetition_penalty_reward": -0.027649471536278725,
+      "rewards/tag_count_reward": 0.9153645932674408,
       "step": 185
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 88.43229484558105,
-      "epoch": 0.5577211394302849,
-      "grad_norm": 1.83258753969118,
-      "kl": 0.9951171875,
-      "learning_rate": 9.737355234729531e-06,
-      "loss": 0.017,
-      "reward": 2.743259012699127,
-      "reward_std": 0.3549557775259018,
-      "rewards/accuracy_reward": 0.8072916865348816,
-      "rewards/reasoning_steps_reward": 0.9722222536802292,
-      "rewards/repetition_penalty_reward": -0.005005000915843993,
-      "rewards/tag_count_reward": 0.96875,
+      "completion_length": 225.0989646911621,
+      "epoch": 0.27896512935883017,
+      "grad_norm": 11.447938130338011,
+      "kl": 5.7109375,
+      "learning_rate": 1.8114760854135168e-05,
+      "loss": -0.0203,
+      "reward": 2.524846374988556,
+      "reward_std": 0.6509635746479034,
+      "rewards/accuracy_reward": 0.7031250149011612,
+      "rewards/reasoning_steps_reward": 0.9392361640930176,
+      "rewards/repetition_penalty_reward": -0.025066897738724947,
+      "rewards/tag_count_reward": 0.9075520932674408,
       "step": 186
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 90.44791984558105,
-      "epoch": 0.56071964017991,
-      "grad_norm": 1.8795466176250475,
-      "kl": 1.1171875,
-      "learning_rate": 9.632337922425106e-06,
-      "loss": -0.0016,
-      "reward": 2.803585946559906,
-      "reward_std": 0.39651068300008774,
-      "rewards/accuracy_reward": 0.8750000149011612,
-      "rewards/reasoning_steps_reward": 0.973958358168602,
-      "rewards/repetition_penalty_reward": -0.008914108970202506,
-      "rewards/tag_count_reward": 0.9635416716337204,
+      "completion_length": 190.2239646911621,
+      "epoch": 0.28046494188226473,
+      "grad_norm": 1.2909186023075674,
+      "kl": 2.044921875,
+      "learning_rate": 1.808399994544222e-05,
+      "loss": -0.163,
+      "reward": 2.393586277961731,
+      "reward_std": 0.7894448935985565,
+      "rewards/accuracy_reward": 0.6770833507180214,
+      "rewards/reasoning_steps_reward": 0.8593750298023224,
+      "rewards/repetition_penalty_reward": -0.01917426590807736,
+      "rewards/tag_count_reward": 0.8763021230697632,
       "step": 187
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 91.19791984558105,
-      "epoch": 0.5637181409295352,
-      "grad_norm": 43.26072438039279,
-      "kl": 3.529296875,
-      "learning_rate": 9.527361198546715e-06,
-      "loss": 0.1745,
-      "reward": 2.712351143360138,
-      "reward_std": 0.5067479014396667,
-      "rewards/accuracy_reward": 0.8385416716337204,
-      "rewards/reasoning_steps_reward": 0.947916716337204,
-      "rewards/repetition_penalty_reward": -0.005096889741253108,
-      "rewards/tag_count_reward": 0.9309895932674408,
+      "completion_length": 216.42708587646484,
+      "epoch": 0.2819647544056993,
+      "grad_norm": 1.5283509521120155,
+      "kl": 1.4580078125,
+      "learning_rate": 1.805301666920138e-05,
+      "loss": -0.0956,
+      "reward": 2.425880491733551,
+      "reward_std": 0.6612659990787506,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9236111491918564,
+      "rewards/repetition_penalty_reward": -0.027678666170686483,
+      "rewards/tag_count_reward": 0.9049479514360428,
       "step": 188
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 91.64062690734863,
-      "epoch": 0.5667166416791605,
-      "grad_norm": 11.228523577104813,
-      "kl": 1.244140625,
-      "learning_rate": 9.422436652105718e-06,
-      "loss": 0.1033,
-      "reward": 2.70283704996109,
-      "reward_std": 0.4315161928534508,
-      "rewards/accuracy_reward": 0.786458358168602,
-      "rewards/reasoning_steps_reward": 0.9652777910232544,
-      "rewards/repetition_penalty_reward": -0.00462843308923766,
-      "rewards/tag_count_reward": 0.9557291716337204,
+      "completion_length": 211.4895896911621,
+      "epoch": 0.28346456692913385,
+      "grad_norm": 0.7866126552096253,
+      "kl": 0.4970703125,
+      "learning_rate": 1.802181187767332e-05,
+      "loss": -0.05,
+      "reward": 2.5987101793289185,
+      "reward_std": 0.4513225108385086,
+      "rewards/accuracy_reward": 0.697916679084301,
+      "rewards/reasoning_steps_reward": 0.973958358168602,
+      "rewards/repetition_penalty_reward": -0.03410231741145253,
+      "rewards/tag_count_reward": 0.9609375149011612,
       "step": 189
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 90.95833778381348,
-      "epoch": 0.5697151424287856,
-      "grad_norm": 6.30477846185162,
-      "kl": 1.2138671875,
-      "learning_rate": 9.317575866353293e-06,
-      "loss": 0.1868,
-      "reward": 2.829277455806732,
-      "reward_std": 0.42368828505277634,
-      "rewards/accuracy_reward": 0.9166666865348816,
-      "rewards/reasoning_steps_reward": 0.9791667014360428,
-      "rewards/repetition_penalty_reward": -0.004055907833389938,
-      "rewards/tag_count_reward": 0.9375000149011612,
+      "completion_length": 210.44271087646484,
+      "epoch": 0.2849643794525684,
+      "grad_norm": 1.8848248356549524,
+      "kl": 0.82763671875,
+      "learning_rate": 1.7990386429211945e-05,
+      "loss": 0.0075,
+      "reward": 2.6982374787330627,
+      "reward_std": 0.35126618295907974,
+      "rewards/accuracy_reward": 0.7760416865348816,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.02789103239774704,
+      "rewards/tag_count_reward": 0.9726562798023224,
       "step": 190
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 93.19791793823242,
-      "epoch": 0.5727136431784108,
-      "grad_norm": 7.430329871632132,
-      "kl": 2.046875,
-      "learning_rate": 9.212790417501688e-06,
-      "loss": 0.1354,
-      "reward": 2.5388087034225464,
-      "reward_std": 0.6883069798350334,
-      "rewards/accuracy_reward": 0.7031250149011612,
-      "rewards/reasoning_steps_reward": 0.9461805820465088,
-      "rewards/repetition_penalty_reward": -0.0050282846204936504,
-      "rewards/tag_count_reward": 0.8945312649011612,
+      "completion_length": 207.36980056762695,
+      "epoch": 0.286464191976003,
+      "grad_norm": 0.709217032508712,
+      "kl": 0.396484375,
+      "learning_rate": 1.7958741188240808e-05,
+      "loss": 0.003,
+      "reward": 2.6602558493614197,
+      "reward_std": 0.3209294006228447,
+      "rewards/accuracy_reward": 0.7135416865348816,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.03635886125266552,
+      "rewards/tag_count_reward": 0.9882812649011612,
       "step": 191
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 98.67187690734863,
-      "epoch": 0.5757121439280359,
-      "grad_norm": 5.377382713696718,
-      "kl": 1.591796875,
-      "learning_rate": 9.108091873446264e-06,
-      "loss": 0.2739,
-      "reward": 2.752804219722748,
-      "reward_std": 0.5407753884792328,
-      "rewards/accuracy_reward": 0.8854166716337204,
-      "rewards/reasoning_steps_reward": 0.9340278059244156,
-      "rewards/repetition_penalty_reward": -0.004140403761994094,
-      "rewards/tag_count_reward": 0.9375000149011612,
+      "completion_length": 196.33855056762695,
+      "epoch": 0.2879640044994376,
+      "grad_norm": 0.690421758006153,
+      "kl": 0.44775390625,
+      "learning_rate": 1.7926877025229313e-05,
+      "loss": 0.0006,
+      "reward": 2.72603303194046,
+      "reward_std": 0.37794774770736694,
+      "rewards/accuracy_reward": 0.7968750149011612,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.03481765231117606,
+      "rewards/tag_count_reward": 0.9778645932674408,
       "step": 192
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 103.41666984558105,
-      "epoch": 0.5787106446776612,
-      "grad_norm": 4.0000714621638105,
-      "kl": 1.837890625,
-      "learning_rate": 9.003491792488438e-06,
-      "loss": 0.2431,
-      "reward": 2.7179980874061584,
-      "reward_std": 0.5703071355819702,
-      "rewards/accuracy_reward": 0.8437500298023224,
-      "rewards/reasoning_steps_reward": 0.9461805671453476,
-      "rewards/repetition_penalty_reward": -0.004224258096655831,
-      "rewards/tag_count_reward": 0.9322916865348816,
+      "completion_length": 209.93750381469727,
+      "epoch": 0.28946381702287216,
+      "grad_norm": 0.6642189099853063,
+      "kl": 0.52783203125,
+      "learning_rate": 1.789479481666878e-05,
+      "loss": 0.0102,
+      "reward": 2.6431049704551697,
+      "reward_std": 0.30012810230255127,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.03528041858226061,
+      "rewards/tag_count_reward": 0.9804687649011612,
       "step": 193
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 109.62500190734863,
-      "epoch": 0.5817091454272864,
-      "grad_norm": 6.784297297705125,
-      "kl": 1.486328125,
-      "learning_rate": 8.899001722059687e-06,
-      "loss": 0.4446,
-      "reward": 2.69238018989563,
-      "reward_std": 0.5326372683048248,
-      "rewards/accuracy_reward": 0.7916667014360428,
-      "rewards/reasoning_steps_reward": 0.9548610895872116,
-      "rewards/repetition_penalty_reward": -0.005970706697553396,
-      "rewards/tag_count_reward": 0.9518229216337204,
+      "completion_length": 194.14063262939453,
+      "epoch": 0.2909636295463067,
+      "grad_norm": 3.7403729265839174,
+      "kl": 1.55224609375,
+      "learning_rate": 1.786249544504834e-05,
+      "loss": -0.0579,
+      "reward": 2.4264142513275146,
+      "reward_std": 0.48074235022068024,
+      "rewards/accuracy_reward": 0.5104166865348816,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.03973169345408678,
+      "rewards/tag_count_reward": 0.9713542014360428,
       "step": 194
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 115.79167175292969,
-      "epoch": 0.5847076461769115,
-      "grad_norm": 15.68094738303537,
-      "kl": 1.1884765625,
-      "learning_rate": 8.79463319744677e-06,
-      "loss": 0.553,
-      "reward": 2.6782643795013428,
-      "reward_std": 0.4203122928738594,
-      "rewards/accuracy_reward": 0.7656250149011612,
-      "rewards/reasoning_steps_reward": 0.9531250447034836,
-      "rewards/repetition_penalty_reward": -0.004027447925182059,
-      "rewards/tag_count_reward": 0.9635416716337204,
+      "completion_length": 199.9322967529297,
+      "epoch": 0.2924634420697413,
+      "grad_norm": 5.690002700824813,
+      "kl": 16.08984375,
+      "learning_rate": 1.7829979798830646e-05,
+      "loss": -0.0034,
+      "reward": 2.7136351466178894,
+      "reward_std": 0.36977453902363777,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.03766712639480829,
+      "rewards/tag_count_reward": 0.970052108168602,
       "step": 195
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 107.8281307220459,
-      "epoch": 0.5877061469265368,
-      "grad_norm": 9.395887357942577,
-      "kl": 1.21875,
-      "learning_rate": 8.690397740518279e-06,
-      "loss": 0.4238,
-      "reward": 2.8701762557029724,
-      "reward_std": 0.38676780834794044,
-      "rewards/accuracy_reward": 0.9375000298023224,
-      "rewards/reasoning_steps_reward": 0.9687500149011612,
-      "rewards/repetition_penalty_reward": -0.0035217873519286513,
-      "rewards/tag_count_reward": 0.9674479365348816,
+      "completion_length": 185.40104293823242,
+      "epoch": 0.29396325459317585,
+      "grad_norm": 1.0626617195260077,
+      "kl": 2.44140625,
+      "learning_rate": 1.779724877242745e-05,
+      "loss": -0.0171,
+      "reward": 2.54072505235672,
+      "reward_std": 0.3664104826748371,
+      "rewards/accuracy_reward": 0.6406250149011612,
+      "rewards/reasoning_steps_reward": 0.9809027910232544,
+      "rewards/repetition_penalty_reward": -0.046948717907071114,
+      "rewards/tag_count_reward": 0.9661458432674408,
       "step": 196
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 106.45312690734863,
-      "epoch": 0.5907046476761619,
-      "grad_norm": 4.57053771862413,
-      "kl": 0.8564453125,
-      "learning_rate": 8.586306858452653e-06,
-      "loss": 0.3129,
-      "reward": 2.78133487701416,
-      "reward_std": 0.38999389111995697,
-      "rewards/accuracy_reward": 0.833333358168602,
-      "rewards/reasoning_steps_reward": 0.9809028059244156,
-      "rewards/repetition_penalty_reward": -0.006859738496132195,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 187.1822967529297,
+      "epoch": 0.2954630671166104,
+      "grad_norm": 2.4457099387311376,
+      "kl": 4.498046875,
+      "learning_rate": 1.776430326617498e-05,
+      "loss": 0.0135,
+      "reward": 2.7412848472595215,
+      "reward_std": 0.20034414064139128,
+      "rewards/accuracy_reward": 0.796875,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.039965324103832245,
+      "rewards/tag_count_reward": 0.989583358168602,
       "step": 197
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 114.46354675292969,
-      "epoch": 0.5937031484257871,
-      "grad_norm": 11.162254121408726,
-      "kl": 1.4287109375,
-      "learning_rate": 8.48237204246785e-06,
-      "loss": 0.5309,
-      "reward": 2.7748392820358276,
-      "reward_std": 0.4289344698190689,
-      "rewards/accuracy_reward": 0.8697917014360428,
-      "rewards/reasoning_steps_reward": 0.9496528059244156,
-      "rewards/repetition_penalty_reward": -0.00554278859635815,
-      "rewards/tag_count_reward": 0.9609375,
+      "completion_length": 199.6354217529297,
+      "epoch": 0.296962879640045,
+      "grad_norm": 0.7391409818658768,
+      "kl": 0.830078125,
+      "learning_rate": 1.77311441863092e-05,
+      "loss": -0.053,
+      "reward": 2.715912103652954,
+      "reward_std": 0.4153308942914009,
+      "rewards/accuracy_reward": 0.8125000149011612,
+      "rewards/reasoning_steps_reward": 0.982638880610466,
+      "rewards/repetition_penalty_reward": -0.04927895776927471,
+      "rewards/tag_count_reward": 0.9700520932674408,
       "step": 198
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 111.04166793823242,
-      "epoch": 0.5967016491754122,
-      "grad_norm": 29.168148040502643,
-      "kl": 1.78515625,
-      "learning_rate": 8.378604766552756e-06,
-      "loss": 0.3664,
-      "reward": 2.7375897765159607,
-      "reward_std": 0.32485630363225937,
-      "rewards/accuracy_reward": 0.7812500149011612,
-      "rewards/reasoning_steps_reward": 0.9895833432674408,
-      "rewards/repetition_penalty_reward": -0.0072020008228719234,
-      "rewards/tag_count_reward": 0.9739583432674408,
+      "completion_length": 185.04687881469727,
+      "epoch": 0.2984626921634796,
+      "grad_norm": 0.7197244320446806,
+      "kl": 0.33837890625,
+      "learning_rate": 1.769777244494086e-05,
+      "loss": 0.0279,
+      "reward": 2.6263469457626343,
+      "reward_std": 0.35684073716402054,
+      "rewards/accuracy_reward": 0.6927083432674408,
+      "rewards/reasoning_steps_reward": 0.9913194626569748,
+      "rewards/repetition_penalty_reward": -0.042055959813296795,
+      "rewards/tag_count_reward": 0.984375,
       "step": 199
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 125.81250381469727,
-      "epoch": 0.5997001499250375,
-      "grad_norm": 7.416359123988612,
-      "kl": 0.955078125,
-      "learning_rate": 8.275016486200498e-06,
-      "loss": 0.623,
-      "reward": 2.769349992275238,
-      "reward_std": 0.4332389682531357,
-      "rewards/accuracy_reward": 0.8385416865348816,
-      "rewards/reasoning_steps_reward": 0.9670139104127884,
-      "rewards/repetition_penalty_reward": -0.004955811775289476,
-      "rewards/tag_count_reward": 0.96875,
+      "completion_length": 187.06250762939453,
+      "epoch": 0.29996250468691416,
+      "grad_norm": 0.7272305025229391,
+      "kl": 0.3505859375,
+      "learning_rate": 1.7664188960030422e-05,
+      "loss": 0.0127,
+      "reward": 2.607051908969879,
+      "reward_std": 0.35574449226260185,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555671453476,
+      "rewards/repetition_penalty_reward": -0.03313909750431776,
+      "rewards/tag_count_reward": 0.9804687649011612,
       "step": 200
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 1024.0,
-      "epoch": 0.6026986506746627,
-      "grad_norm": 27.882830639662522,
-      "kl": 2.0904541015625,
-      "learning_rate": 8.17161863714381e-06,
-      "loss": 0.0835,
-      "reward": 0.0013020833721384406,
-      "reward_std": 0.0052083334885537624,
-      "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.0,
-      "rewards/repetition_penalty_reward": 0.0,
-      "rewards/tag_count_reward": 0.0013020833721384406,
+      "completion_length": 172.6354217529297,
+      "epoch": 0.3014623172103487,
+      "grad_norm": 1.8628860952936386,
+      "kl": 0.56298828125,
+      "learning_rate": 1.7630394655362798e-05,
+      "loss": -0.0152,
+      "reward": 2.3712204694747925,
+      "reward_std": 0.8324102908372879,
+      "rewards/accuracy_reward": 0.6302083432674408,
+      "rewards/reasoning_steps_reward": 0.9062500596046448,
+      "rewards/repetition_penalty_reward": -0.035029674880206585,
+      "rewards/tag_count_reward": 0.8697916865348816,
       "step": 201
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 1014.3333435058594,
-      "epoch": 0.6056971514242878,
-      "grad_norm": 0.6042533284398771,
-      "kl": 0.113037109375,
-      "learning_rate": 8.06842263409257e-06,
-      "loss": 0.0419,
-      "reward": 0.029826793004758656,
-      "reward_std": 0.11930717201903462,
-      "rewards/accuracy_reward": 0.0052083334885537624,
-      "rewards/reasoning_steps_reward": 0.010416666977107525,
-      "rewards/repetition_penalty_reward": -0.00012112403055652976,
-      "rewards/tag_count_reward": 0.014322917093522847,
+      "completion_length": 163.6354217529297,
+      "epoch": 0.3029621297337833,
+      "grad_norm": 1.0953361912511788,
+      "kl": 0.48681640625,
+      "learning_rate": 1.7596390460521946e-05,
+      "loss": -0.0349,
+      "reward": 2.3163662552833557,
+      "reward_std": 0.572496585547924,
+      "rewards/accuracy_reward": 0.463541679084301,
+      "rewards/reasoning_steps_reward": 0.9496527910232544,
+      "rewards/repetition_penalty_reward": -0.025213704211637378,
+      "rewards/tag_count_reward": 0.9283854365348816,
       "step": 202
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 1019.1822967529297,
-      "epoch": 0.6086956521739131,
-      "grad_norm": 1.3832204376438657,
-      "kl": 0.21142578125,
-      "learning_rate": 7.965439869473664e-06,
-      "loss": 0.0271,
-      "reward": 0.018031382700428367,
-      "reward_std": 0.06882480159401894,
-      "rewards/accuracy_reward": 0.0052083334885537624,
-      "rewards/reasoning_steps_reward": 0.0052083334885537624,
-      "rewards/repetition_penalty_reward": -0.00019778480054810643,
-      "rewards/tag_count_reward": 0.007812500232830644,
+      "completion_length": 153.9427146911621,
+      "epoch": 0.30446194225721784,
+      "grad_norm": 1.9105972336877117,
+      "kl": 0.41455078125,
+      "learning_rate": 1.7562177310865296e-05,
+      "loss": 0.1019,
+      "reward": 2.7670071125030518,
+      "reward_std": 0.4459151364862919,
+      "rewards/accuracy_reward": 0.8593750298023224,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.021621438674628735,
+      "rewards/tag_count_reward": 0.9622395932674408,
       "step": 203
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 1024.0,
-      "epoch": 0.6116941529235382,
-      "grad_norm": 0.9850061937944327,
-      "kl": 0.1051025390625,
-      "learning_rate": 7.862681712173304e-06,
-      "loss": 0.0042,
-      "reward": 0.007812500232830644,
-      "reward_std": 0.027949271723628044,
-      "rewards/accuracy_reward": 0.0,
-      "rewards/reasoning_steps_reward": 0.0,
-      "rewards/repetition_penalty_reward": 0.0,
-      "rewards/tag_count_reward": 0.007812500232830644,
+      "completion_length": 147.7395896911621,
+      "epoch": 0.3059617547806524,
+      "grad_norm": 1.1424371878407156,
+      "kl": 0.4150390625,
+      "learning_rate": 1.7527756147498026e-05,
+      "loss": 0.007,
+      "reward": 2.516347885131836,
+      "reward_std": 0.38063713908195496,
+      "rewards/accuracy_reward": 0.609375,
+      "rewards/reasoning_steps_reward": 0.9687500447034836,
+      "rewards/repetition_penalty_reward": -0.024016746319830418,
+      "rewards/tag_count_reward": 0.9622395932674408,
       "step": 204
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 1009.2812652587891,
-      "epoch": 0.6146926536731634,
-      "grad_norm": 0.44253170115395524,
-      "kl": 0.1737060546875,
-      "learning_rate": 7.760159506281955e-06,
-      "loss": 0.0639,
-      "reward": 0.04947916732635349,
-      "reward_std": 0.19791666930541396,
-      "rewards/accuracy_reward": 0.015625000465661287,
-      "rewards/reasoning_steps_reward": 0.015625000465661287,
-      "rewards/repetition_penalty_reward": 0.0,
-      "rewards/tag_count_reward": 0.01822916732635349,
+      "completion_length": 142.0208396911621,
+      "epoch": 0.30746156730408697,
+      "grad_norm": 0.8745105132943645,
+      "kl": 0.33447265625,
+      "learning_rate": 1.7493127917247168e-05,
+      "loss": 0.017,
+      "reward": 2.7387551069259644,
+      "reward_std": 0.25662550423294306,
+      "rewards/accuracy_reward": 0.8020833432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.022095664869993925,
+      "rewards/tag_count_reward": 0.9726562649011612,
       "step": 205
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 995.3854370117188,
-      "epoch": 0.6176911544227887,
-      "grad_norm": 3.8020369247401415,
-      "kl": 0.369384765625,
-      "learning_rate": 7.65788456984198e-06,
-      "loss": 0.1046,
-      "reward": 0.10146949626505375,
-      "reward_std": 0.3067319616675377,
-      "rewards/accuracy_reward": 0.026041667442768812,
-      "rewards/reasoning_steps_reward": 0.031250000931322575,
-      "rewards/repetition_penalty_reward": -9.300596138928086e-05,
-      "rewards/tag_count_reward": 0.044270834885537624,
+      "completion_length": 144.41667556762695,
+      "epoch": 0.3089613798275216,
+      "grad_norm": 0.841666968800414,
+      "kl": 0.3515625,
+      "learning_rate": 1.7458293572635573e-05,
+      "loss": -0.0625,
+      "reward": 2.5762619972229004,
+      "reward_std": 0.39808357134461403,
+      "rewards/accuracy_reward": 0.6406250149011612,
+      "rewards/reasoning_steps_reward": 0.9861111044883728,
+      "rewards/repetition_penalty_reward": -0.025734632275998592,
+      "rewards/tag_count_reward": 0.9752604216337204,
       "step": 206
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 980.5520935058594,
-      "epoch": 0.6206896551724138,
-      "grad_norm": 2.4622521421222046,
-      "kl": 0.35595703125,
-      "learning_rate": 7.555868193598188e-06,
-      "loss": 0.1499,
-      "reward": 0.15076165180653334,
-      "reward_std": 0.4339568931609392,
-      "rewards/accuracy_reward": 0.041666666977107525,
-      "rewards/reasoning_steps_reward": 0.046875000931322575,
-      "rewards/repetition_penalty_reward": -0.0002800179208861664,
-      "rewards/tag_count_reward": 0.06250000186264515,
+      "completion_length": 145.5989646911621,
+      "epoch": 0.31046119235095615,
+      "grad_norm": 0.8971484142091678,
+      "kl": 0.4033203125,
+      "learning_rate": 1.7423254071855696e-05,
+      "loss": 0.0317,
+      "reward": 2.6069366335868835,
+      "reward_std": 0.4525081217288971,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.020233482588082552,
+      "rewards/tag_count_reward": 0.9518229365348816,
       "step": 207
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 985.7864685058594,
-      "epoch": 0.623688155922039,
-      "grad_norm": 508.3775666051304,
-      "kl": 21.75,
-      "learning_rate": 7.4541216397513705e-06,
-      "loss": 1.0583,
-      "reward": 0.15980212949216366,
-      "reward_std": 0.5328295826911926,
-      "rewards/accuracy_reward": 0.03645833441987634,
-      "rewards/reasoning_steps_reward": 0.057291668839752674,
-      "rewards/repetition_penalty_reward": -0.0003541221594787203,
-      "rewards/tag_count_reward": 0.06640625093132257,
+      "completion_length": 136.8541717529297,
+      "epoch": 0.3119610048743907,
+      "grad_norm": 0.7957437127746465,
+      "kl": 0.4130859375,
+      "learning_rate": 1.7388010378743255e-05,
+      "loss": -0.0478,
+      "reward": 2.5852027535438538,
+      "reward_std": 0.3762405589222908,
+      "rewards/accuracy_reward": 0.6614583432674408,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.015057688346132636,
+      "rewards/tag_count_reward": 0.970052108168602,
       "step": 208
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 980.5989685058594,
-      "epoch": 0.6266866566716641,
-      "grad_norm": 16.61142685491831,
-      "kl": 1.8037109375,
-      "learning_rate": 7.352656140715006e-06,
-      "loss": 0.1989,
-      "reward": 0.16810668725520372,
-      "reward_std": 0.42627183347940445,
-      "rewards/accuracy_reward": 0.03645833441987634,
-      "rewards/reasoning_steps_reward": 0.05381944612599909,
-      "rewards/repetition_penalty_reward": -0.00029610148339997977,
-      "rewards/tag_count_reward": 0.07812500139698386,
+      "completion_length": 141.765625,
+      "epoch": 0.3134608173978253,
+      "grad_norm": 0.7389331807682286,
+      "kl": 0.45556640625,
+      "learning_rate": 1.735256346275071e-05,
+      "loss": -0.0314,
+      "reward": 2.709416389465332,
+      "reward_std": 0.4428819492459297,
+      "rewards/accuracy_reward": 0.7968750298023224,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.022788605885580182,
+      "rewards/tag_count_reward": 0.9700520932674408,
       "step": 209
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 990.5625152587891,
-      "epoch": 0.6296851574212894,
-      "grad_norm": 5.532107109224189,
-      "kl": 0.662109375,
-      "learning_rate": 7.2514828978752434e-06,
-      "loss": 0.1355,
-      "reward": 0.1534007415175438,
-      "reward_std": 0.41936828941106796,
-      "rewards/accuracy_reward": 0.031250000931322575,
-      "rewards/reasoning_steps_reward": 0.04687500232830644,
-      "rewards/repetition_penalty_reward": -0.0002450980609864928,
-      "rewards/tag_count_reward": 0.07552083488553762,
+      "completion_length": 144.4114646911621,
+      "epoch": 0.31496062992125984,
+      "grad_norm": 0.8326796660055926,
+      "kl": 0.36376953125,
+      "learning_rate": 1.7316914298920592e-05,
+      "loss": -0.0442,
+      "reward": 2.6346763372421265,
+      "reward_std": 0.4088163301348686,
+      "rewards/accuracy_reward": 0.6979167014360428,
+      "rewards/reasoning_steps_reward": 0.9722222536802292,
+      "rewards/repetition_penalty_reward": -0.01853547664359212,
+      "rewards/tag_count_reward": 0.9830729216337204,
       "step": 210
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 942.3229370117188,
-      "epoch": 0.6326836581709145,
-      "grad_norm": 14.765094590304768,
-      "kl": 0.7763671875,
-      "learning_rate": 7.150613080354315e-06,
-      "loss": 0.262,
-      "reward": 0.32499945536255836,
-      "reward_std": 0.7303483635187149,
-      "rewards/accuracy_reward": 0.06770833535119891,
-      "rewards/reasoning_steps_reward": 0.118055559694767,
-      "rewards/repetition_penalty_reward": -0.0013894452131353319,
-      "rewards/tag_count_reward": 0.1406250037252903,
+      "completion_length": 153.4947967529297,
+      "epoch": 0.3164604424446944,
+      "grad_norm": 0.8162698667548337,
+      "kl": 0.4296875,
+      "learning_rate": 1.7281063867858687e-05,
+      "loss": -0.0033,
+      "reward": 2.47624808549881,
+      "reward_std": 0.4159582331776619,
+      "rewards/accuracy_reward": 0.5520833507180214,
+      "rewards/reasoning_steps_reward": 0.9809028059244156,
+      "rewards/repetition_penalty_reward": -0.025488153100013733,
+      "rewards/tag_count_reward": 0.9687500149011612,
       "step": 211
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 893.5156402587891,
-      "epoch": 0.6356821589205397,
-      "grad_norm": 13.402868555399907,
-      "kl": 2.154296875,
-      "learning_rate": 7.050057823777503e-06,
-      "loss": 0.4338,
-      "reward": 0.4721004366874695,
-      "reward_std": 0.9947518408298492,
-      "rewards/accuracy_reward": 0.13020833767950535,
-      "rewards/reasoning_steps_reward": 0.1545138955116272,
-      "rewards/repetition_penalty_reward": -0.0014238738513085991,
-      "rewards/tag_count_reward": 0.1888020895421505,
+      "completion_length": 173.90625381469727,
+      "epoch": 0.31796025496812896,
+      "grad_norm": 0.9992737614199144,
+      "kl": 0.435546875,
+      "learning_rate": 1.7245013155707076e-05,
+      "loss": 0.0243,
+      "reward": 2.806881844997406,
+      "reward_std": 0.2670608460903168,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9895833283662796,
+      "rewards/repetition_penalty_reward": -0.0316599381621927,
+      "rewards/tag_count_reward": 0.9947916716337204,
       "step": 212
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 882.8437652587891,
-      "epoch": 0.638680659670165,
-      "grad_norm": 533.9503112402125,
-      "kl": 46.984375,
-      "learning_rate": 6.9498282290438235e-06,
-      "loss": 2.4742,
-      "reward": 0.5224730595946312,
-      "reward_std": 0.9336346387863159,
-      "rewards/accuracy_reward": 0.1354166716337204,
-      "rewards/reasoning_steps_reward": 0.1666666716337204,
-      "rewards/repetition_penalty_reward": -0.0009644359670346603,
-      "rewards/tag_count_reward": 0.221354179084301,
+      "completion_length": 187.99480056762695,
+      "epoch": 0.3194600674915636,
+      "grad_norm": 15.20697089186289,
+      "kl": 1.06494140625,
+      "learning_rate": 1.7208763154116973e-05,
+      "loss": 0.1191,
+      "reward": 2.6377468705177307,
+      "reward_std": 0.27828123420476913,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.028485802467912436,
+      "rewards/tag_count_reward": 0.9856770932674408,
       "step": 213
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 821.7448120117188,
-      "epoch": 0.6416791604197901,
-      "grad_norm": 149.18286801865162,
-      "kl": 16.4375,
-      "learning_rate": 6.849935361100522e-06,
-      "loss": 1.2104,
-      "reward": 0.7398104518651962,
-      "reward_std": 1.1720031201839447,
-      "rewards/accuracy_reward": 0.2083333432674408,
-      "rewards/reasoning_steps_reward": 0.2586805671453476,
-      "rewards/repetition_penalty_reward": -0.0019430473039392382,
-      "rewards/tag_count_reward": 0.2747395932674408,
+      "completion_length": 301.7708435058594,
+      "epoch": 0.32095988001499814,
+      "grad_norm": 20.978080391787007,
+      "kl": 1.6455078125,
+      "learning_rate": 1.7172314860221494e-05,
+      "loss": 0.2695,
+      "reward": 2.531617820262909,
+      "reward_std": 0.30965887755155563,
+      "rewards/accuracy_reward": 0.6093750149011612,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.037392658181488514,
+      "rewards/tag_count_reward": 0.9752604216337204,
       "step": 214
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 840.9218902587891,
-      "epoch": 0.6446776611694153,
-      "grad_norm": 30.355372142343207,
-      "kl": 3.22265625,
-      "learning_rate": 6.750390247721549e-06,
-      "loss": 0.562,
-      "reward": 0.6818769425153732,
-      "reward_std": 1.1269759833812714,
-      "rewards/accuracy_reward": 0.17708333767950535,
-      "rewards/reasoning_steps_reward": 0.2413194589316845,
-      "rewards/repetition_penalty_reward": -0.0021508438512682915,
-      "rewards/tag_count_reward": 0.2656250074505806,
+      "completion_length": 445.5572967529297,
+      "epoch": 0.3224596925384327,
+      "grad_norm": 1.4225466343917932,
+      "kl": 1.603515625,
+      "learning_rate": 1.713566927660818e-05,
+      "loss": 0.0372,
+      "reward": 2.569119691848755,
+      "reward_std": 0.31232137233018875,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.03374499548226595,
+      "rewards/tag_count_reward": 0.9882812649011612,
       "step": 215
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 803.9166870117188,
-      "epoch": 0.6476761619190404,
-      "grad_norm": 23.89552805694713,
-      "kl": 1.5625,
-      "learning_rate": 6.651203878290139e-06,
-      "loss": 0.5275,
-      "reward": 0.8299884647130966,
-      "reward_std": 1.1851888000965118,
-      "rewards/accuracy_reward": 0.2031250037252903,
-      "rewards/reasoning_steps_reward": 0.2934028059244156,
-      "rewards/repetition_penalty_reward": -0.0011747851967811584,
-      "rewards/tag_count_reward": 0.334635429084301,
+      "completion_length": 265.4427146911621,
+      "epoch": 0.32395950506186727,
+      "grad_norm": 5.079986124787856,
+      "kl": 1.427734375,
+      "learning_rate": 1.7098827411291474e-05,
+      "loss": 0.1262,
+      "reward": 2.6483259201049805,
+      "reward_std": 0.43077797442674637,
+      "rewards/accuracy_reward": 0.723958358168602,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.03570183180272579,
+      "rewards/tag_count_reward": 0.973958358168602,
       "step": 216
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 718.4583435058594,
-      "epoch": 0.6506746626686657,
-      "grad_norm": 30.60183508089706,
-      "kl": 2.451171875,
-      "learning_rate": 6.552387202585629e-06,
-      "loss": 0.7066,
-      "reward": 1.0522598177194595,
-      "reward_std": 1.2560299634933472,
-      "rewards/accuracy_reward": 0.2708333432674408,
-      "rewards/reasoning_steps_reward": 0.3732639104127884,
-      "rewards/repetition_penalty_reward": -0.0032958039082586765,
-      "rewards/tag_count_reward": 0.4114583432674408,
+      "completion_length": 216.0260467529297,
+      "epoch": 0.32545931758530183,
+      "grad_norm": 8.779280281061196,
+      "kl": 0.5869140625,
+      "learning_rate": 1.7061790277684935e-05,
+      "loss": 0.0993,
+      "reward": 2.6443240642547607,
+      "reward_std": 0.3985915258526802,
+      "rewards/accuracy_reward": 0.7447917014360428,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.03753373399376869,
+      "rewards/tag_count_reward": 0.9648437649011612,
       "step": 217
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 718.8177185058594,
-      "epoch": 0.6536731634182908,
-      "grad_norm": 64.25492952390515,
-      "kl": 7.8125,
-      "learning_rate": 6.453951129574644e-06,
-      "loss": 1.02,
-      "reward": 1.1462399363517761,
-      "reward_std": 1.311941385269165,
-      "rewards/accuracy_reward": 0.3229166753590107,
-      "rewards/reasoning_steps_reward": 0.3888889104127884,
-      "rewards/repetition_penalty_reward": -0.0017635486146900803,
-      "rewards/tag_count_reward": 0.4361979216337204,
+      "completion_length": 196.91146087646484,
+      "epoch": 0.3269591301087364,
+      "grad_norm": 8.182772315553562,
+      "kl": 0.6435546875,
+      "learning_rate": 1.7024558894573408e-05,
+      "loss": 0.135,
+      "reward": 2.516254484653473,
+      "reward_std": 0.3703230023384094,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.03148850845173001,
+      "rewards/tag_count_reward": 0.9765625149011612,
       "step": 218
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 574.0937652587891,
-      "epoch": 0.656671664167916,
-      "grad_norm": 108.73570895032785,
-      "kl": 14.359375,
-      "learning_rate": 6.355906526206788e-06,
-      "loss": 1.8085,
-      "reward": 1.5105059146881104,
-      "reward_std": 1.3671056032180786,
-      "rewards/accuracy_reward": 0.432291679084301,
-      "rewards/reasoning_steps_reward": 0.5277777910232544,
-      "rewards/repetition_penalty_reward": -0.004251136677339673,
-      "rewards/tag_count_reward": 0.5546875149011612,
+      "completion_length": 202.86458587646484,
+      "epoch": 0.32845894263217096,
+      "grad_norm": 121464435.65252294,
+      "kl": 3831808.0,
+      "learning_rate": 1.698713428608497e-05,
+      "loss": 331049.0938,
+      "reward": 2.498434364795685,
+      "reward_std": 0.5815113484859467,
+      "rewards/accuracy_reward": 0.6197916865348816,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.02500331262126565,
+      "rewards/tag_count_reward": 0.934895858168602,
       "step": 219
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 675.6041870117188,
-      "epoch": 0.6596701649175413,
-      "grad_norm": 68.16871801389341,
-      "kl": 10.140625,
-      "learning_rate": 6.2582642162149775e-06,
-      "loss": 1.2516,
-      "reward": 1.230929583311081,
-      "reward_std": 1.2748659253120422,
-      "rewards/accuracy_reward": 0.3072916716337204,
-      "rewards/reasoning_steps_reward": 0.4531250149011612,
-      "rewards/repetition_penalty_reward": -0.003445470007136464,
-      "rewards/tag_count_reward": 0.4739583507180214,
+      "completion_length": 234.6614646911621,
+      "epoch": 0.3299587551556056,
+      "grad_norm": 5166178.874158303,
+      "kl": 174346.3125,
+      "learning_rate": 1.694951748166278e-05,
+      "loss": 24312.8574,
+      "reward": 2.523725211620331,
+      "reward_std": 0.5246653333306313,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.9687500447034836,
+      "rewards/repetition_penalty_reward": -0.043983266688883305,
+      "rewards/tag_count_reward": 0.942708358168602,
       "step": 220
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 565.9896087646484,
-      "epoch": 0.6626686656671664,
-      "grad_norm": 37.36448608931068,
-      "kl": 7.1015625,
-      "learning_rate": 6.161034978920555e-06,
-      "loss": 1.2308,
-      "reward": 1.6033472418785095,
-      "reward_std": 1.2990168035030365,
-      "rewards/accuracy_reward": 0.4375000149011612,
-      "rewards/reasoning_steps_reward": 0.5798611491918564,
-      "rewards/repetition_penalty_reward": -0.003857686650007963,
-      "rewards/tag_count_reward": 0.5898437649011612,
+      "completion_length": 223.1145896911621,
+      "epoch": 0.33145856767904014,
+      "grad_norm": 431.25017200680634,
+      "kl": 21.69677734375,
+      "learning_rate": 1.6911709516036755e-05,
+      "loss": 3.13,
+      "reward": 2.538953959941864,
+      "reward_std": 0.5608213990926743,
+      "rewards/accuracy_reward": 0.6718750149011612,
+      "rewards/reasoning_steps_reward": 0.9600694924592972,
+      "rewards/repetition_penalty_reward": -0.029188551474362612,
+      "rewards/tag_count_reward": 0.9361979216337204,
       "step": 221
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 613.5521087646484,
-      "epoch": 0.6656671664167916,
-      "grad_norm": 42.67953665381653,
-      "kl": 6.75,
-      "learning_rate": 6.064229548043272e-06,
-      "loss": 1.1467,
-      "reward": 1.5173589289188385,
-      "reward_std": 1.3208832442760468,
-      "rewards/accuracy_reward": 0.416666679084301,
-      "rewards/reasoning_steps_reward": 0.5625000447034836,
-      "rewards/repetition_penalty_reward": -0.004776577930897474,
-      "rewards/tag_count_reward": 0.5429687649011612,
+      "completion_length": 276.19792556762695,
+      "epoch": 0.3329583802024747,
+      "grad_norm": 16.73765270319652,
+      "kl": 0.51953125,
+      "learning_rate": 1.6873711429195095e-05,
+      "loss": 0.3655,
+      "reward": 2.327203631401062,
+      "reward_std": 0.42651342228055,
+      "rewards/accuracy_reward": 0.4739583358168602,
+      "rewards/reasoning_steps_reward": 0.9635417014360428,
+      "rewards/repetition_penalty_reward": -0.03868194296956062,
+      "rewards/tag_count_reward": 0.9283854365348816,
       "step": 222
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 550.3437652587891,
-      "epoch": 0.6686656671664168,
-      "grad_norm": 33.134335694451195,
-      "kl": 5.140625,
-      "learning_rate": 5.9678586105163535e-06,
-      "loss": 1.071,
-      "reward": 1.621566891670227,
-      "reward_std": 1.2871178090572357,
-      "rewards/accuracy_reward": 0.4427083432674408,
-      "rewards/reasoning_steps_reward": 0.6006944626569748,
-      "rewards/repetition_penalty_reward": -0.005169248324818909,
-      "rewards/tag_count_reward": 0.583333358168602,
+      "completion_length": 263.83333587646484,
+      "epoch": 0.33445819272590926,
+      "grad_norm": 16.356024789507156,
+      "kl": 0.49072265625,
+      "learning_rate": 1.6835524266355698e-05,
+      "loss": 0.4278,
+      "reward": 2.46044385433197,
+      "reward_std": 0.501777321100235,
+      "rewards/accuracy_reward": 0.5989583432674408,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.034347846638411283,
+      "rewards/tag_count_reward": 0.9166666865348816,
       "step": 223
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 479.4739761352539,
-      "epoch": 0.671664167916042,
-      "grad_norm": 47.63738327986576,
-      "kl": 4.26953125,
-      "learning_rate": 5.8719328053066886e-06,
-      "loss": 1.1644,
-      "reward": 1.8401104509830475,
-      "reward_std": 1.2440669536590576,
-      "rewards/accuracy_reward": 0.5156250149011612,
-      "rewards/reasoning_steps_reward": 0.6840278059244156,
-      "rewards/repetition_penalty_reward": -0.005375679756980389,
-      "rewards/tag_count_reward": 0.645833358168602,
+      "completion_length": 249.49480438232422,
+      "epoch": 0.3359580052493438,
+      "grad_norm": 2132.408739618289,
+      "kl": 61.46533203125,
+      "learning_rate": 1.6797149077937395e-05,
+      "loss": 6.8572,
+      "reward": 2.410385847091675,
+      "reward_std": 0.3996973782777786,
+      "rewards/accuracy_reward": 0.5364583507180214,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.038832977414131165,
+      "rewards/tag_count_reward": 0.9283854365348816,
       "step": 224
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 488.96356201171875,
-      "epoch": 0.6746626686656672,
-      "grad_norm": 78.62849172701118,
-      "kl": 9.546875,
-      "learning_rate": 5.776462722240337e-06,
-      "loss": 1.7087,
-      "reward": 1.9351984560489655,
-      "reward_std": 1.149868130683899,
-      "rewards/accuracy_reward": 0.5104166865348816,
-      "rewards/reasoning_steps_reward": 0.763888955116272,
-      "rewards/repetition_penalty_reward": -0.004471685038879514,
-      "rewards/tag_count_reward": 0.665364608168602,
+      "completion_length": 238.79167556762695,
+      "epoch": 0.3374578177727784,
+      "grad_norm": 1.5282888356192013,
+      "kl": 0.42333984375,
+      "learning_rate": 1.6758586919531054e-05,
+      "loss": 0.4067,
+      "reward": 2.5930500626564026,
+      "reward_std": 0.40433521568775177,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9687499701976776,
+      "rewards/repetition_penalty_reward": -0.039762526750564575,
+      "rewards/tag_count_reward": 0.934895858168602,
       "step": 225
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 463.96875762939453,
-      "epoch": 0.6776611694152923,
-      "grad_norm": 50.31185538870822,
-      "kl": 9.921875,
-      "learning_rate": 5.6814589008334475e-06,
-      "loss": 1.669,
-      "reward": 1.9726946949958801,
-      "reward_std": 1.1515787243843079,
-      "rewards/accuracy_reward": 0.5312500149011612,
-      "rewards/reasoning_steps_reward": 0.7743055671453476,
-      "rewards/repetition_penalty_reward": -0.006038067163899541,
-      "rewards/tag_count_reward": 0.6731770932674408,
+      "completion_length": 230.0520896911621,
+      "epoch": 0.33895763029621295,
+      "grad_norm": 0.8528067604314957,
+      "kl": 0.41748046875,
+      "learning_rate": 1.671983885187055e-05,
+      "loss": 0.2675,
+      "reward": 2.4493818879127502,
+      "reward_std": 0.3525548577308655,
+      "rewards/accuracy_reward": 0.5468750074505806,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.028916888870298862,
+      "rewards/tag_count_reward": 0.9557291865348816,
       "step": 226
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 484.86981201171875,
-      "epoch": 0.6806596701649176,
-      "grad_norm": 25.321634983297866,
-      "kl": 7.1796875,
-      "learning_rate": 5.58693182912875e-06,
-      "loss": 1.4453,
-      "reward": 2.014142394065857,
-      "reward_std": 1.1831817626953125,
-      "rewards/accuracy_reward": 0.567708358168602,
-      "rewards/reasoning_steps_reward": 0.788194477558136,
-      "rewards/repetition_penalty_reward": -0.0032188262266572565,
-      "rewards/tag_count_reward": 0.661458358168602,
+      "completion_length": 258.1666793823242,
+      "epoch": 0.34045744281964757,
+      "grad_norm": 2.1183154515503513,
+      "kl": 0.462890625,
+      "learning_rate": 1.6680905940803596e-05,
+      "loss": 0.7181,
+      "reward": 2.582502543926239,
+      "reward_std": 0.494766004383564,
+      "rewards/accuracy_reward": 0.697916679084301,
+      "rewards/reasoning_steps_reward": 0.9739583730697632,
+      "rewards/repetition_penalty_reward": -0.02947685099206865,
+      "rewards/tag_count_reward": 0.9401041865348816,
       "step": 227
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 444.3385543823242,
-      "epoch": 0.6836581709145427,
-      "grad_norm": 31.357353768527005,
-      "kl": 2.78125,
-      "learning_rate": 5.4928919425377035e-06,
-      "loss": 1.1606,
-      "reward": 2.098558932542801,
-      "reward_std": 1.0963150560855865,
-      "rewards/accuracy_reward": 0.5885416865348816,
-      "rewards/reasoning_steps_reward": 0.8263889402151108,
-      "rewards/repetition_penalty_reward": -0.0038717142306268215,
-      "rewards/tag_count_reward": 0.6875000149011612,
+      "completion_length": 244.5989646911621,
+      "epoch": 0.34195725534308213,
+      "grad_norm": 4.12408868442433,
+      "kl": 0.43505859375,
+      "learning_rate": 1.66417892572624e-05,
+      "loss": 0.3667,
+      "reward": 2.5001248717308044,
+      "reward_std": 0.44768981635570526,
+      "rewards/accuracy_reward": 0.5989583432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.04154195077717304,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 228
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 514.4323043823242,
-      "epoch": 0.6866566716641679,
-      "grad_norm": 30.871800901209355,
-      "kl": 2.44921875,
-      "learning_rate": 5.399349622688479e-06,
-      "loss": 0.9849,
-      "reward": 1.9208540618419647,
-      "reward_std": 1.0992612540721893,
-      "rewards/accuracy_reward": 0.4739583358168602,
-      "rewards/reasoning_steps_reward": 0.8159722685813904,
-      "rewards/repetition_penalty_reward": -0.005795358796603978,
-      "rewards/tag_count_reward": 0.6367187649011612,
+      "completion_length": 252.33855056762695,
+      "epoch": 0.3434570678665167,
+      "grad_norm": 29.415667963205827,
+      "kl": 0.52783203125,
+      "learning_rate": 1.6602489877234235e-05,
+      "loss": 0.4238,
+      "reward": 2.625883162021637,
+      "reward_std": 0.4509083032608032,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9618055671453476,
+      "rewards/repetition_penalty_reward": -0.03514129016548395,
+      "rewards/tag_count_reward": 0.9492187649011612,
       "step": 229
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 407.44793701171875,
-      "epoch": 0.6896551724137931,
-      "grad_norm": 29.451600398259565,
-      "kl": 2.1015625,
-      "learning_rate": 5.306315196279864e-06,
-      "loss": 1.1623,
-      "reward": 2.2249255180358887,
-      "reward_std": 1.0277338027954102,
-      "rewards/accuracy_reward": 0.630208358168602,
-      "rewards/reasoning_steps_reward": 0.8836805820465088,
-      "rewards/repetition_penalty_reward": -0.0051092767680529505,
-      "rewards/tag_count_reward": 0.7161458432674408,
+      "completion_length": 259.85417556762695,
+      "epoch": 0.34495688038995126,
+      "grad_norm": 18.06563359145896,
+      "kl": 0.6884765625,
+      "learning_rate": 1.656300888173181e-05,
+      "loss": 0.6687,
+      "reward": 2.646353542804718,
+      "reward_std": 0.518807977437973,
+      "rewards/accuracy_reward": 0.770833358168602,
+      "rewards/reasoning_steps_reward": 0.9652778506278992,
+      "rewards/repetition_penalty_reward": -0.028559860307723284,
+      "rewards/tag_count_reward": 0.938802108168602,
       "step": 230
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 408.5989685058594,
-      "epoch": 0.6926536731634183,
-      "grad_norm": 33.38250055371977,
-      "kl": 5.451171875,
-      "learning_rate": 5.213798933941237e-06,
-      "loss": 1.5075,
-      "reward": 2.2532132267951965,
-      "reward_std": 1.0094469636678696,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.8767361044883728,
-      "rewards/repetition_penalty_reward": -0.003731334174517542,
-      "rewards/tag_count_reward": 0.7343750298023224,
+      "completion_length": 231.1197967529297,
+      "epoch": 0.3464566929133858,
+      "grad_norm": 15.68283618975048,
+      "kl": 3.099609375,
+      "learning_rate": 1.6523347356763572e-05,
+      "loss": 0.8542,
+      "reward": 2.695122182369232,
+      "reward_std": 0.4237038269639015,
+      "rewards/accuracy_reward": 0.8072916865348816,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.034478533547371626,
+      "rewards/tag_count_reward": 0.9570312649011612,
       "step": 231
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 442.53125762939453,
-      "epoch": 0.6956521739130435,
-      "grad_norm": 31.956941133841717,
-      "kl": 6.71875,
-      "learning_rate": 5.121811049098728e-06,
-      "loss": 1.5082,
-      "reward": 2.063460409641266,
-      "reward_std": 1.0131369531154633,
-      "rewards/accuracy_reward": 0.5104166865348816,
-      "rewards/reasoning_steps_reward": 0.864583358168602,
-      "rewards/repetition_penalty_reward": -0.006852172431536019,
-      "rewards/tag_count_reward": 0.6953125149011612,
+      "completion_length": 302.8698081970215,
+      "epoch": 0.3479565054368204,
+      "grad_norm": 193.87550647562585,
+      "kl": 9.875,
+      "learning_rate": 1.6483506393303807e-05,
+      "loss": 1.5385,
+      "reward": 2.4312188625335693,
+      "reward_std": 0.5264012217521667,
+      "rewards/accuracy_reward": 0.598958358168602,
+      "rewards/reasoning_steps_reward": 0.9496528208255768,
+      "rewards/repetition_penalty_reward": -0.03275693487375975,
+      "rewards/tag_count_reward": 0.915364608168602,
       "step": 232
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 461.07813262939453,
-      "epoch": 0.6986506746626686,
-      "grad_norm": 14.583503473769614,
-      "kl": 4.49609375,
-      "learning_rate": 5.030361696847706e-06,
-      "loss": 1.3138,
-      "reward": 2.1591392755508423,
-      "reward_std": 1.082452654838562,
-      "rewards/accuracy_reward": 0.6041666865348816,
-      "rewards/reasoning_steps_reward": 0.8767361640930176,
-      "rewards/repetition_penalty_reward": -0.004055280558532104,
-      "rewards/tag_count_reward": 0.6822916865348816,
+      "completion_length": 264.13542556762695,
+      "epoch": 0.34945631796025495,
+      "grad_norm": 36.22860317274317,
+      "kl": 0.8671875,
+      "learning_rate": 1.644348708726263e-05,
+      "loss": 0.6915,
+      "reward": 2.4658846259117126,
+      "reward_std": 0.6284756064414978,
+      "rewards/accuracy_reward": 0.5885416716337204,
+      "rewards/reasoning_steps_reward": 0.9618055820465088,
+      "rewards/repetition_penalty_reward": -0.023264775052666664,
+      "rewards/tag_count_reward": 0.9388020932674408,
       "step": 233
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 465.6146011352539,
-      "epoch": 0.7016491754122939,
-      "grad_norm": 31.93736272659645,
-      "kl": 2.73046875,
-      "learning_rate": 4.939460972831684e-06,
-      "loss": 1.1015,
-      "reward": 2.054699957370758,
-      "reward_std": 1.0152440816164017,
-      "rewards/accuracy_reward": 0.5208333432674408,
-      "rewards/reasoning_steps_reward": 0.8628472238779068,
-      "rewards/repetition_penalty_reward": -0.004761879798024893,
-      "rewards/tag_count_reward": 0.6757812649011612,
+      "completion_length": 286.3333435058594,
+      "epoch": 0.35095613048368957,
+      "grad_norm": 9.424469802346994,
+      "kl": 2.83984375,
+      "learning_rate": 1.640329053945585e-05,
+      "loss": 0.8706,
+      "reward": 2.417014181613922,
+      "reward_std": 0.6688820198178291,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.940972238779068,
+      "rewards/repetition_penalty_reward": -0.023958142613992095,
+      "rewards/tag_count_reward": 0.895833358168602,
       "step": 234
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 424.4479293823242,
-      "epoch": 0.704647676161919,
-      "grad_norm": 22.78953660465126,
-      "kl": 1.619140625,
-      "learning_rate": 4.849118912127817e-06,
-      "loss": 1.1123,
-      "reward": 2.1616870760917664,
-      "reward_std": 1.1319475769996643,
-      "rewards/accuracy_reward": 0.6354166865348816,
-      "rewards/reasoning_steps_reward": 0.8315972685813904,
-      "rewards/repetition_penalty_reward": -0.004545638337731361,
-      "rewards/tag_count_reward": 0.6992187649011612,
+      "completion_length": 234.5885467529297,
+      "epoch": 0.35245594300712413,
+      "grad_norm": 194.2142920957737,
+      "kl": 16.7421875,
+      "learning_rate": 1.6362917855574694e-05,
+      "loss": 1.9856,
+      "reward": 2.713364839553833,
+      "reward_std": 0.46920711547136307,
+      "rewards/accuracy_reward": 0.817708358168602,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.029690792318433523,
+      "rewards/tag_count_reward": 0.9479166716337204,
       "step": 235
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 432.97918701171875,
-      "epoch": 0.7076461769115442,
-      "grad_norm": 32.912310324473026,
-      "kl": 3.009765625,
-      "learning_rate": 4.759345488139054e-06,
-      "loss": 1.2209,
-      "reward": 2.037692070007324,
-      "reward_std": 1.0419560819864273,
-      "rewards/accuracy_reward": 0.5208333507180214,
-      "rewards/reasoning_steps_reward": 0.8315972238779068,
-      "rewards/repetition_penalty_reward": -0.0022385247866623104,
-      "rewards/tag_count_reward": 0.6875000298023224,
+      "completion_length": 288.375,
+      "epoch": 0.3539557555305587,
+      "grad_norm": 27.915108177299196,
+      "kl": 5.3359375,
+      "learning_rate": 1.6322370146155372e-05,
+      "loss": 1.2636,
+      "reward": 2.4878005385398865,
+      "reward_std": 0.5347049832344055,
+      "rewards/accuracy_reward": 0.630208358168602,
+      "rewards/reasoning_steps_reward": 0.9652778059244156,
+      "rewards/repetition_penalty_reward": -0.028258674778044224,
+      "rewards/tag_count_reward": 0.9205729514360428,
       "step": 236
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 415.2864685058594,
-      "epoch": 0.7106446776611695,
-      "grad_norm": 11.37715164851099,
-      "kl": 1.5185546875,
-      "learning_rate": 4.670150611493116e-06,
-      "loss": 1.1303,
-      "reward": 2.193199932575226,
-      "reward_std": 1.067429170012474,
-      "rewards/accuracy_reward": 0.6406250149011612,
-      "rewards/reasoning_steps_reward": 0.8437500447034836,
-      "rewards/repetition_penalty_reward": -0.002112621790729463,
-      "rewards/tag_count_reward": 0.7109375149011612,
+      "completion_length": 230.7604217529297,
+      "epoch": 0.35545556805399325,
+      "grad_norm": 17.59840442919455,
+      "kl": 0.9892578125,
+      "learning_rate": 1.6281648526548556e-05,
+      "loss": 0.6013,
+      "reward": 2.604456603527069,
+      "reward_std": 0.3973645493388176,
+      "rewards/accuracy_reward": 0.734375,
+      "rewards/reasoning_steps_reward": 0.9392361789941788,
+      "rewards/repetition_penalty_reward": -0.030092121567577124,
+      "rewards/tag_count_reward": 0.9609375149011612,
       "step": 237
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 432.12500762939453,
-      "epoch": 0.7136431784107946,
-      "grad_norm": 21.21508512413439,
-      "kl": 2.18359375,
-      "learning_rate": 4.581544128948413e-06,
-      "loss": 1.1413,
-      "reward": 2.1001856327056885,
-      "reward_std": 1.1043085157871246,
-      "rewards/accuracy_reward": 0.5885416865348816,
-      "rewards/reasoning_steps_reward": 0.822916716337204,
-      "rewards/repetition_penalty_reward": -0.005283167352899909,
-      "rewards/tag_count_reward": 0.6940104365348816,
+      "completion_length": 259.2395935058594,
+      "epoch": 0.3569553805774278,
+      "grad_norm": 24.885038866025347,
+      "kl": 1.595703125,
+      "learning_rate": 1.6240754116888673e-05,
+      "loss": 0.8111,
+      "reward": 2.5495967864990234,
+      "reward_std": 0.5210294723510742,
+      "rewards/accuracy_reward": 0.6822916865348816,
+      "rewards/reasoning_steps_reward": 0.9531250149011612,
+      "rewards/repetition_penalty_reward": -0.027226188685745,
+      "rewards/tag_count_reward": 0.9414062798023224,
       "step": 238
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 455.89064025878906,
-      "epoch": 0.7166416791604198,
-      "grad_norm": 3.7024792757899943,
-      "kl": 1.119140625,
-      "learning_rate": 4.493535822306993e-06,
-      "loss": 1.0115,
-      "reward": 1.9977717995643616,
-      "reward_std": 1.0762813091278076,
-      "rewards/accuracy_reward": 0.546875,
-      "rewards/reasoning_steps_reward": 0.7881944477558136,
-      "rewards/repetition_penalty_reward": -0.00396440684562549,
-      "rewards/tag_count_reward": 0.6666666865348816,
+      "completion_length": 241.46875762939453,
+      "epoch": 0.3584551931008624,
+      "grad_norm": 563.2951701790159,
+      "kl": 37.115234375,
+      "learning_rate": 1.6199688042063118e-05,
+      "loss": 4.5597,
+      "reward": 2.6697729229927063,
+      "reward_std": 0.3740931283682585,
+      "rewards/accuracy_reward": 0.7760416865348816,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.03378618019632995,
+      "rewards/tag_count_reward": 0.95703125,
       "step": 239
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 377.89583587646484,
-      "epoch": 0.719640179910045,
-      "grad_norm": 3.6760487188389717,
-      "kl": 0.953125,
-      "learning_rate": 4.406135407334669e-06,
-      "loss": 1.1086,
-      "reward": 2.1464386582374573,
-      "reward_std": 1.1137472093105316,
-      "rewards/accuracy_reward": 0.5989583432674408,
-      "rewards/reasoning_steps_reward": 0.8159722238779068,
-      "rewards/repetition_penalty_reward": -0.002866994822397828,
-      "rewards/tag_count_reward": 0.7343750298023224,
+      "completion_length": 207.78646087646484,
+      "epoch": 0.35995500562429694,
+      "grad_norm": 194.22792255309713,
+      "kl": 11.2109375,
+      "learning_rate": 1.6158451431681292e-05,
+      "loss": 2.2416,
+      "reward": 2.570555090904236,
+      "reward_std": 0.43018101900815964,
+      "rewards/accuracy_reward": 0.6510416865348816,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.0240629562176764,
+      "rewards/tag_count_reward": 0.9713541716337204,
       "step": 240
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 384.5364685058594,
-      "epoch": 0.7226386806596702,
-      "grad_norm": 19.529434988342086,
-      "kl": 1.9482421875,
-      "learning_rate": 4.319352532688444e-06,
-      "loss": 1.2755,
-      "reward": 2.1653665900230408,
-      "reward_std": 1.1197065114974976,
-      "rewards/accuracy_reward": 0.6250000149011612,
-      "rewards/reasoning_steps_reward": 0.8142361342906952,
-      "rewards/repetition_penalty_reward": -0.003036284673726186,
-      "rewards/tag_count_reward": 0.7291666865348816,
+      "completion_length": 227.88021087646484,
+      "epoch": 0.36145481814773156,
+      "grad_norm": 16.659702726061884,
+      "kl": 1.123046875,
+      "learning_rate": 1.6117045420043545e-05,
+      "loss": 0.5796,
+      "reward": 2.6110920906066895,
+      "reward_std": 0.44603806734085083,
+      "rewards/accuracy_reward": 0.6979166865348816,
+      "rewards/reasoning_steps_reward": 0.9809028059244156,
+      "rewards/repetition_penalty_reward": -0.03126898966729641,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 241
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 377.8020935058594,
-      "epoch": 0.7256371814092953,
-      "grad_norm": 46.6982672472414,
-      "kl": 2.0419921875,
-      "learning_rate": 4.2331967788513295e-06,
-      "loss": 1.2806,
-      "reward": 2.2072007060050964,
-      "reward_std": 1.220471739768982,
-      "rewards/accuracy_reward": 0.6979166865348816,
-      "rewards/reasoning_steps_reward": 0.7881944924592972,
-      "rewards/repetition_penalty_reward": -0.0028688511229120195,
-      "rewards/tag_count_reward": 0.723958358168602,
+      "completion_length": 225.95833587646484,
+      "epoch": 0.3629546306711661,
+      "grad_norm": 7.320890528168202,
+      "kl": 2.55078125,
+      "learning_rate": 1.6075471146109957e-05,
+      "loss": 0.7141,
+      "reward": 2.701682925224304,
+      "reward_std": 0.4787740185856819,
+      "rewards/accuracy_reward": 0.7968750149011612,
+      "rewards/reasoning_steps_reward": 0.9652778059244156,
+      "rewards/repetition_penalty_reward": -0.026615716982632875,
+      "rewards/tag_count_reward": 0.966145858168602,
       "step": 242
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 396.3802261352539,
-      "epoch": 0.7286356821589205,
-      "grad_norm": 51.973275524782814,
-      "kl": 2.06640625,
-      "learning_rate": 4.1476776570747065e-06,
-      "loss": 1.0713,
-      "reward": 1.942684918642044,
-      "reward_std": 1.1059914082288742,
-      "rewards/accuracy_reward": 0.4687500223517418,
-      "rewards/reasoning_steps_reward": 0.7656250447034836,
-      "rewards/repetition_penalty_reward": -0.005231783725321293,
-      "rewards/tag_count_reward": 0.7135416865348816,
+      "completion_length": 247.0260467529297,
+      "epoch": 0.3644544431946007,
+      "grad_norm": 112.81198496662604,
+      "kl": 10.6669921875,
+      "learning_rate": 1.603372975346903e-05,
+      "loss": 1.84,
+      "reward": 2.496858596801758,
+      "reward_std": 0.453279048204422,
+      "rewards/accuracy_reward": 0.6093750074505806,
+      "rewards/reasoning_steps_reward": 0.9618055820465088,
+      "rewards/repetition_penalty_reward": -0.02614493854343891,
+      "rewards/tag_count_reward": 0.9518229365348816,
       "step": 243
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 428.0104293823242,
-      "epoch": 0.7316341829085458,
-      "grad_norm": 4.890418211224268,
-      "kl": 1.2119140625,
-      "learning_rate": 4.0628046083283134e-06,
-      "loss": 1.0296,
-      "reward": 1.890177071094513,
-      "reward_std": 1.1783712059259415,
-      "rewards/accuracy_reward": 0.5000000149011612,
-      "rewards/reasoning_steps_reward": 0.7187500149011612,
-      "rewards/repetition_penalty_reward": -0.0030522070010192692,
-      "rewards/tag_count_reward": 0.6744791865348816,
+      "completion_length": 246.4895896911621,
+      "epoch": 0.36595425571803525,
+      "grad_norm": 21.266417808536865,
+      "kl": 2.26171875,
+      "learning_rate": 1.599182239030621e-05,
+      "loss": 0.7589,
+      "reward": 2.5812110900878906,
+      "reward_std": 0.5132277607917786,
+      "rewards/accuracy_reward": 0.6770833507180214,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.025559830013662577,
+      "rewards/tag_count_reward": 0.9505208432674408,
       "step": 244
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 387.3073043823242,
-      "epoch": 0.7346326836581709,
-      "grad_norm": 5.532289024750771,
-      "kl": 1.078125,
-      "learning_rate": 3.9785870022580075e-06,
-      "loss": 1.1278,
-      "reward": 2.0230748057365417,
-      "reward_std": 1.2535961270332336,
-      "rewards/accuracy_reward": 0.5833333432674408,
-      "rewards/reasoning_steps_reward": 0.7326389104127884,
-      "rewards/repetition_penalty_reward": -0.005137050000485033,
-      "rewards/tag_count_reward": 0.7122395932674408,
+      "completion_length": 256.6614646911621,
+      "epoch": 0.3674540682414698,
+      "grad_norm": 256.9075653283988,
+      "kl": 19.359375,
+      "learning_rate": 1.594975020937233e-05,
+      "loss": 2.8854,
+      "reward": 2.6606767177581787,
+      "reward_std": 0.455551914870739,
+      "rewards/accuracy_reward": 0.7760417014360428,
+      "rewards/reasoning_steps_reward": 0.9687500298023224,
+      "rewards/repetition_penalty_reward": -0.025521302595734596,
+      "rewards/tag_count_reward": 0.9414062649011612,
       "step": 245
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 348.17189025878906,
-      "epoch": 0.7376311844077961,
-      "grad_norm": 9.03974459360704,
-      "kl": 0.9912109375,
-      "learning_rate": 3.895034136151388e-06,
-      "loss": 1.2011,
-      "reward": 2.1917267441749573,
-      "reward_std": 1.1604805290699005,
-      "rewards/accuracy_reward": 0.661458358168602,
-      "rewards/reasoning_steps_reward": 0.7812499850988388,
-      "rewards/repetition_penalty_reward": -0.004887988092377782,
-      "rewards/tag_count_reward": 0.7539062649011612,
+      "completion_length": 260.8906364440918,
+      "epoch": 0.3689538807649044,
+      "grad_norm": 54.830877036198636,
+      "kl": 7.4140625,
+      "learning_rate": 1.590751436795186e-05,
+      "loss": 1.4602,
+      "reward": 2.603227972984314,
+      "reward_std": 0.5533816516399384,
+      "rewards/accuracy_reward": 0.7187500149011612,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.025678453966975212,
+      "rewards/tag_count_reward": 0.946614608168602,
       "step": 246
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 371.46876525878906,
-      "epoch": 0.7406296851574213,
-      "grad_norm": 10.803578128694179,
-      "kl": 0.986328125,
-      "learning_rate": 3.8121552339114166e-06,
-      "loss": 1.1353,
-      "reward": 2.1688408851623535,
-      "reward_std": 1.189782053232193,
-      "rewards/accuracy_reward": 0.645833358168602,
-      "rewards/reasoning_steps_reward": 0.7847222536802292,
-      "rewards/repetition_penalty_reward": -0.00390218710526824,
-      "rewards/tag_count_reward": 0.7421875298023224,
+      "completion_length": 281.8541679382324,
+      "epoch": 0.37045369328833894,
+      "grad_norm": 30.40863534370435,
+      "kl": 0.8173828125,
+      "learning_rate": 1.5865116027831123e-05,
+      "loss": 0.6622,
+      "reward": 2.464226543903351,
+      "reward_std": 0.5860697254538536,
+      "rewards/accuracy_reward": 0.5989583432674408,
+      "rewards/reasoning_steps_reward": 0.9722222536802292,
+      "rewards/repetition_penalty_reward": -0.030131183564662933,
+      "rewards/tag_count_reward": 0.9231770932674408,
       "step": 247
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 352.3073043823242,
-      "epoch": 0.7436281859070465,
-      "grad_norm": 15.262258125639379,
-      "kl": 0.8876953125,
-      "learning_rate": 3.729959445038136e-06,
-      "loss": 1.1792,
-      "reward": 2.193057656288147,
-      "reward_std": 1.1223880648612976,
-      "rewards/accuracy_reward": 0.6458333656191826,
-      "rewards/reasoning_steps_reward": 0.7951389104127884,
-      "rewards/repetition_penalty_reward": -0.004425037943292409,
-      "rewards/tag_count_reward": 0.7565104365348816,
+      "completion_length": 252.2604217529297,
+      "epoch": 0.37195350581177355,
+      "grad_norm": 23.77354715994608,
+      "kl": 0.6748046875,
+      "learning_rate": 1.5822556355266302e-05,
+      "loss": 0.6163,
+      "reward": 2.6034846901893616,
+      "reward_std": 0.5462495759129524,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9722222685813904,
+      "rewards/repetition_penalty_reward": -0.02498761680908501,
+      "rewards/tag_count_reward": 0.9479166865348816,
       "step": 248
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 383.93751525878906,
-      "epoch": 0.7466266866566716,
-      "grad_norm": 2160.279743864228,
-      "kl": 34.333984375,
-      "learning_rate": 3.6484558436185936e-06,
-      "loss": 4.9009,
-      "reward": 2.1113908290863037,
-      "reward_std": 1.0678627490997314,
-      "rewards/accuracy_reward": 0.5625000149011612,
-      "rewards/reasoning_steps_reward": 0.8263889253139496,
-      "rewards/repetition_penalty_reward": -0.005362731404602528,
-      "rewards/tag_count_reward": 0.7278645932674408,
+      "completion_length": 221.90105056762695,
+      "epoch": 0.3734533183352081,
+      "grad_norm": 14.516331745406989,
+      "kl": 1.0048828125,
+      "learning_rate": 1.577983652095137e-05,
+      "loss": 0.4184,
+      "reward": 2.332468032836914,
+      "reward_std": 0.43112215772271156,
+      "rewards/accuracy_reward": 0.4322916716337204,
+      "rewards/reasoning_steps_reward": 0.967013955116272,
+      "rewards/repetition_penalty_reward": -0.03168142307549715,
+      "rewards/tag_count_reward": 0.9648437649011612,
       "step": 249
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 367.54689025878906,
-      "epoch": 0.7496251874062968,
-      "grad_norm": 43.212145640693585,
-      "kl": 1.115234375,
-      "learning_rate": 3.5676534273251072e-06,
-      "loss": 1.1604,
-      "reward": 2.2231311798095703,
-      "reward_std": 1.064414381980896,
-      "rewards/accuracy_reward": 0.6406250074505806,
-      "rewards/reasoning_steps_reward": 0.8472222685813904,
-      "rewards/repetition_penalty_reward": -0.004299435357097536,
-      "rewards/tag_count_reward": 0.739583358168602,
+      "completion_length": 211.7447967529297,
+      "epoch": 0.3749531308586427,
+      "grad_norm": 46.523365934279404,
+      "kl": 4.62451171875,
+      "learning_rate": 1.5736957699985887e-05,
+      "loss": 0.9192,
+      "reward": 2.580030918121338,
+      "reward_std": 0.3245397238060832,
+      "rewards/accuracy_reward": 0.6510416865348816,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.031948281452059746,
+      "rewards/tag_count_reward": 0.9713541716337204,
       "step": 250
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 352.95314025878906,
-      "epoch": 0.7526236881559221,
-      "grad_norm": 1377.7529772591768,
-      "kl": 48.345703125,
-      "learning_rate": 3.487561116421958e-06,
-      "loss": 4.8422,
-      "reward": 2.2985642552375793,
-      "reward_std": 0.997696116566658,
-      "rewards/accuracy_reward": 0.6822917014360428,
-      "rewards/reasoning_steps_reward": 0.8715277910232544,
-      "rewards/repetition_penalty_reward": -0.003953163628466427,
-      "rewards/tag_count_reward": 0.7486979365348816,
+      "completion_length": 229.50521087646484,
+      "epoch": 0.37645294338207724,
+      "grad_norm": 46.335938002602354,
+      "kl": 4.9765625,
+      "learning_rate": 1.5693921071842688e-05,
+      "loss": 1.1019,
+      "reward": 2.6523364782333374,
+      "reward_std": 0.4070918932557106,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.03212546557188034,
+      "rewards/tag_count_reward": 0.9674479216337204,
       "step": 251
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 307.4739646911621,
-      "epoch": 0.7556221889055472,
-      "grad_norm": 237.6223200520559,
-      "kl": 7.603515625,
-      "learning_rate": 3.408187752780624e-06,
-      "loss": 1.9096,
-      "reward": 2.386752665042877,
-      "reward_std": 0.99474136531353,
-      "rewards/accuracy_reward": 0.7083333432674408,
-      "rewards/reasoning_steps_reward": 0.901041716337204,
-      "rewards/repetition_penalty_reward": -0.01298706023953855,
-      "rewards/tag_count_reward": 0.7903646230697632,
+      "completion_length": 237.23438262939453,
+      "epoch": 0.3779527559055118,
+      "grad_norm": 19.678447582967497,
+      "kl": 1.244140625,
+      "learning_rate": 1.5650727820335417e-05,
+      "loss": 0.624,
+      "reward": 2.570410132408142,
+      "reward_std": 0.5100410208106041,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.032888590823858976,
+      "rewards/tag_count_reward": 0.9453125149011612,
       "step": 252
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 272.8541793823242,
-      "epoch": 0.7586206896551724,
-      "grad_norm": 103.11957636266683,
-      "kl": 1.3759765625,
-      "learning_rate": 3.329542098903674e-06,
-      "loss": 1.2511,
-      "reward": 2.4556241035461426,
-      "reward_std": 0.754611574113369,
-      "rewards/accuracy_reward": 0.708333358168602,
-      "rewards/reasoning_steps_reward": 0.9288194626569748,
-      "rewards/repetition_penalty_reward": -0.010955846635624766,
-      "rewards/tag_count_reward": 0.829427108168602,
+      "completion_length": 247.0572967529297,
+      "epoch": 0.37945256842894637,
+      "grad_norm": 10.000060356187625,
+      "kl": 1.7998046875,
+      "learning_rate": 1.5607379133585978e-05,
+      "loss": 0.4729,
+      "reward": 2.7505863904953003,
+      "reward_std": 0.4299619048833847,
+      "rewards/accuracy_reward": 0.859375,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.02675750060006976,
+      "rewards/tag_count_reward": 0.9544270932674408,
       "step": 253
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 216.80209350585938,
-      "epoch": 0.7616191904047976,
-      "grad_norm": 72.74924199958468,
-      "kl": 1.1376953125,
-      "learning_rate": 3.2516328369574247e-06,
-      "loss": 1.1836,
-      "reward": 2.5425453782081604,
-      "reward_std": 0.6245445907115936,
-      "rewards/accuracy_reward": 0.713541679084301,
-      "rewards/reasoning_steps_reward": 0.9479166865348816,
-      "rewards/repetition_penalty_reward": -0.005631837877444923,
-      "rewards/tag_count_reward": 0.8867187649011612,
+      "completion_length": 255.86980056762695,
+      "epoch": 0.38095238095238093,
+      "grad_norm": 41.51304950449503,
+      "kl": 6.734375,
+      "learning_rate": 1.5563876203991856e-05,
+      "loss": 1.5806,
+      "reward": 2.535602629184723,
+      "reward_std": 0.5331927165389061,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.958333358168602,
+      "rewards/repetition_penalty_reward": -0.02299123560078442,
+      "rewards/tag_count_reward": 0.9335937649011612,
       "step": 254
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 288.1458396911621,
-      "epoch": 0.7646176911544228,
-      "grad_norm": 18.047951176172486,
-      "kl": 0.7626953125,
-      "learning_rate": 3.174468567813461e-06,
-      "loss": 1.2569,
-      "reward": 2.4596019983291626,
-      "reward_std": 0.9773845970630646,
-      "rewards/accuracy_reward": 0.755208358168602,
-      "rewards/reasoning_steps_reward": 0.8854167014360428,
-      "rewards/repetition_penalty_reward": -0.003939674003049731,
-      "rewards/tag_count_reward": 0.8229166865348816,
+      "completion_length": 231.61979293823242,
+      "epoch": 0.38245219347581555,
+      "grad_norm": 18.09973578264405,
+      "kl": 4.19921875,
+      "learning_rate": 1.55202202281933e-05,
+      "loss": 0.9442,
+      "reward": 2.68158096075058,
+      "reward_std": 0.38162920624017715,
+      "rewards/accuracy_reward": 0.7656250298023224,
+      "rewards/reasoning_steps_reward": 0.9826389253139496,
+      "rewards/repetition_penalty_reward": -0.025016394443809986,
+      "rewards/tag_count_reward": 0.9583333432674408,
       "step": 255
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 279.9635467529297,
-      "epoch": 0.767616191904048,
-      "grad_norm": 2617.8075543625873,
-      "kl": 1.8046875,
-      "learning_rate": 3.0980578100991356e-06,
-      "loss": 0.9846,
-      "reward": 2.420164167881012,
-      "reward_std": 0.901138424873352,
+      "completion_length": 236.7083396911621,
+      "epoch": 0.3839520059992501,
+      "grad_norm": 23.810294289503528,
+      "kl": 1.501953125,
+      "learning_rate": 1.5476412407040445e-05,
+      "loss": 0.6148,
+      "reward": 2.6228463649749756,
+      "reward_std": 0.43365930393338203,
       "rewards/accuracy_reward": 0.739583358168602,
-      "rewards/reasoning_steps_reward": 0.8559027910232544,
-      "rewards/repetition_penalty_reward": -0.0034470059908926487,
-      "rewards/tag_count_reward": 0.8281250149011612,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.02689332398585975,
+      "rewards/tag_count_reward": 0.946614608168602,
       "step": 256
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 244.65104293823242,
-      "epoch": 0.7706146926536732,
-      "grad_norm": 10.063702176948555,
-      "kl": 0.7568359375,
-      "learning_rate": 3.022408999257148e-06,
-      "loss": 1.0871,
-      "reward": 2.4764610528945923,
-      "reward_std": 0.8722782582044601,
-      "rewards/accuracy_reward": 0.7656250298023224,
-      "rewards/reasoning_steps_reward": 0.8697917014360428,
-      "rewards/repetition_penalty_reward": -0.004007782437838614,
-      "rewards/tag_count_reward": 0.845052108168602,
+      "completion_length": 292.3333435058594,
+      "epoch": 0.3854518185226847,
+      "grad_norm": 40.613619721197324,
+      "kl": 2.0693359375,
+      "learning_rate": 1.5432453945560223e-05,
+      "loss": 0.8716,
+      "reward": 2.567777395248413,
+      "reward_std": 0.5245895758271217,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9531250149011612,
+      "rewards/repetition_penalty_reward": -0.028576852288097143,
+      "rewards/tag_count_reward": 0.9140625298023224,
       "step": 257
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 236.68750762939453,
-      "epoch": 0.7736131934032984,
-      "grad_norm": 57.97300190458974,
-      "kl": 0.9248046875,
-      "learning_rate": 2.947530486614303e-06,
-      "loss": 1.1472,
-      "reward": 2.488058924674988,
-      "reward_std": 0.8399006128311157,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9062500149011612,
-      "rewards/repetition_penalty_reward": -0.008034905651584268,
-      "rewards/tag_count_reward": 0.860677108168602,
+      "completion_length": 249.99480438232422,
+      "epoch": 0.38695163104611924,
+      "grad_norm": 21.07252470454373,
+      "kl": 3.251953125,
+      "learning_rate": 1.5388346052923268e-05,
+      "loss": 1.0205,
+      "reward": 2.599771797657013,
+      "reward_std": 0.545002818107605,
+      "rewards/accuracy_reward": 0.7187500074505806,
+      "rewards/reasoning_steps_reward": 0.9722222536802292,
+      "rewards/repetition_penalty_reward": -0.028700455324724317,
+      "rewards/tag_count_reward": 0.9375000298023224,
       "step": 258
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 290.46875762939453,
-      "epoch": 0.7766116941529235,
-      "grad_norm": 911.2077972417934,
-      "kl": 33.609375,
-      "learning_rate": 2.8734305384595598e-06,
-      "loss": 1.6782,
-      "reward": 2.371084213256836,
-      "reward_std": 0.9771549105644226,
-      "rewards/accuracy_reward": 0.6822916865348816,
-      "rewards/reasoning_steps_reward": 0.8663194626569748,
-      "rewards/repetition_penalty_reward": -0.005651986633893102,
-      "rewards/tag_count_reward": 0.8281250149011612,
+      "completion_length": 210.9166717529297,
+      "epoch": 0.3884514435695538,
+      "grad_norm": 142.3585656392472,
+      "kl": 10.0546875,
+      "learning_rate": 1.534408994241063e-05,
+      "loss": 1.7735,
+      "reward": 2.710343062877655,
+      "reward_std": 0.4035199508070946,
+      "rewards/accuracy_reward": 0.7916666716337204,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.03227862901985645,
+      "rewards/tag_count_reward": 0.9752604216337204,
       "step": 259
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 237.38542556762695,
-      "epoch": 0.7796101949025487,
-      "grad_norm": 326.19482464697984,
-      "kl": 3.037109375,
-      "learning_rate": 2.8001173351314625e-06,
-      "loss": 1.277,
-      "reward": 2.5145729780197144,
-      "reward_std": 0.8064324706792831,
-      "rewards/accuracy_reward": 0.770833358168602,
-      "rewards/reasoning_steps_reward": 0.8888888955116272,
-      "rewards/repetition_penalty_reward": -0.007128427678253502,
-      "rewards/tag_count_reward": 0.8619791716337204,
+      "completion_length": 237.3958396911621,
+      "epoch": 0.38995125609298836,
+      "grad_norm": 377.6752575854368,
+      "kl": 24.46875,
+      "learning_rate": 1.5299686831380395e-05,
+      "loss": 3.8202,
+      "reward": 2.407591462135315,
+      "reward_std": 0.3708171471953392,
+      "rewards/accuracy_reward": 0.5208333414047956,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.04206139035522938,
+      "rewards/tag_count_reward": 0.958333358168602,
       "step": 260
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 247.31771087646484,
-      "epoch": 0.782608695652174,
-      "grad_norm": 38.273912583736674,
-      "kl": 1.0126953125,
-      "learning_rate": 2.7275989701150684e-06,
-      "loss": 1.3591,
-      "reward": 2.555409252643585,
-      "reward_std": 0.9130035191774368,
-      "rewards/accuracy_reward": 0.8125000149011612,
-      "rewards/reasoning_steps_reward": 0.8854166865348816,
-      "rewards/repetition_penalty_reward": -0.004486680845730007,
-      "rewards/tag_count_reward": 0.8619791865348816,
+      "completion_length": 232.5989646911621,
+      "epoch": 0.3914510686164229,
+      "grad_norm": 930.2326367615931,
+      "kl": 29.53125,
+      "learning_rate": 1.5255137941234228e-05,
+      "loss": 5.1202,
+      "reward": 2.6035396456718445,
+      "reward_std": 0.47209347784519196,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.01972440304234624,
+      "rewards/tag_count_reward": 0.9635417014360428,
       "step": 261
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 211.36458587646484,
-      "epoch": 0.7856071964017991,
-      "grad_norm": 63.967595082731656,
-      "kl": 0.875,
-      "learning_rate": 2.6558834491484576e-06,
-      "loss": 1.0617,
-      "reward": 2.4956788420677185,
-      "reward_std": 0.7778256386518478,
-      "rewards/accuracy_reward": 0.708333358168602,
-      "rewards/reasoning_steps_reward": 0.9062500149011612,
-      "rewards/repetition_penalty_reward": -0.005623297765851021,
-      "rewards/tag_count_reward": 0.8867187798023224,
+      "completion_length": 253.92188262939453,
+      "epoch": 0.39295088113985754,
+      "grad_norm": 56.25971418456335,
+      "kl": 7.8515625,
+      "learning_rate": 1.5210444497383745e-05,
+      "loss": 1.6828,
+      "reward": 2.4639751315116882,
+      "reward_std": 0.519036740064621,
+      "rewards/accuracy_reward": 0.5885416865348816,
+      "rewards/reasoning_steps_reward": 0.958333358168602,
+      "rewards/repetition_penalty_reward": -0.020400056848302484,
+      "rewards/tag_count_reward": 0.9375000149011612,
       "step": 262
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 203.95312881469727,
-      "epoch": 0.7886056971514243,
-      "grad_norm": 195.88629364008162,
-      "kl": 1.525390625,
-      "learning_rate": 2.5849786893389296e-06,
-      "loss": 1.5584,
-      "reward": 2.52648663520813,
-      "reward_std": 0.8215616047382355,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9131944626569748,
-      "rewards/repetition_penalty_reward": -0.0038954283227212727,
-      "rewards/tag_count_reward": 0.8880208432674408,
+      "completion_length": 206.9895896911621,
+      "epoch": 0.3944506936632921,
+      "grad_norm": 37.53129740951539,
+      "kl": 0.9560546875,
+      "learning_rate": 1.5165607729216822e-05,
+      "loss": 0.5769,
+      "reward": 2.6286195516586304,
+      "reward_std": 0.4291228875517845,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.97743059694767,
+      "rewards/repetition_penalty_reward": -0.025894425809383392,
+      "rewards/tag_count_reward": 0.9687500149011612,
       "step": 263
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 234.9270896911621,
-      "epoch": 0.7916041979010495,
-      "grad_norm": 111.05889429266634,
-      "kl": 2.357421875,
-      "learning_rate": 2.514892518288988e-06,
-      "loss": 1.6347,
-      "reward": 2.5823854207992554,
-      "reward_std": 0.9507413059473038,
-      "rewards/accuracy_reward": 0.848958358168602,
-      "rewards/reasoning_steps_reward": 0.8750000149011612,
-      "rewards/repetition_penalty_reward": -0.004854308412177488,
-      "rewards/tag_count_reward": 0.8632812649011612,
+      "completion_length": 224.0260467529297,
+      "epoch": 0.39595050618672667,
+      "grad_norm": 40.41532757864787,
+      "kl": 0.7841796875,
+      "learning_rate": 1.5120628870063772e-05,
+      "loss": 0.5366,
+      "reward": 2.569106698036194,
+      "reward_std": 0.5048164129257202,
+      "rewards/accuracy_reward": 0.6770833507180214,
+      "rewards/reasoning_steps_reward": 0.9548611342906952,
+      "rewards/repetition_penalty_reward": -0.02377540967427194,
+      "rewards/tag_count_reward": 0.9609375149011612,
       "step": 264
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 236.2187614440918,
-      "epoch": 0.7946026986506747,
-      "grad_norm": 71.84249826406462,
-      "kl": 2.79296875,
-      "learning_rate": 2.445632673232208e-06,
-      "loss": 1.4097,
-      "reward": 2.507424294948578,
-      "reward_std": 0.8585045337677002,
-      "rewards/accuracy_reward": 0.770833358168602,
-      "rewards/reasoning_steps_reward": 0.8802083432674408,
-      "rewards/repetition_penalty_reward": -0.002992400841321796,
-      "rewards/tag_count_reward": 0.8593750298023224,
+      "completion_length": 232.37500381469727,
+      "epoch": 0.39745031871016123,
+      "grad_norm": 36.78339467024414,
+      "kl": 1.271484375,
+      "learning_rate": 1.5075509157163422e-05,
+      "loss": 0.6666,
+      "reward": 2.532996416091919,
+      "reward_std": 0.406702384352684,
+      "rewards/accuracy_reward": 0.614583358168602,
+      "rewards/reasoning_steps_reward": 0.9739583730697632,
+      "rewards/repetition_penalty_reward": -0.017784894444048405,
+      "rewards/tag_count_reward": 0.962239608168602,
       "step": 265
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 287.5520935058594,
-      "epoch": 0.7976011994002998,
-      "grad_norm": 952.7545080808868,
-      "kl": 22.484375,
-      "learning_rate": 2.3772068001790682e-06,
-      "loss": 3.905,
-      "reward": 2.385596811771393,
-      "reward_std": 0.9874599426984787,
-      "rewards/accuracy_reward": 0.6979166865348816,
-      "rewards/reasoning_steps_reward": 0.8645833730697632,
-      "rewards/repetition_penalty_reward": -0.0063302937196567655,
-      "rewards/tag_count_reward": 0.8294270932674408,
+      "completion_length": 269.87500762939453,
+      "epoch": 0.3989501312335958,
+      "grad_norm": 5.827263683756955,
+      "kl": 5.0390625,
+      "learning_rate": 1.503024983162908e-05,
+      "loss": 1.1664,
+      "reward": 2.5889384150505066,
+      "reward_std": 0.5250253453850746,
+      "rewards/accuracy_reward": 0.723958358168602,
+      "rewards/reasoning_steps_reward": 0.9618055820465088,
+      "rewards/repetition_penalty_reward": -0.025211102329194546,
+      "rewards/tag_count_reward": 0.9283854216337204,
       "step": 266
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 205.85938262939453,
-      "epoch": 0.800599700149925,
-      "grad_norm": 345.0356082309632,
-      "kl": 10.75,
-      "learning_rate": 2.309622453072867e-06,
-      "loss": 2.8112,
-      "reward": 2.638276517391205,
-      "reward_std": 0.6796813756227493,
-      "rewards/accuracy_reward": 0.8229166716337204,
-      "rewards/reasoning_steps_reward": 0.9236111342906952,
-      "rewards/repetition_penalty_reward": -0.005386793913203292,
-      "rewards/tag_count_reward": 0.8971354514360428,
+      "completion_length": 237.25000381469727,
+      "epoch": 0.40044994375703036,
+      "grad_norm": 50.76618742939356,
+      "kl": 6.140625,
+      "learning_rate": 1.49848521384144e-05,
+      "loss": 1.4081,
+      "reward": 2.7229838967323303,
+      "reward_std": 0.47072865813970566,
+      "rewards/accuracy_reward": 0.8229166865348816,
+      "rewards/reasoning_steps_reward": 0.970486119389534,
+      "rewards/repetition_penalty_reward": -0.03396061668172479,
+      "rewards/tag_count_reward": 0.9635416865348816,
       "step": 267
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 261.25000381469727,
-      "epoch": 0.8035982008995503,
-      "grad_norm": 243.7268824601945,
-      "kl": 3.1484375,
-      "learning_rate": 2.2428870929558012e-06,
-      "loss": 1.6183,
-      "reward": 2.373494029045105,
-      "reward_std": 0.9536427110433578,
-      "rewards/accuracy_reward": 0.6718750149011612,
-      "rewards/reasoning_steps_reward": 0.8663194626569748,
-      "rewards/repetition_penalty_reward": -0.005846357671543956,
-      "rewards/tag_count_reward": 0.8411458432674408,
+      "completion_length": 216.90625762939453,
+      "epoch": 0.4019497562804649,
+      "grad_norm": 5.736957116926136,
+      "kl": 3.244140625,
+      "learning_rate": 1.4939317326279125e-05,
+      "loss": 1.0251,
+      "reward": 2.5634613633155823,
+      "reward_std": 0.4809972904622555,
+      "rewards/accuracy_reward": 0.6718750298023224,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.02681647054851055,
+      "rewards/tag_count_reward": 0.953125,
       "step": 268
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 212.04688262939453,
-      "epoch": 0.8065967016491754,
-      "grad_norm": 203.47240043545727,
-      "kl": 2.35546875,
-      "learning_rate": 2.177008087145286e-06,
-      "loss": 1.358,
-      "reward": 2.610089957714081,
-      "reward_std": 0.7462358474731445,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 0.923611119389534,
-      "rewards/repetition_penalty_reward": -0.006229604536201805,
-      "rewards/tag_count_reward": 0.8958333432674408,
+      "completion_length": 234.33855056762695,
+      "epoch": 0.40344956880389954,
+      "grad_norm": 23.899798832527537,
+      "kl": 4.408203125,
+      "learning_rate": 1.489364664775475e-05,
+      "loss": 0.9419,
+      "reward": 2.684835433959961,
+      "reward_std": 0.39333367347717285,
+      "rewards/accuracy_reward": 0.770833358168602,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.03261255333200097,
+      "rewards/tag_count_reward": 0.9622395932674408,
       "step": 269
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 220.5156364440918,
-      "epoch": 0.8095952023988006,
-      "grad_norm": 104.0415374329618,
-      "kl": 4.044921875,
-      "learning_rate": 2.111992708420646e-06,
-      "loss": 1.7589,
-      "reward": 2.612178146839142,
-      "reward_std": 0.8539304882287979,
-      "rewards/accuracy_reward": 0.8385416865348816,
-      "rewards/reasoning_steps_reward": 0.89930559694767,
-      "rewards/repetition_penalty_reward": -0.004575358587317169,
-      "rewards/tag_count_reward": 0.8789062798023224,
+      "completion_length": 237.65625762939453,
+      "epoch": 0.4049493813273341,
+      "grad_norm": 2.2657313852356853,
+      "kl": 2.919921875,
+      "learning_rate": 1.4847841359110058e-05,
+      "loss": 0.7915,
+      "reward": 2.613522946834564,
+      "reward_std": 0.3725850097835064,
+      "rewards/accuracy_reward": 0.7135416939854622,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.027970252325758338,
+      "rewards/tag_count_reward": 0.9557291865348816,
       "step": 270
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 228.5260467529297,
-      "epoch": 0.8125937031484258,
-      "grad_norm": 162.74223326789044,
-      "kl": 11.875,
-      "learning_rate": 2.047848134220213e-06,
-      "loss": 2.9084,
-      "reward": 2.4650405645370483,
-      "reward_std": 0.8374519795179367,
-      "rewards/accuracy_reward": 0.6979166865348816,
-      "rewards/reasoning_steps_reward": 0.9027778059244156,
-      "rewards/repetition_penalty_reward": -0.008049842552281916,
-      "rewards/tag_count_reward": 0.872395858168602,
+      "completion_length": 230.23438262939453,
+      "epoch": 0.40644919385076866,
+      "grad_norm": 11.972730389593089,
+      "kl": 3.646484375,
+      "learning_rate": 1.480190272031657e-05,
+      "loss": 0.8564,
+      "reward": 2.5715832710266113,
+      "reward_std": 0.4187803417444229,
+      "rewards/accuracy_reward": 0.6770833507180214,
+      "rewards/reasoning_steps_reward": 0.9704861640930176,
+      "rewards/repetition_penalty_reward": -0.030413302592933178,
+      "rewards/tag_count_reward": 0.954427108168602,
       "step": 271
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 274.6197929382324,
-      "epoch": 0.815592203898051,
-      "grad_norm": 103.66672953735151,
-      "kl": 8.3125,
-      "learning_rate": 1.984581445848981e-06,
-      "loss": 2.2752,
-      "reward": 2.4658045768737793,
-      "reward_std": 0.9265211671590805,
-      "rewards/accuracy_reward": 0.7656250298023224,
-      "rewards/reasoning_steps_reward": 0.8697917014360428,
-      "rewards/repetition_penalty_reward": -0.00554958607244771,
-      "rewards/tag_count_reward": 0.8359375298023224,
+      "completion_length": 203.9166717529297,
+      "epoch": 0.4079490063742032,
+      "grad_norm": 17.38669681328992,
+      "kl": 1.04296875,
+      "learning_rate": 1.475583199501389e-05,
+      "loss": 0.4368,
+      "reward": 2.6774919033050537,
+      "reward_std": 0.440396323800087,
+      "rewards/accuracy_reward": 0.7604166865348816,
+      "rewards/reasoning_steps_reward": 0.9722222238779068,
+      "rewards/repetition_penalty_reward": -0.03170954994857311,
+      "rewards/tag_count_reward": 0.9765625149011612,
       "step": 272
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 287.3697967529297,
-      "epoch": 0.8185907046476761,
-      "grad_norm": 106.41271848435063,
-      "kl": 6.5703125,
-      "learning_rate": 1.9221996276968523e-06,
-      "loss": 2.0641,
-      "reward": 2.389641523361206,
-      "reward_std": 1.0108384490013123,
-      "rewards/accuracy_reward": 0.7135416865348816,
-      "rewards/reasoning_steps_reward": 0.8628472536802292,
-      "rewards/repetition_penalty_reward": -0.007059881230816245,
-      "rewards/tag_count_reward": 0.8203125149011612,
+      "completion_length": 230.77605056762695,
+      "epoch": 0.4094488188976378,
+      "grad_norm": 14.315967716823838,
+      "kl": 1.7646484375,
+      "learning_rate": 1.4709630450474936e-05,
+      "loss": 0.679,
+      "reward": 2.7193071246147156,
+      "reward_std": 0.4441717490553856,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9618055820465088,
+      "rewards/repetition_penalty_reward": -0.03676935099065304,
+      "rewards/tag_count_reward": 0.9557291865348816,
       "step": 273
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 218.31250762939453,
-      "epoch": 0.8215892053973014,
-      "grad_norm": 78.45170527728588,
-      "kl": 3.9296875,
-      "learning_rate": 1.8607095664675868e-06,
-      "loss": 1.7482,
-      "reward": 2.5906487703323364,
-      "reward_std": 0.8199838548898697,
-      "rewards/accuracy_reward": 0.8020833432674408,
-      "rewards/reasoning_steps_reward": 0.9062500149011612,
-      "rewards/repetition_penalty_reward": -0.007007573178270832,
-      "rewards/tag_count_reward": 0.8893229365348816,
+      "completion_length": 216.60938262939453,
+      "epoch": 0.41094863142107235,
+      "grad_norm": 12.679917824503232,
+      "kl": 2.2265625,
+      "learning_rate": 1.466329935757109e-05,
+      "loss": 0.6142,
+      "reward": 2.5890852212905884,
+      "reward_std": 0.3576386868953705,
+      "rewards/accuracy_reward": 0.6614583507180214,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.03157464787364006,
+      "rewards/tag_count_reward": 0.9713541865348816,
       "step": 274
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 255.67708587646484,
-      "epoch": 0.8245877061469266,
-      "grad_norm": 28.349825823882362,
-      "kl": 5.97265625,
-      "learning_rate": 1.8001180504185401e-06,
-      "loss": 1.9051,
-      "reward": 2.504952549934387,
-      "reward_std": 0.9115692526102066,
-      "rewards/accuracy_reward": 0.755208358168602,
-      "rewards/reasoning_steps_reward": 0.8923611491918564,
-      "rewards/repetition_penalty_reward": -0.0032940262462943792,
-      "rewards/tag_count_reward": 0.860677108168602,
+      "completion_length": 210.6302146911621,
+      "epoch": 0.4124484439445069,
+      "grad_norm": 6.724833772098711,
+      "kl": 1.8125,
+      "learning_rate": 1.4616839990737232e-05,
+      "loss": 0.4751,
+      "reward": 2.6611116528511047,
+      "reward_std": 0.3041349947452545,
+      "rewards/accuracy_reward": 0.7291666716337204,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.028124571312218904,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 275
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 223.79688262939453,
-      "epoch": 0.8275862068965517,
-      "grad_norm": 82.66659054217332,
-      "kl": 9.0390625,
-      "learning_rate": 1.7404317686112638e-06,
-      "loss": 2.4944,
-      "reward": 2.659183382987976,
-      "reward_std": 0.785422757267952,
-      "rewards/accuracy_reward": 0.8541667014360428,
-      "rewards/reasoning_steps_reward": 0.923611119389534,
-      "rewards/repetition_penalty_reward": -0.007917315931990743,
-      "rewards/tag_count_reward": 0.8893229365348816,
+      "completion_length": 241.07812881469727,
+      "epoch": 0.41394825646794153,
+      "grad_norm": 12.361659642014535,
+      "kl": 3.11328125,
+      "learning_rate": 1.4570253627936693e-05,
+      "loss": 0.8863,
+      "reward": 2.647101581096649,
+      "reward_std": 0.40448398888111115,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.030849804636090994,
+      "rewards/tag_count_reward": 0.9453125149011612,
       "step": 276
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 225.1197967529297,
-      "epoch": 0.8305847076461769,
-      "grad_norm": 115.53315284387001,
-      "kl": 9.71875,
-      "learning_rate": 1.6816573101730637e-06,
-      "loss": 2.136,
-      "reward": 2.557182252407074,
-      "reward_std": 0.806065671145916,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 0.8923610895872116,
-      "rewards/repetition_penalty_reward": -0.004449750704225153,
-      "rewards/tag_count_reward": 0.8723958432674408,
+      "completion_length": 203.4791717529297,
+      "epoch": 0.4154480689913761,
+      "grad_norm": 10.28432474158588,
+      "kl": 0.76171875,
+      "learning_rate": 1.4523541550626093e-05,
+      "loss": 0.349,
+      "reward": 2.594405770301819,
+      "reward_std": 0.1998637057840824,
+      "rewards/accuracy_reward": 0.6666667014360428,
+      "rewards/reasoning_steps_reward": 0.9843750298023224,
+      "rewards/repetition_penalty_reward": -0.04101098608225584,
+      "rewards/tag_count_reward": 0.984375,
       "step": 277
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 259.59375762939453,
-      "epoch": 0.8335832083958021,
-      "grad_norm": 55.03228946038301,
-      "kl": 5.38671875,
-      "learning_rate": 1.6238011635695849e-06,
-      "loss": 1.6557,
-      "reward": 2.512112319469452,
-      "reward_std": 0.8234718143939972,
-      "rewards/accuracy_reward": 0.7968750298023224,
-      "rewards/reasoning_steps_reward": 0.8802083730697632,
-      "rewards/repetition_penalty_reward": -0.0048148492351174355,
-      "rewards/tag_count_reward": 0.8398437649011612,
+      "completion_length": 200.7708396911621,
+      "epoch": 0.41694788151481066,
+      "grad_norm": 8.121334172436347,
+      "kl": 0.89990234375,
+      "learning_rate": 1.4476705043720099e-05,
+      "loss": 0.3742,
+      "reward": 2.8078061938285828,
+      "reward_std": 0.304847190156579,
+      "rewards/accuracy_reward": 0.8593750149011612,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.02552723279222846,
+      "rewards/tag_count_reward": 0.984375,
       "step": 278
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 262.9635543823242,
-      "epoch": 0.8365817091454273,
-      "grad_norm": 64.0876715179611,
-      "kl": 6.34375,
-      "learning_rate": 1.5668697158885104e-06,
-      "loss": 2.0553,
-      "reward": 2.3902793526649475,
-      "reward_std": 0.9917967170476913,
-      "rewards/accuracy_reward": 0.6979166865348816,
-      "rewards/reasoning_steps_reward": 0.8541666567325592,
-      "rewards/repetition_penalty_reward": -0.0029499368683900684,
-      "rewards/tag_count_reward": 0.8411458432674408,
+      "completion_length": 205.4895896911621,
+      "epoch": 0.4184476940382452,
+      "grad_norm": 8.922984245945559,
+      "kl": 1.37744140625,
+      "learning_rate": 1.4429745395556073e-05,
+      "loss": 0.5918,
+      "reward": 2.665529727935791,
+      "reward_std": 0.33949872478842735,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555671453476,
+      "rewards/repetition_penalty_reward": -0.024140472058206797,
+      "rewards/tag_count_reward": 0.9674479365348816,
       "step": 279
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 214.36980056762695,
-      "epoch": 0.8395802098950524,
-      "grad_norm": 71.70526659015454,
-      "kl": 8.18359375,
-      "learning_rate": 1.5108692521344526e-06,
-      "loss": 2.302,
-      "reward": 2.5640260577201843,
-      "reward_std": 0.7752020061016083,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.9097222536802292,
-      "rewards/repetition_penalty_reward": -0.005852422269526869,
-      "rewards/tag_count_reward": 0.8893229365348816,
+      "completion_length": 197.20312881469727,
+      "epoch": 0.4199475065616798,
+      "grad_norm": 11.885955592817725,
+      "kl": 1.419921875,
+      "learning_rate": 1.4382663897858647e-05,
+      "loss": 0.3527,
+      "reward": 2.600801944732666,
+      "reward_std": 0.2686375230550766,
+      "rewards/accuracy_reward": 0.6510416865348816,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.030708489008247852,
+      "rewards/tag_count_reward": 0.985677108168602,
       "step": 280
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 207.5260467529297,
-      "epoch": 0.8425787106446777,
-      "grad_norm": 41.429427085551154,
-      "kl": 7.34765625,
-      "learning_rate": 1.4558059545351144e-06,
-      "loss": 2.4269,
-      "reward": 2.5451850295066833,
-      "reward_std": 0.7376502603292465,
-      "rewards/accuracy_reward": 0.7447917014360428,
-      "rewards/reasoning_steps_reward": 0.9184028208255768,
-      "rewards/repetition_penalty_reward": -0.004728195344796404,
-      "rewards/tag_count_reward": 0.8867187798023224,
+      "completion_length": 277.1823043823242,
+      "epoch": 0.42144731908511435,
+      "grad_norm": 34.196754104798,
+      "kl": 5.865234375,
+      "learning_rate": 1.4335461845704173e-05,
+      "loss": 1.1386,
+      "reward": 2.416446477174759,
+      "reward_std": 0.47412654757499695,
+      "rewards/accuracy_reward": 0.5729166865348816,
+      "rewards/reasoning_steps_reward": 0.9548611342906952,
+      "rewards/repetition_penalty_reward": -0.024091816041618586,
+      "rewards/tag_count_reward": 0.9127604365348816,
       "step": 281
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 176.7864646911621,
-      "epoch": 0.8455772113943029,
-      "grad_norm": 83.02833138285867,
-      "kl": 2.6708984375,
-      "learning_rate": 1.4016859018587958e-06,
-      "loss": 1.3215,
-      "reward": 2.6956359148025513,
-      "reward_std": 0.533905953168869,
-      "rewards/accuracy_reward": 0.8437500298023224,
-      "rewards/reasoning_steps_reward": 0.9357638955116272,
-      "rewards/repetition_penalty_reward": -0.005753145087510347,
-      "rewards/tag_count_reward": 0.9218750149011612,
+      "completion_length": 220.23958587646484,
+      "epoch": 0.4229471316085489,
+      "grad_norm": 7.827483997656616,
+      "kl": 2.251953125,
+      "learning_rate": 1.428814053748512e-05,
+      "loss": 0.72,
+      "reward": 2.669346511363983,
+      "reward_std": 0.410087987780571,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.9791667014360428,
+      "rewards/repetition_penalty_reward": -0.02857015887275338,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 282
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 224.3854217529297,
-      "epoch": 0.848575712143928,
-      "grad_norm": 85.4845823002127,
-      "kl": 5.06640625,
-      "learning_rate": 1.3485150687433168e-06,
-      "loss": 2.0253,
-      "reward": 2.631519079208374,
-      "reward_std": 0.8587917536497116,
-      "rewards/accuracy_reward": 0.8593750149011612,
-      "rewards/reasoning_steps_reward": 0.9062499850988388,
-      "rewards/repetition_penalty_reward": -0.0038975586649030447,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 195.53125,
+      "epoch": 0.42444694413198353,
+      "grad_norm": 3.0528671937109757,
+      "kl": 0.5556640625,
+      "learning_rate": 1.4240701274874331e-05,
+      "loss": 0.1773,
+      "reward": 2.6338695287704468,
+      "reward_std": 0.2932188091799617,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.03236319124698639,
+      "rewards/tag_count_reward": 0.9908854216337204,
       "step": 283
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 234.18230056762695,
-      "epoch": 0.8515742128935532,
-      "grad_norm": 74.49346765573262,
-      "kl": 10.109375,
-      "learning_rate": 1.2962993250364541e-06,
-      "loss": 2.5552,
-      "reward": 2.4909926652908325,
-      "reward_std": 0.8654274046421051,
-      "rewards/accuracy_reward": 0.739583358168602,
-      "rewards/reasoning_steps_reward": 0.8906250447034836,
-      "rewards/repetition_penalty_reward": -0.009007335815113038,
-      "rewards/tag_count_reward": 0.8697916716337204,
+      "completion_length": 196.04687881469727,
+      "epoch": 0.4259467566554181,
+      "grad_norm": 16.521107944862717,
+      "kl": 1.06494140625,
+      "learning_rate": 1.419314536278925e-05,
+      "loss": 0.6407,
+      "reward": 2.675931394100189,
+      "reward_std": 0.3121615252457559,
+      "rewards/accuracy_reward": 0.755208358168602,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.030665938276797533,
+      "rewards/tag_count_reward": 0.9687500149011612,
       "step": 284
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 238.6614646911621,
-      "epoch": 0.8545727136431784,
-      "grad_norm": 172.140865328852,
-      "kl": 10.84375,
-      "learning_rate": 1.2450444351479196e-06,
-      "loss": 2.3458,
-      "reward": 2.4398276805877686,
-      "reward_std": 0.7756916880607605,
-      "rewards/accuracy_reward": 0.6614583507180214,
-      "rewards/reasoning_steps_reward": 0.9114583432674408,
-      "rewards/repetition_penalty_reward": -0.006786908605135977,
-      "rewards/tag_count_reward": 0.8736979365348816,
+      "completion_length": 204.18750381469727,
+      "epoch": 0.42744656917885265,
+      "grad_norm": 5.497622639890418,
+      "kl": 0.99462890625,
+      "learning_rate": 1.4145474109356008e-05,
+      "loss": 0.4354,
+      "reward": 2.717790961265564,
+      "reward_std": 0.28716015443205833,
+      "rewards/accuracy_reward": 0.7760416865348816,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.02656660182401538,
+      "rewards/tag_count_reward": 0.9804687649011612,
       "step": 285
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 184.40625762939453,
-      "epoch": 0.8575712143928036,
-      "grad_norm": 33.813378116083186,
-      "kl": 4.0390625,
-      "learning_rate": 1.1947560574130013e-06,
-      "loss": 1.2931,
-      "reward": 2.638872265815735,
-      "reward_std": 0.5787665322422981,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 0.9392361640930176,
-      "rewards/repetition_penalty_reward": -0.006093142030294985,
-      "rewards/tag_count_reward": 0.9088541865348816,
+      "completion_length": 246.9531364440918,
+      "epoch": 0.4289463817022872,
+      "grad_norm": 11.12837490424536,
+      "kl": 3.10546875,
+      "learning_rate": 1.4097688825873437e-05,
+      "loss": 0.8672,
+      "reward": 2.6280420422554016,
+      "reward_std": 0.39313384145498276,
+      "rewards/accuracy_reward": 0.7343750149011612,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.03905859449878335,
+      "rewards/tag_count_reward": 0.9570312798023224,
       "step": 286
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 267.8020935058594,
-      "epoch": 0.8605697151424287,
-      "grad_norm": 29.38596871781673,
-      "kl": 7.890625,
-      "learning_rate": 1.1454397434679022e-06,
-      "loss": 2.2911,
-      "reward": 2.4633301496505737,
-      "reward_std": 1.00086210668087,
-      "rewards/accuracy_reward": 0.755208358168602,
-      "rewards/reasoning_steps_reward": 0.8715277910232544,
-      "rewards/repetition_penalty_reward": -0.0019476997549645603,
-      "rewards/tag_count_reward": 0.8385416865348816,
+      "completion_length": 230.0572967529297,
+      "epoch": 0.4304461942257218,
+      "grad_norm": 9.306320853861074,
+      "kl": 2.4326171875,
+      "learning_rate": 1.4049790826777016e-05,
+      "loss": 0.8369,
+      "reward": 2.5040841102600098,
+      "reward_std": 0.3208254538476467,
+      "rewards/accuracy_reward": 0.5781250074505806,
+      "rewards/reasoning_steps_reward": 0.9965277910232544,
+      "rewards/repetition_penalty_reward": -0.031506254337728024,
+      "rewards/tag_count_reward": 0.9609375149011612,
       "step": 287
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 195.26562881469727,
-      "epoch": 0.863568215892054,
-      "grad_norm": 22.60777483816971,
-      "kl": 5.671875,
-      "learning_rate": 1.0971009376368614e-06,
-      "loss": 1.9524,
-      "reward": 2.4915042221546173,
-      "reward_std": 0.6768720299005508,
-      "rewards/accuracy_reward": 0.6718750149011612,
-      "rewards/reasoning_steps_reward": 0.923611119389534,
-      "rewards/repetition_penalty_reward": -0.00632570160087198,
-      "rewards/tag_count_reward": 0.9023437798023224,
+      "completion_length": 217.78125381469727,
+      "epoch": 0.43194600674915634,
+      "grad_norm": 5.585862488308667,
+      "kl": 1.44921875,
+      "learning_rate": 1.4001781429602704e-05,
+      "loss": 0.6499,
+      "reward": 2.556925117969513,
+      "reward_std": 0.42149753123521805,
+      "rewards/accuracy_reward": 0.6093750260770321,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.019897868391126394,
+      "rewards/tag_count_reward": 0.97265625,
       "step": 288
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 230.11458587646484,
-      "epoch": 0.8665667166416792,
-      "grad_norm": 203.49207627713312,
-      "kl": 12.6796875,
-      "learning_rate": 1.049744976331124e-06,
-      "loss": 3.1976,
-      "reward": 2.5564926862716675,
-      "reward_std": 0.9073340892791748,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 0.8958333432674408,
-      "rewards/repetition_penalty_reward": -0.0047051976434886456,
-      "rewards/tag_count_reward": 0.868489608168602,
+      "completion_length": 207.7552146911621,
+      "epoch": 0.4334458192725909,
+      "grad_norm": 5.688055347158159,
+      "kl": 1.3486328125,
+      "learning_rate": 1.3953661954950693e-05,
+      "loss": 0.5048,
+      "reward": 2.4292953610420227,
+      "reward_std": 0.3075702078640461,
+      "rewards/accuracy_reward": 0.5000000102445483,
+      "rewards/reasoning_steps_reward": 0.9913194477558136,
+      "rewards/repetition_penalty_reward": -0.03207629453390837,
+      "rewards/tag_count_reward": 0.970052108168602,
       "step": 289
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 198.89062881469727,
-      "epoch": 0.8695652173913043,
-      "grad_norm": 124.52484685260382,
-      "kl": 9.375,
-      "learning_rate": 1.0033770874598226e-06,
-      "loss": 2.8325,
-      "reward": 2.5556147694587708,
-      "reward_std": 0.8008880615234375,
-      "rewards/accuracy_reward": 0.7447916865348816,
-      "rewards/reasoning_steps_reward": 0.9201388955116272,
-      "rewards/repetition_penalty_reward": -0.003847024345304817,
-      "rewards/tag_count_reward": 0.8945312798023224,
+      "completion_length": 187.9166717529297,
+      "epoch": 0.4349456317960255,
+      "grad_norm": 1.8549641962399313,
+      "kl": 0.71044921875,
+      "learning_rate": 1.3905433726449102e-05,
+      "loss": 0.251,
+      "reward": 2.6722521781921387,
+      "reward_std": 0.3378063030540943,
+      "rewards/accuracy_reward": 0.7239583432674408,
+      "rewards/reasoning_steps_reward": 0.9965277910232544,
+      "rewards/repetition_penalty_reward": -0.03651529923081398,
+      "rewards/tag_count_reward": 0.9882812649011612,
       "step": 290
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 214.3177146911621,
-      "epoch": 0.8725637181409296,
-      "grad_norm": 58.07106802782898,
-      "kl": 4.95703125,
-      "learning_rate": 9.580023898528346e-07,
-      "loss": 2.0119,
-      "reward": 2.5350549817085266,
-      "reward_std": 0.7193208187818527,
-      "rewards/accuracy_reward": 0.729166679084301,
-      "rewards/reasoning_steps_reward": 0.923611119389534,
-      "rewards/repetition_penalty_reward": -0.012254099652636796,
-      "rewards/tag_count_reward": 0.8945312649011612,
+      "completion_length": 193.82813262939453,
+      "epoch": 0.4364454443194601,
+      "grad_norm": 2.2180974772052453,
+      "kl": 1.0986328125,
+      "learning_rate": 1.3857098070717543e-05,
+      "loss": 0.5164,
+      "reward": 2.7617053389549255,
+      "reward_std": 0.3312147632241249,
+      "rewards/accuracy_reward": 0.8177083432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.022582892794162035,
+      "rewards/tag_count_reward": 0.9804687649011612,
       "step": 291
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 212.4583396911621,
-      "epoch": 0.8755622188905547,
-      "grad_norm": 56.388506410538476,
-      "kl": 5.796875,
-      "learning_rate": 9.136258926956887e-07,
-      "loss": 1.9009,
-      "reward": 2.553084135055542,
-      "reward_std": 0.7310795336961746,
-      "rewards/accuracy_reward": 0.7031250149011612,
-      "rewards/reasoning_steps_reward": 0.9531250298023224,
-      "rewards/repetition_penalty_reward": -0.006811692088376731,
-      "rewards/tag_count_reward": 0.9036458432674408,
+      "completion_length": 235.54688262939453,
+      "epoch": 0.43794525684289465,
+      "grad_norm": 12.18787699766214,
+      "kl": 2.185546875,
+      "learning_rate": 1.3808656317330646e-05,
+      "loss": 0.7891,
+      "reward": 2.4147286415100098,
+      "reward_std": 0.5803365781903267,
+      "rewards/accuracy_reward": 0.5572916828095913,
+      "rewards/reasoning_steps_reward": 0.9565972685813904,
+      "rewards/repetition_penalty_reward": -0.05619149189442396,
+      "rewards/tag_count_reward": 0.9570312649011612,
       "step": 292
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 241.4270896911621,
-      "epoch": 0.8785607196401799,
-      "grad_norm": 28.637264291998722,
-      "kl": 9.59375,
-      "learning_rate": 8.702524949765645e-07,
-      "loss": 2.5645,
-      "reward": 2.4492486715316772,
-      "reward_std": 0.9317169636487961,
-      "rewards/accuracy_reward": 0.7135416865348816,
-      "rewards/reasoning_steps_reward": 0.8854167014360428,
-      "rewards/repetition_penalty_reward": -0.003876364848110825,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 226.35937881469727,
+      "epoch": 0.4394450693663292,
+      "grad_norm": 14.298340510742703,
+      "kl": 1.6005859375,
+      "learning_rate": 1.3760109798781489e-05,
+      "loss": 0.6926,
+      "reward": 2.744252860546112,
+      "reward_std": 0.3814915865659714,
+      "rewards/accuracy_reward": 0.8229166716337204,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.053924234583973885,
+      "rewards/tag_count_reward": 0.9752604365348816,
       "step": 293
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 176.4166717529297,
-      "epoch": 0.881559220389805,
-      "grad_norm": 62.774529626199765,
-      "kl": 5.64453125,
-      "learning_rate": 8.278869849454718e-07,
-      "loss": 2.3349,
-      "reward": 2.748142123222351,
-      "reward_std": 0.7123552411794662,
-      "rewards/accuracy_reward": 0.9010416865348816,
-      "rewards/reasoning_steps_reward": 0.9340278208255768,
-      "rewards/repetition_penalty_reward": -0.003594138892367482,
-      "rewards/tag_count_reward": 0.9166666865348816,
+      "completion_length": 206.0677146911621,
+      "epoch": 0.4409448818897638,
+      "grad_norm": 1.5380278951289539,
+      "kl": 0.818359375,
+      "learning_rate": 1.3711459850444923e-05,
+      "loss": 0.4797,
+      "reward": 2.565054178237915,
+      "reward_std": 0.43462975323200226,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.982638880610466,
+      "rewards/repetition_penalty_reward": -0.047793143428862095,
+      "rewards/tag_count_reward": 0.9739583432674408,
       "step": 294
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 257.51042556762695,
-      "epoch": 0.8845577211394303,
-      "grad_norm": 128.72898033686099,
-      "kl": 11.6328125,
-      "learning_rate": 7.865340395856325e-07,
-      "loss": 2.9018,
-      "reward": 2.4993616342544556,
-      "reward_std": 0.9201382249593735,
-      "rewards/accuracy_reward": 0.7656250149011612,
-      "rewards/reasoning_steps_reward": 0.8836805820465088,
-      "rewards/repetition_penalty_reward": -0.0041107177385129035,
-      "rewards/tag_count_reward": 0.8541666865348816,
+      "completion_length": 216.1822967529297,
+      "epoch": 0.44244469441319834,
+      "grad_norm": 2.6940921154273263,
+      "kl": 0.70263671875,
+      "learning_rate": 1.3662707810540867e-05,
+      "loss": 0.4852,
+      "reward": 2.7508795261383057,
+      "reward_std": 0.4059586077928543,
+      "rewards/accuracy_reward": 0.8281250298023224,
+      "rewards/reasoning_steps_reward": 0.9965277761220932,
+      "rewards/repetition_penalty_reward": -0.04642965644598007,
+      "rewards/tag_count_reward": 0.9726562649011612,
       "step": 295
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 208.9895896911621,
-      "epoch": 0.8875562218890555,
-      "grad_norm": 19.58108496310109,
-      "kl": 6.859375,
-      "learning_rate": 7.461982240971799e-07,
-      "loss": 2.4699,
-      "reward": 2.625687062740326,
-      "reward_std": 0.8259201794862747,
-      "rewards/accuracy_reward": 0.8072916865348816,
-      "rewards/reasoning_steps_reward": 0.9322916716337204,
-      "rewards/repetition_penalty_reward": -0.0032192860962823033,
-      "rewards/tag_count_reward": 0.8893229216337204,
+      "completion_length": 205.18230056762695,
+      "epoch": 0.4439445069366329,
+      "grad_norm": 3.103564524085595,
+      "kl": 0.60400390625,
+      "learning_rate": 1.3613855020097477e-05,
+      "loss": 0.4706,
+      "reward": 2.7546513080596924,
+      "reward_std": 0.3886885233223438,
+      "rewards/accuracy_reward": 0.8437500149011612,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.05133842211216688,
+      "rewards/tag_count_reward": 0.9726562649011612,
       "step": 296
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 212.17708587646484,
-      "epoch": 0.8905547226386806,
-      "grad_norm": 33.259888962364116,
-      "kl": 6.0,
-      "learning_rate": 7.068839913931646e-07,
-      "loss": 2.0135,
-      "reward": 2.681620717048645,
-      "reward_std": 0.7299316972494125,
-      "rewards/accuracy_reward": 0.8645833432674408,
-      "rewards/reasoning_steps_reward": 0.9322916865348816,
-      "rewards/repetition_penalty_reward": -0.004577249084832147,
-      "rewards/tag_count_reward": 0.8893229365348816,
+      "completion_length": 208.4427146911621,
+      "epoch": 0.4454443194600675,
+      "grad_norm": 2.4514035939782155,
+      "kl": 0.61279296875,
+      "learning_rate": 1.3564902822914274e-05,
+      "loss": 0.4279,
+      "reward": 2.7078776359558105,
+      "reward_std": 0.3973502516746521,
+      "rewards/accuracy_reward": 0.7968750149011612,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.05340719223022461,
+      "rewards/tag_count_reward": 0.9817708432674408,
       "step": 297
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 241.8229217529297,
-      "epoch": 0.8935532233883059,
-      "grad_norm": 47.82236374571516,
-      "kl": 7.6875,
-      "learning_rate": 6.685956816079753e-07,
-      "loss": 2.4039,
-      "reward": 2.5991902351379395,
-      "reward_std": 0.863490641117096,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9062500149011612,
-      "rewards/repetition_penalty_reward": -0.004976554249878973,
-      "rewards/tag_count_reward": 0.8645833730697632,
+      "completion_length": 203.4322967529297,
+      "epoch": 0.4469441319835021,
+      "grad_norm": 16.752659312125665,
+      "kl": 0.775390625,
+      "learning_rate": 1.3515852565525167e-05,
+      "loss": 0.3563,
+      "reward": 2.8165610432624817,
+      "reward_std": 0.28164857253432274,
+      "rewards/accuracy_reward": 0.880208358168602,
+      "rewards/reasoning_steps_reward": 0.9965277910232544,
+      "rewards/repetition_penalty_reward": -0.05106059880927205,
+      "rewards/tag_count_reward": 0.9908854365348816,
       "step": 298
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 199.1666717529297,
-      "epoch": 0.896551724137931,
-      "grad_norm": 21.908160424151905,
-      "kl": 6.2890625,
-      "learning_rate": 6.313375216182039e-07,
-      "loss": 2.1573,
-      "reward": 2.705993890762329,
-      "reward_std": 0.7558012455701828,
-      "rewards/accuracy_reward": 0.8750000149011612,
-      "rewards/reasoning_steps_reward": 0.9288194626569748,
-      "rewards/repetition_penalty_reward": -0.00277355604339391,
-      "rewards/tag_count_reward": 0.9049479365348816,
+      "completion_length": 196.20833587646484,
+      "epoch": 0.44844394450693664,
+      "grad_norm": 29.22658507606216,
+      "kl": 0.84423828125,
+      "learning_rate": 1.3466705597161416e-05,
+      "loss": 0.4067,
+      "reward": 2.764431357383728,
+      "reward_std": 0.3144514746963978,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9809028208255768,
+      "rewards/repetition_penalty_reward": -0.05240898672491312,
+      "rewards/tag_count_reward": 0.9817708432674408,
       "step": 299
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 204.30730056762695,
-      "epoch": 0.8995502248875562,
-      "grad_norm": 122.95452439926706,
-      "kl": 8.9375,
-      "learning_rate": 5.951136245760181e-07,
-      "loss": 2.666,
-      "reward": 2.5854055881500244,
-      "reward_std": 0.6643490642309189,
-      "rewards/accuracy_reward": 0.7291666865348816,
-      "rewards/reasoning_steps_reward": 0.9496528059244156,
-      "rewards/repetition_penalty_reward": -0.0061743835103698075,
-      "rewards/tag_count_reward": 0.9127604365348816,
+      "completion_length": 188.73437881469727,
+      "epoch": 0.4499437570303712,
+      "grad_norm": 0.9955419431414791,
+      "kl": 0.42578125,
+      "learning_rate": 1.3417463269714525e-05,
+      "loss": 0.2115,
+      "reward": 2.706632077693939,
+      "reward_std": 0.2604191384743899,
+      "rewards/accuracy_reward": 0.7656250149011612,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.03555556107312441,
+      "rewards/tag_count_reward": 0.9869791865348816,
       "step": 300
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 276.95314025878906,
-      "epoch": 0.9025487256371814,
-      "grad_norm": 295.5147824374305,
-      "kl": 15.828125,
-      "learning_rate": 5.599279894550824e-07,
-      "loss": 3.1455,
-      "reward": 2.4131381511688232,
-      "reward_std": 0.8852957636117935,
-      "rewards/accuracy_reward": 0.7031250149011612,
-      "rewards/reasoning_steps_reward": 0.8819444477558136,
-      "rewards/repetition_penalty_reward": -0.003962728020269424,
-      "rewards/tag_count_reward": 0.8320312649011612,
+      "completion_length": 212.2395896911621,
+      "epoch": 0.45144356955380577,
+      "grad_norm": 1.461352147416173,
+      "kl": 0.5546875,
+      "learning_rate": 1.3368126937699055e-05,
+      "loss": 0.4384,
+      "reward": 2.648770272731781,
+      "reward_std": 0.4800826385617256,
+      "rewards/accuracy_reward": 0.760416679084301,
+      "rewards/reasoning_steps_reward": 0.9739583432674408,
+      "rewards/repetition_penalty_reward": -0.06346932239830494,
+      "rewards/tag_count_reward": 0.9778645932674408,
       "step": 301
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 212.93750762939453,
-      "epoch": 0.9055472263868066,
-      "grad_norm": 70.53830308263373,
-      "kl": 8.5390625,
-      "learning_rate": 5.257845006090911e-07,
-      "loss": 2.4792,
-      "reward": 2.5973342061042786,
-      "reward_std": 0.7405329048633575,
-      "rewards/accuracy_reward": 0.786458358168602,
-      "rewards/reasoning_steps_reward": 0.9322916865348816,
-      "rewards/repetition_penalty_reward": -0.005530498514417559,
-      "rewards/tag_count_reward": 0.884114608168602,
+      "completion_length": 228.32813262939453,
+      "epoch": 0.45294338207724033,
+      "grad_norm": 1.0886524612409492,
+      "kl": 0.60986328125,
+      "learning_rate": 1.3318697958215358e-05,
+      "loss": 0.4636,
+      "reward": 2.561100661754608,
+      "reward_std": 0.5553513169288635,
+      "rewards/accuracy_reward": 0.6979166865348816,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.06823972798883915,
+      "rewards/tag_count_reward": 0.9609375149011612,
       "step": 302
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 257.29167556762695,
-      "epoch": 0.9085457271364318,
-      "grad_norm": 22.013195631274943,
-      "kl": 6.3125,
-      "learning_rate": 4.926869273429447e-07,
-      "loss": 1.8309,
-      "reward": 2.465520918369293,
-      "reward_std": 0.7833161950111389,
-      "rewards/accuracy_reward": 0.7135416865348816,
-      "rewards/reasoning_steps_reward": 0.9062500298023224,
-      "rewards/repetition_penalty_reward": -0.009739560075104237,
-      "rewards/tag_count_reward": 0.8554687798023224,
+      "completion_length": 191.9166717529297,
+      "epoch": 0.4544431946006749,
+      "grad_norm": 4.083408704872753,
+      "kl": 0.5341796875,
+      "learning_rate": 1.3269177690912244e-05,
+      "loss": 0.4274,
+      "reward": 2.4237236976623535,
+      "reward_std": 0.3847580924630165,
+      "rewards/accuracy_reward": 0.5000000149011612,
+      "rewards/reasoning_steps_reward": 0.9826389253139496,
+      "rewards/repetition_penalty_reward": -0.040686123073101044,
+      "rewards/tag_count_reward": 0.9817708432674408,
       "step": 303
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 252.58854293823242,
-      "epoch": 0.9115442278860569,
-      "grad_norm": 88.95393283353653,
-      "kl": 5.8515625,
-      "learning_rate": 4.606389234966424e-07,
-      "loss": 1.9041,
-      "reward": 2.496855139732361,
-      "reward_std": 0.8814668357372284,
-      "rewards/accuracy_reward": 0.7708333432674408,
-      "rewards/reasoning_steps_reward": 0.88368059694767,
-      "rewards/repetition_penalty_reward": -0.00531507984851487,
-      "rewards/tag_count_reward": 0.8476562649011612,
+      "completion_length": 180.50000381469727,
+      "epoch": 0.4559430071241095,
+      "grad_norm": 5.326097299416955,
+      "kl": 0.65478515625,
+      "learning_rate": 1.3219567497949603e-05,
+      "loss": 0.2812,
+      "reward": 2.7030047178268433,
+      "reward_std": 0.3447668179869652,
+      "rewards/accuracy_reward": 0.7760416865348816,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.04786340892314911,
+      "rewards/tag_count_reward": 0.9869791865348816,
       "step": 304
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 211.03125381469727,
-      "epoch": 0.9145427286356822,
-      "grad_norm": 152.53165269997837,
-      "kl": 4.96484375,
-      "learning_rate": 4.2964402704190555e-07,
-      "loss": 2.153,
-      "reward": 2.586493730545044,
-      "reward_std": 0.8061617463827133,
-      "rewards/accuracy_reward": 0.786458358168602,
-      "rewards/reasoning_steps_reward": 0.920138880610466,
-      "rewards/repetition_penalty_reward": -0.0042182166944257915,
-      "rewards/tag_count_reward": 0.8841145932674408,
+      "completion_length": 197.6145896911621,
+      "epoch": 0.4574428196475441,
+      "grad_norm": 3.907793025870675,
+      "kl": 0.7373046875,
+      "learning_rate": 1.3169868743960904e-05,
+      "loss": 0.6404,
+      "reward": 2.686201572418213,
+      "reward_std": 0.4964783936738968,
+      "rewards/accuracy_reward": 0.802083358168602,
+      "rewards/reasoning_steps_reward": 0.970486119389534,
+      "rewards/repetition_penalty_reward": -0.05251379404217005,
+      "rewards/tag_count_reward": 0.9661458432674408,
       "step": 305
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 216.83333587646484,
-      "epoch": 0.9175412293853074,
-      "grad_norm": 88.75542795978963,
-      "kl": 6.0703125,
-      "learning_rate": 3.997056596916038e-07,
-      "loss": 2.1542,
-      "reward": 2.656059443950653,
-      "reward_std": 0.79727503657341,
-      "rewards/accuracy_reward": 0.8645833432674408,
-      "rewards/reasoning_steps_reward": 0.91493059694767,
-      "rewards/repetition_penalty_reward": -0.0036628567904699594,
-      "rewards/tag_count_reward": 0.8802083432674408,
+      "completion_length": 227.15625381469727,
+      "epoch": 0.45894263217097864,
+      "grad_norm": 1.3757226995738794,
+      "kl": 0.8720703125,
+      "learning_rate": 1.3120082796015694e-05,
+      "loss": 0.5179,
+      "reward": 2.551712214946747,
+      "reward_std": 0.4617387279868126,
+      "rewards/accuracy_reward": 0.7031250223517418,
+      "rewards/reasoning_steps_reward": 0.963541716337204,
+      "rewards/repetition_penalty_reward": -0.06938168965280056,
+      "rewards/tag_count_reward": 0.954427108168602,
       "step": 306
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 208.17188262939453,
-      "epoch": 0.9205397301349325,
-      "grad_norm": 75.82177154852495,
-      "kl": 8.06640625,
-      "learning_rate": 3.708271265220087e-07,
-      "loss": 2.1738,
-      "reward": 2.6760451197624207,
-      "reward_std": 0.7257892712950706,
-      "rewards/accuracy_reward": 0.880208358168602,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.006246598088182509,
-      "rewards/tag_count_reward": 0.8854166865348816,
+      "completion_length": 191.1302146911621,
+      "epoch": 0.4604424446944132,
+      "grad_norm": 2.380244199934681,
+      "kl": 0.49951171875,
+      "learning_rate": 1.3070211023581959e-05,
+      "loss": 0.3394,
+      "reward": 2.5620276927948,
+      "reward_std": 0.2889845799654722,
+      "rewards/accuracy_reward": 0.635416679084301,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.047347418032586575,
+      "rewards/tag_count_reward": 0.9843750149011612,
       "step": 307
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 226.62500381469727,
-      "epoch": 0.9235382308845578,
-      "grad_norm": 167.48098467184784,
-      "kl": 10.15625,
-      "learning_rate": 3.430116156079277e-07,
-      "loss": 2.5838,
-      "reward": 2.539145290851593,
-      "reward_std": 0.8195622712373734,
-      "rewards/accuracy_reward": 0.7552083432674408,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.006427670712582767,
-      "rewards/tag_count_reward": 0.8736979216337204,
+      "completion_length": 195.4479217529297,
+      "epoch": 0.46194225721784776,
+      "grad_norm": 3.432394493893828,
+      "kl": 0.86669921875,
+      "learning_rate": 1.302025479848847e-05,
+      "loss": 0.6401,
+      "reward": 2.608788251876831,
+      "reward_std": 0.3985390290617943,
+      "rewards/accuracy_reward": 0.7135416865348816,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.05050008138641715,
+      "rewards/tag_count_reward": 0.970052108168602,
       "step": 308
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 250.46875381469727,
-      "epoch": 0.9265367316341829,
-      "grad_norm": 188.4999317880181,
-      "kl": 12.703125,
-      "learning_rate": 3.1626219767075584e-07,
-      "loss": 2.8945,
-      "reward": 2.4836193919181824,
-      "reward_std": 0.8600171506404877,
-      "rewards/accuracy_reward": 0.739583358168602,
-      "rewards/reasoning_steps_reward": 0.9079861342906952,
-      "rewards/repetition_penalty_reward": -0.01551259565167129,
-      "rewards/tag_count_reward": 0.8515625149011612,
+      "completion_length": 182.7604217529297,
+      "epoch": 0.4634420697412823,
+      "grad_norm": 1.720744202323982,
+      "kl": 0.62353515625,
+      "learning_rate": 1.2970215494887057e-05,
+      "loss": 0.496,
+      "reward": 2.7085583209991455,
+      "reward_std": 0.3612633068114519,
+      "rewards/accuracy_reward": 0.7812500298023224,
+      "rewards/reasoning_steps_reward": 0.989583358168602,
+      "rewards/repetition_penalty_reward": -0.04144172929227352,
+      "rewards/tag_count_reward": 0.9791666865348816,
       "step": 309
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 245.3020896911621,
-      "epoch": 0.9295352323838081,
-      "grad_norm": 61.9794266708475,
-      "kl": 9.3046875,
-      "learning_rate": 2.905818257394799e-07,
-      "loss": 2.2241,
-      "reward": 2.4837332367897034,
-      "reward_std": 0.7677433341741562,
-      "rewards/accuracy_reward": 0.7291666716337204,
-      "rewards/reasoning_steps_reward": 0.9079861491918564,
-      "rewards/repetition_penalty_reward": -0.004982162558007985,
-      "rewards/tag_count_reward": 0.8515625149011612,
+      "completion_length": 182.78646087646484,
+      "epoch": 0.4649418822647169,
+      "grad_norm": 3.1536070169922437,
+      "kl": 0.6181640625,
+      "learning_rate": 1.2920094489214794e-05,
+      "loss": 0.4513,
+      "reward": 2.6318033933639526,
+      "reward_std": 0.3417645953595638,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.049186245538294315,
+      "rewards/tag_count_reward": 0.9830729365348816,
       "step": 310
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 223.09375762939453,
-      "epoch": 0.9325337331334332,
-      "grad_norm": 38.9834806127014,
-      "kl": 6.96875,
-      "learning_rate": 2.659733348246685e-07,
-      "loss": 2.2927,
-      "reward": 2.5947685837745667,
-      "reward_std": 0.8158188462257385,
-      "rewards/accuracy_reward": 0.8072916716337204,
-      "rewards/reasoning_steps_reward": 0.9149305671453476,
-      "rewards/repetition_penalty_reward": -0.005057800153736025,
-      "rewards/tag_count_reward": 0.8776041716337204,
+      "completion_length": 172.5208396911621,
+      "epoch": 0.4664416947881515,
+      "grad_norm": 2.2535506878920915,
+      "kl": 0.5439453125,
+      "learning_rate": 1.2869893160156144e-05,
+      "loss": 0.2928,
+      "reward": 2.554309666156769,
+      "reward_std": 0.3680322840809822,
+      "rewards/accuracy_reward": 0.635416679084301,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.042912650387734175,
+      "rewards/tag_count_reward": 0.9791666716337204,
       "step": 311
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 237.89063262939453,
-      "epoch": 0.9355322338830585,
-      "grad_norm": 93.39130786178272,
-      "kl": 5.546875,
-      "learning_rate": 2.4243944160550757e-07,
-      "loss": 1.9383,
-      "reward": 2.5787097811698914,
-      "reward_std": 0.834644541144371,
-      "rewards/accuracy_reward": 0.817708358168602,
-      "rewards/reasoning_steps_reward": 0.9010417014360428,
-      "rewards/repetition_penalty_reward": -0.007227733498439193,
-      "rewards/tag_count_reward": 0.8671875149011612,
+      "completion_length": 169.9322967529297,
+      "epoch": 0.46794150731158607,
+      "grad_norm": 1.1774122557707811,
+      "kl": 0.40576171875,
+      "learning_rate": 1.2819612888605038e-05,
+      "loss": 0.3948,
+      "reward": 2.694904148578644,
+      "reward_std": 0.34949270635843277,
+      "rewards/accuracy_reward": 0.7552083432674408,
+      "rewards/reasoning_steps_reward": 0.989583358168602,
+      "rewards/repetition_penalty_reward": -0.03426253283396363,
+      "rewards/tag_count_reward": 0.9843750149011612,
       "step": 312
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 220.5572967529297,
-      "epoch": 0.9385307346326837,
-      "grad_norm": 75.42828643371548,
-      "kl": 5.5546875,
-      "learning_rate": 2.199827441298863e-07,
-      "loss": 1.8684,
-      "reward": 2.61844664812088,
-      "reward_std": 0.8282945156097412,
-      "rewards/accuracy_reward": 0.8437500149011612,
-      "rewards/reasoning_steps_reward": 0.9062500298023224,
-      "rewards/repetition_penalty_reward": -0.006553410203196108,
-      "rewards/tag_count_reward": 0.8750000149011612,
+      "completion_length": 163.50000381469727,
+      "epoch": 0.46944131983502063,
+      "grad_norm": 0.6398131246264415,
+      "kl": 0.33056640625,
+      "learning_rate": 1.2769255057626879e-05,
+      "loss": 0.1586,
+      "reward": 2.7207802534103394,
+      "reward_std": 0.12394981307443231,
+      "rewards/accuracy_reward": 0.7552083432674408,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.0240114142652601,
+      "rewards/tag_count_reward": 0.9947916865348816,
       "step": 313
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 254.90625762939453,
-      "epoch": 0.9415292353823088,
-      "grad_norm": 55.21000753237673,
-      "kl": 6.890625,
-      "learning_rate": 1.986057215275816e-07,
-      "loss": 2.2089,
-      "reward": 2.4655805230140686,
-      "reward_std": 0.8631928116083145,
-      "rewards/accuracy_reward": 0.7135416865348816,
-      "rewards/reasoning_steps_reward": 0.9010416716337204,
-      "rewards/repetition_penalty_reward": -0.004471502033993602,
-      "rewards/tag_count_reward": 0.8554687649011612,
+      "completion_length": 162.06250762939453,
+      "epoch": 0.4709411323584552,
+      "grad_norm": 1.0331986505694868,
+      "kl": 0.3564453125,
+      "learning_rate": 1.2718821052420518e-05,
+      "loss": 0.3133,
+      "reward": 2.7166225910186768,
+      "reward_std": 0.23217968584503978,
+      "rewards/accuracy_reward": 0.7656250298023224,
+      "rewards/reasoning_steps_reward": 0.9913194626569748,
+      "rewards/repetition_penalty_reward": -0.029905199247878045,
+      "rewards/tag_count_reward": 0.9895833432674408,
       "step": 314
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 199.2916717529297,
-      "epoch": 0.9445277361319341,
-      "grad_norm": 13.351484586247325,
-      "kl": 4.96875,
-      "learning_rate": 1.7831073373657527e-07,
-      "loss": 1.2923,
-      "reward": 2.658263087272644,
-      "reward_std": 0.573756992816925,
-      "rewards/accuracy_reward": 0.8333333432674408,
-      "rewards/reasoning_steps_reward": 0.9236111491918564,
-      "rewards/repetition_penalty_reward": -0.00753560135490261,
-      "rewards/tag_count_reward": 0.9088541716337204,
+      "completion_length": 159.98438262939453,
+      "epoch": 0.47244094488188976,
+      "grad_norm": 0.8428674192959495,
+      "kl": 0.3876953125,
+      "learning_rate": 1.2668312260280136e-05,
+      "loss": 0.1753,
+      "reward": 2.8209707736968994,
+      "reward_std": 0.2351000364869833,
+      "rewards/accuracy_reward": 0.8750000149011612,
+      "rewards/reasoning_steps_reward": 0.9843750298023224,
+      "rewards/repetition_penalty_reward": -0.025383584201335907,
+      "rewards/tag_count_reward": 0.9869791865348816,
       "step": 315
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 211.89583587646484,
-      "epoch": 0.9475262368815592,
-      "grad_norm": 14.409679464629958,
-      "kl": 7.03125,
-      "learning_rate": 1.5910002124251979e-07,
-      "loss": 2.3063,
-      "reward": 2.6654553413391113,
-      "reward_std": 0.7150935083627701,
-      "rewards/accuracy_reward": 0.8281250298023224,
-      "rewards/reasoning_steps_reward": 0.9444444924592972,
-      "rewards/repetition_penalty_reward": -0.002947573288111016,
-      "rewards/tag_count_reward": 0.895833358168602,
+      "completion_length": 181.640625,
+      "epoch": 0.4739407574053243,
+      "grad_norm": 1.5853800434552652,
+      "kl": 0.482421875,
+      "learning_rate": 1.2617730070557079e-05,
+      "loss": 0.6332,
+      "reward": 2.6595771312713623,
+      "reward_std": 0.5877653062343597,
+      "rewards/accuracy_reward": 0.770833358168602,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.0517944535240531,
+      "rewards/tag_count_reward": 0.9648437649011612,
       "step": 316
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 269.00521087646484,
-      "epoch": 0.9505247376311844,
-      "grad_norm": 157.41546141813902,
-      "kl": 12.15625,
-      "learning_rate": 1.4097570483140642e-07,
-      "loss": 2.5569,
-      "reward": 2.467340648174286,
-      "reward_std": 0.888208270072937,
-      "rewards/accuracy_reward": 0.7239583432674408,
-      "rewards/reasoning_steps_reward": 0.9097222685813904,
-      "rewards/repetition_penalty_reward": -0.00878784217638895,
-      "rewards/tag_count_reward": 0.8424479365348816,
+      "completion_length": 196.3593864440918,
+      "epoch": 0.4754405699287589,
+      "grad_norm": 0.6434608924686329,
+      "kl": 0.37841796875,
+      "learning_rate": 1.2567075874621658e-05,
+      "loss": 0.3751,
+      "reward": 2.7092297673225403,
+      "reward_std": 0.3774634450674057,
+      "rewards/accuracy_reward": 0.817708358168602,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.05769744282588363,
+      "rewards/tag_count_reward": 0.9700520932674408,
       "step": 317
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 203.60938262939453,
-      "epoch": 0.9535232383808095,
-      "grad_norm": 141.14810050857622,
-      "kl": 9.65625,
-      "learning_rate": 1.239397853554336e-07,
-      "loss": 3.0646,
-      "reward": 2.610135555267334,
-      "reward_std": 0.7995207458734512,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 0.9253472685813904,
-      "rewards/repetition_penalty_reward": -0.005315949267242104,
-      "rewards/tag_count_reward": 0.8932291716337204,
+      "completion_length": 250.70313262939453,
+      "epoch": 0.4769403824521935,
+      "grad_norm": 3.085162084216117,
+      "kl": 0.607421875,
+      "learning_rate": 1.2516351065824864e-05,
+      "loss": 0.9745,
+      "reward": 2.462671160697937,
+      "reward_std": 0.7659648507833481,
+      "rewards/accuracy_reward": 0.7135416865348816,
+      "rewards/reasoning_steps_reward": 0.973958358168602,
+      "rewards/repetition_penalty_reward": -0.12196437083184719,
+      "rewards/tag_count_reward": 0.8971354365348816,
       "step": 318
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 220.1302146911621,
-      "epoch": 0.9565217391304348,
-      "grad_norm": 104.39513312177709,
-      "kl": 9.4296875,
-      "learning_rate": 1.0799414351212234e-07,
-      "loss": 2.6495,
-      "reward": 2.5906622409820557,
-      "reward_std": 0.7361035495996475,
-      "rewards/accuracy_reward": 0.786458358168602,
-      "rewards/reasoning_steps_reward": 0.9218750596046448,
-      "rewards/repetition_penalty_reward": -0.006994032271904871,
-      "rewards/tag_count_reward": 0.8893229365348816,
+      "completion_length": 337.12500762939453,
+      "epoch": 0.47844019497562806,
+      "grad_norm": 2.5910813299992617,
+      "kl": 0.67919921875,
+      "learning_rate": 1.2465557039460048e-05,
+      "loss": 0.9239,
+      "reward": 2.1963090300559998,
+      "reward_std": 0.973323866724968,
+      "rewards/accuracy_reward": 0.598958358168602,
+      "rewards/reasoning_steps_reward": 0.960069477558136,
+      "rewards/repetition_penalty_reward": -0.21167714521288872,
+      "rewards/tag_count_reward": 0.848958358168602,
       "step": 319
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 260.72396087646484,
-      "epoch": 0.95952023988006,
-      "grad_norm": 172.54963273214497,
-      "kl": 12.796875,
-      "learning_rate": 9.314053963669245e-08,
-      "loss": 2.6928,
-      "reward": 2.501668691635132,
-      "reward_std": 0.9300008714199066,
-      "rewards/accuracy_reward": 0.7968750149011612,
-      "rewards/reasoning_steps_reward": 0.8750000596046448,
-      "rewards/repetition_penalty_reward": -0.006144002894870937,
-      "rewards/tag_count_reward": 0.8359375298023224,
+      "completion_length": 282.5729293823242,
+      "epoch": 0.4799400074990626,
+      "grad_norm": 0.6692834432378754,
+      "kl": 0.4453125,
+      "learning_rate": 1.241469519272453e-05,
+      "loss": 1.0469,
+      "reward": 2.4757498502731323,
+      "reward_std": 0.9635120928287506,
+      "rewards/accuracy_reward": 0.8072916716337204,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.1574968732893467,
+      "rewards/tag_count_reward": 0.860677108168602,
       "step": 320
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 222.2656364440918,
-      "epoch": 0.9625187406296851,
-      "grad_norm": 58.41749025051054,
-      "kl": 8.359375,
-      "learning_rate": 7.938061350773241e-08,
-      "loss": 2.5917,
-      "reward": 2.5556360483169556,
-      "reward_std": 0.8327264338731766,
-      "rewards/accuracy_reward": 0.7656250298023224,
-      "rewards/reasoning_steps_reward": 0.9166666865348816,
-      "rewards/repetition_penalty_reward": -0.006864078342914581,
-      "rewards/tag_count_reward": 0.880208358168602,
+      "completion_length": 239.3645896911621,
+      "epoch": 0.4814398200224972,
+      "grad_norm": 1.2551570949769288,
+      "kl": 0.373046875,
+      "learning_rate": 1.2363766924681178e-05,
+      "loss": 0.9825,
+      "reward": 2.528091847896576,
+      "reward_std": 0.9040358066558838,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.9635416716337204,
+      "rewards/repetition_penalty_reward": -0.10602277517318726,
+      "rewards/tag_count_reward": 0.8580729365348816,
       "step": 321
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 199.64583587646484,
-      "epoch": 0.9655172413793104,
-      "grad_norm": 21.274250522963346,
-      "kl": 6.21875,
-      "learning_rate": 6.671588416617081e-08,
-      "loss": 2.0353,
-      "reward": 2.5732468962669373,
-      "reward_std": 0.6762567013502121,
-      "rewards/accuracy_reward": 0.7343750149011612,
-      "rewards/reasoning_steps_reward": 0.934027761220932,
-      "rewards/repetition_penalty_reward": -0.00661420589312911,
-      "rewards/tag_count_reward": 0.911458358168602,
+      "completion_length": 236.21875381469727,
+      "epoch": 0.48293963254593175,
+      "grad_norm": 0.7609484269615335,
+      "kl": 0.39013671875,
+      "learning_rate": 1.2312773636219919e-05,
+      "loss": 0.6847,
+      "reward": 2.2356700897216797,
+      "reward_std": 0.7553437650203705,
+      "rewards/accuracy_reward": 0.5729166828095913,
+      "rewards/reasoning_steps_reward": 0.9513889253139496,
+      "rewards/repetition_penalty_reward": -0.10764593631029129,
+      "rewards/tag_count_reward": 0.8190104365348816,
       "step": 322
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 248.5416717529297,
-      "epoch": 0.9685157421289355,
-      "grad_norm": 30.694601447554664,
-      "kl": 9.1796875,
-      "learning_rate": 5.5147749747582744e-08,
-      "loss": 2.2922,
-      "reward": 2.4823118448257446,
-      "reward_std": 0.845875695347786,
-      "rewards/accuracy_reward": 0.7343750149011612,
-      "rewards/reasoning_steps_reward": 0.8975694477558136,
-      "rewards/repetition_penalty_reward": -0.0037993842852301896,
-      "rewards/tag_count_reward": 0.8541667014360428,
+      "completion_length": 282.00521087646484,
+      "epoch": 0.4844394450693663,
+      "grad_norm": 1.332331758605412,
+      "kl": 0.44873046875,
+      "learning_rate": 1.2261716730019202e-05,
+      "loss": 0.9536,
+      "reward": 2.294684648513794,
+      "reward_std": 1.0781968086957932,
+      "rewards/accuracy_reward": 0.7187500149011612,
+      "rewards/reasoning_steps_reward": 0.9131944626569748,
+      "rewards/repetition_penalty_reward": -0.14585364237427711,
+      "rewards/tag_count_reward": 0.8085937649011612,
       "step": 323
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 226.02605056762695,
-      "epoch": 0.9715142428785607,
-      "grad_norm": 37.717634935459415,
-      "kl": 6.8359375,
-      "learning_rate": 4.467748732783994e-08,
-      "loss": 2.167,
-      "reward": 2.6481770277023315,
-      "reward_std": 0.7722356021404266,
-      "rewards/accuracy_reward": 0.8385416716337204,
-      "rewards/reasoning_steps_reward": 0.9270833283662796,
-      "rewards/repetition_penalty_reward": -0.0015625922533217818,
-      "rewards/tag_count_reward": 0.884114608168602,
+      "completion_length": 238.16146850585938,
+      "epoch": 0.4859392575928009,
+      "grad_norm": 6163963.310883091,
+      "kl": 671744.279296875,
+      "learning_rate": 1.2210597610507418e-05,
+      "loss": 21419.6289,
+      "reward": 2.206323742866516,
+      "reward_std": 0.8206494450569153,
+      "rewards/accuracy_reward": 0.5312500298023224,
+      "rewards/reasoning_steps_reward": 0.9496528208255768,
+      "rewards/repetition_penalty_reward": -0.10661044530570507,
+      "rewards/tag_count_reward": 0.8320312649011612,
       "step": 324
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 195.2916717529297,
-      "epoch": 0.974512743628186,
-      "grad_norm": 116.25145619245001,
-      "kl": 4.23828125,
-      "learning_rate": 3.530625278212685e-08,
-      "loss": 1.9012,
-      "reward": 2.604085922241211,
-      "reward_std": 0.7135767489671707,
+      "completion_length": 267.7760543823242,
+      "epoch": 0.4874390701162355,
+      "grad_norm": 1.168918265059183,
+      "kl": 0.455078125,
+      "learning_rate": 1.2159417683824266e-05,
+      "loss": 0.9769,
+      "reward": 2.380450427532196,
+      "reward_std": 0.9683381170034409,
       "rewards/accuracy_reward": 0.7604166865348816,
-      "rewards/reasoning_steps_reward": 0.9392361491918564,
-      "rewards/repetition_penalty_reward": -0.004421164951054379,
-      "rewards/tag_count_reward": 0.9088541865348816,
+      "rewards/reasoning_steps_reward": 0.9357639104127884,
+      "rewards/repetition_penalty_reward": -0.13864684104919434,
+      "rewards/tag_count_reward": 0.8229166865348816,
       "step": 325
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 225.10937881469727,
-      "epoch": 0.9775112443778111,
-      "grad_norm": 47.023072554995835,
-      "kl": 6.8359375,
-      "learning_rate": 2.7035080657338287e-08,
-      "loss": 2.2076,
-      "reward": 2.577496290206909,
-      "reward_std": 0.8721825927495956,
-      "rewards/accuracy_reward": 0.8125000298023224,
-      "rewards/reasoning_steps_reward": 0.9010417014360428,
-      "rewards/repetition_penalty_reward": -0.005837142816744745,
-      "rewards/tag_count_reward": 0.8697916865348816,
+      "completion_length": 225.47396087646484,
+      "epoch": 0.48893888263967006,
+      "grad_norm": 1.2436386439703448,
+      "kl": 0.51953125,
+      "learning_rate": 1.2108178357782079e-05,
+      "loss": 0.7213,
+      "reward": 2.281162142753601,
+      "reward_std": 0.5486664846539497,
+      "rewards/accuracy_reward": 0.5052083432674408,
+      "rewards/reasoning_steps_reward": 0.9791667014360428,
+      "rewards/repetition_penalty_reward": -0.08342129364609718,
+      "rewards/tag_count_reward": 0.880208358168602,
       "step": 326
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 232.28125762939453,
-      "epoch": 0.9805097451274363,
-      "grad_norm": 91.66318973005008,
-      "kl": 6.0703125,
-      "learning_rate": 1.986488405786524e-08,
-      "loss": 2.2354,
-      "reward": 2.548638343811035,
-      "reward_std": 0.8752985298633575,
-      "rewards/accuracy_reward": 0.7812500149011612,
-      "rewards/reasoning_steps_reward": 0.9079861640930176,
-      "rewards/repetition_penalty_reward": -0.006483202683739364,
-      "rewards/tag_count_reward": 0.8658854365348816,
+      "completion_length": 214.96875762939453,
+      "epoch": 0.4904386951631046,
+      "grad_norm": 0.7171717418281628,
+      "kl": 0.3896484375,
+      "learning_rate": 1.205688104182709e-05,
+      "loss": 0.574,
+      "reward": 2.455682098865509,
+      "reward_std": 0.5591145232319832,
+      "rewards/accuracy_reward": 0.6770833432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.07252970896661282,
+      "rewards/tag_count_reward": 0.8789062649011612,
       "step": 327
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 203.97396087646484,
-      "epoch": 0.9835082458770614,
-      "grad_norm": 93.30979532325884,
-      "kl": 5.15625,
-      "learning_rate": 1.3796454544796612e-08,
-      "loss": 2.018,
-      "reward": 2.5490445494651794,
-      "reward_std": 0.7066824734210968,
-      "rewards/accuracy_reward": 0.7343750149011612,
-      "rewards/reasoning_steps_reward": 0.921875,
-      "rewards/repetition_penalty_reward": -0.005643073527608067,
-      "rewards/tag_count_reward": 0.8984375149011612,
+      "completion_length": 203.1354217529297,
+      "epoch": 0.4919385076865392,
+      "grad_norm": 627.8308768620077,
+      "kl": 74.83154296875,
+      "learning_rate": 1.2005527147000663e-05,
+      "loss": 2.9222,
+      "reward": 2.3696823120117188,
+      "reward_std": 0.6213457994163036,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9670138955116272,
+      "rewards/repetition_penalty_reward": -0.06347744725644588,
+      "rewards/tag_count_reward": 0.903645858168602,
       "step": 328
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 209.05208587646484,
-      "epoch": 0.9865067466266867,
-      "grad_norm": 55.95983139614906,
-      "kl": 5.65625,
-      "learning_rate": 8.83046204853133e-09,
-      "loss": 2.0135,
-      "reward": 2.6783204674720764,
-      "reward_std": 0.7424749433994293,
-      "rewards/accuracy_reward": 0.8697916865348816,
-      "rewards/reasoning_steps_reward": 0.9201388955116272,
-      "rewards/repetition_penalty_reward": -0.0035372423008084297,
-      "rewards/tag_count_reward": 0.8919271230697632,
+      "completion_length": 213.8541717529297,
+      "epoch": 0.49343832020997375,
+      "grad_norm": 1.0700058348332315,
+      "kl": 0.490234375,
+      "learning_rate": 1.1954118085900503e-05,
+      "loss": 0.4632,
+      "reward": 2.4596543312072754,
+      "reward_std": 0.5267078801989555,
+      "rewards/accuracy_reward": 0.6354166865348816,
+      "rewards/reasoning_steps_reward": 0.9704861491918564,
+      "rewards/repetition_penalty_reward": -0.07202969677746296,
+      "rewards/tag_count_reward": 0.9257812649011612,
       "step": 329
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 248.3541717529297,
-      "epoch": 0.9895052473763118,
-      "grad_norm": 45.74323780759467,
-      "kl": 6.8671875,
-      "learning_rate": 4.967454794823079e-09,
-      "loss": 2.1326,
-      "reward": 2.388830304145813,
-      "reward_std": 0.7917287796735764,
-      "rewards/accuracy_reward": 0.6354167014360428,
-      "rewards/reasoning_steps_reward": 0.8993055522441864,
-      "rewards/repetition_penalty_reward": -0.006569080462213606,
-      "rewards/tag_count_reward": 0.8606771230697632,
+      "completion_length": 257.5260467529297,
+      "epoch": 0.4949381327334083,
+      "grad_norm": 1.1628503464475846,
+      "kl": 0.36474609375,
+      "learning_rate": 1.1902655272641756e-05,
+      "loss": 0.6424,
+      "reward": 2.5043947100639343,
+      "reward_std": 0.7565517127513885,
+      "rewards/accuracy_reward": 0.7187500149011612,
+      "rewards/reasoning_steps_reward": 0.9756944477558136,
+      "rewards/repetition_penalty_reward": -0.10411231219768524,
+      "rewards/tag_count_reward": 0.9140625149011612,
       "step": 330
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 226.32291793823242,
-      "epoch": 0.992503748125937,
-      "grad_norm": 18.948649605644253,
-      "kl": 6.3515625,
-      "learning_rate": 2.2078592442553725e-09,
-      "loss": 1.9433,
-      "reward": 2.566007614135742,
-      "reward_std": 0.7248884737491608,
-      "rewards/accuracy_reward": 0.7656250149011612,
-      "rewards/reasoning_steps_reward": 0.923611119389534,
-      "rewards/repetition_penalty_reward": -0.006041057640686631,
-      "rewards/tag_count_reward": 0.8828125149011612,
+      "completion_length": 261.04688262939453,
+      "epoch": 0.49643794525684287,
+      "grad_norm": 0.6255524381969977,
+      "kl": 0.37353515625,
+      "learning_rate": 1.1851140122818155e-05,
+      "loss": 0.5753,
+      "reward": 2.406662940979004,
+      "reward_std": 0.8139507248997688,
+      "rewards/accuracy_reward": 0.661458358168602,
+      "rewards/reasoning_steps_reward": 0.958333358168602,
+      "rewards/repetition_penalty_reward": -0.11417047679424286,
+      "rewards/tag_count_reward": 0.9010416865348816,
       "step": 331
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 199.53646087646484,
-      "epoch": 0.9955022488755623,
-      "grad_norm": 54.47996371401269,
-      "kl": 5.13671875,
-      "learning_rate": 5.519800451625479e-10,
-      "loss": 1.9846,
-      "reward": 2.375428855419159,
-      "reward_std": 0.6744664013385773,
-      "rewards/accuracy_reward": 0.5416666865348816,
-      "rewards/reasoning_steps_reward": 0.937500074505806,
-      "rewards/repetition_penalty_reward": -0.00477961910655722,
-      "rewards/tag_count_reward": 0.9010416716337204,
+      "completion_length": 257.12500381469727,
+      "epoch": 0.4979377577802775,
+      "grad_norm": 0.7463997657803588,
+      "kl": 0.373046875,
+      "learning_rate": 1.1799574053463048e-05,
+      "loss": 0.6145,
+      "reward": 2.637652814388275,
+      "reward_std": 0.6368973329663277,
+      "rewards/accuracy_reward": 0.8229166865348816,
+      "rewards/reasoning_steps_reward": 0.9652777761220932,
+      "rewards/repetition_penalty_reward": -0.09976036194711924,
+      "rewards/tag_count_reward": 0.9492187649011612,
       "step": 332
     },
     {
       "clip_ratio": 0.0,
-      "completion_length": 230.6927146911621,
-      "epoch": 0.9985007496251874,
-      "grad_norm": 58.000699042461555,
-      "kl": 7.21875,
-      "learning_rate": 0.0,
-      "loss": 2.3714,
-      "reward": 2.6160709261894226,
-      "reward_std": 0.8584436923265457,
-      "rewards/accuracy_reward": 0.8437500149011612,
-      "rewards/reasoning_steps_reward": 0.9062500149011612,
-      "rewards/repetition_penalty_reward": -0.00632495793979615,
-      "rewards/tag_count_reward": 0.8723958432674408,
+      "completion_length": 238.8333396911621,
+      "epoch": 0.49943757030371205,
+      "grad_norm": 0.7834690979097384,
+      "kl": 0.35546875,
+      "learning_rate": 1.1747958483010438e-05,
+      "loss": 0.5165,
+      "reward": 2.6003527641296387,
+      "reward_std": 0.7082152366638184,
+      "rewards/accuracy_reward": 0.8020833432674408,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.09148765727877617,
+      "rewards/tag_count_reward": 0.9244791865348816,
       "step": 333
     },
     {
-      "epoch": 0.9985007496251874,
-      "step": 333,
+      "clip_ratio": 0.0,
+      "completion_length": 236.93750381469727,
+      "epoch": 0.5009373828271466,
+      "grad_norm": 0.7440577499317964,
+      "kl": 0.36376953125,
+      "learning_rate": 1.1696294831255961e-05,
+      "loss": 0.5263,
+      "reward": 2.6781486868858337,
+      "reward_std": 0.6196286901831627,
+      "rewards/accuracy_reward": 0.8593750149011612,
+      "rewards/reasoning_steps_reward": 0.9531250149011612,
+      "rewards/repetition_penalty_reward": -0.09008055832237005,
+      "rewards/tag_count_reward": 0.9557291865348816,
+      "step": 334
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 286.8072967529297,
+      "epoch": 0.5024371953505812,
+      "grad_norm": 0.736069132286744,
+      "kl": 0.4169921875,
+      "learning_rate": 1.1644584519317828e-05,
+      "loss": 0.7123,
+      "reward": 2.399460792541504,
+      "reward_std": 0.8988363444805145,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9479167014360428,
+      "rewards/repetition_penalty_reward": -0.12918509356677532,
+      "rewards/tag_count_reward": 0.9140625149011612,
+      "step": 335
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 276.28125762939453,
+      "epoch": 0.5039370078740157,
+      "grad_norm": 0.5872022790584697,
+      "kl": 0.36669921875,
+      "learning_rate": 1.159282896959774e-05,
+      "loss": 0.6407,
+      "reward": 2.4547627568244934,
+      "reward_std": 0.7832493036985397,
+      "rewards/accuracy_reward": 0.6822916865348816,
+      "rewards/reasoning_steps_reward": 0.9687500447034836,
+      "rewards/repetition_penalty_reward": -0.11945611983537674,
+      "rewards/tag_count_reward": 0.9231770932674408,
+      "step": 336
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.8177146911621,
+      "epoch": 0.5054368203974503,
+      "grad_norm": 0.7563101618131821,
+      "kl": 0.3818359375,
+      "learning_rate": 1.1541029605741758e-05,
+      "loss": 0.68,
+      "reward": 2.5487101078033447,
+      "reward_std": 0.7277443781495094,
+      "rewards/accuracy_reward": 0.7395833432674408,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.10580387711524963,
+      "rewards/tag_count_reward": 0.9375000298023224,
+      "step": 337
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.54166793823242,
+      "epoch": 0.5069366329208849,
+      "grad_norm": 0.7629172892618677,
+      "kl": 0.34228515625,
+      "learning_rate": 1.1489187852601147e-05,
+      "loss": 0.4758,
+      "reward": 2.738324999809265,
+      "reward_std": 0.49979202449321747,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.056813917588442564,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 338
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 286.77083587646484,
+      "epoch": 0.5084364454443194,
+      "grad_norm": 1.2875833340510152,
+      "kl": 0.4716796875,
+      "learning_rate": 1.143730513619317e-05,
+      "loss": 0.6866,
+      "reward": 2.457393527030945,
+      "reward_std": 0.8834079504013062,
+      "rewards/accuracy_reward": 0.739583358168602,
+      "rewards/reasoning_steps_reward": 0.9322916865348816,
+      "rewards/repetition_penalty_reward": -0.1311482023447752,
+      "rewards/tag_count_reward": 0.9166666865348816,
+      "step": 339
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 262.95312881469727,
+      "epoch": 0.509936257967754,
+      "grad_norm": 1.9340766781030916,
+      "kl": 0.64697265625,
+      "learning_rate": 1.1385382883661881e-05,
+      "loss": 0.4332,
+      "reward": 2.5942657589912415,
+      "reward_std": 0.6046310998499393,
+      "rewards/accuracy_reward": 0.8177083432674408,
+      "rewards/reasoning_steps_reward": 0.9600694626569748,
+      "rewards/repetition_penalty_reward": -0.11710584536194801,
+      "rewards/tag_count_reward": 0.9335937649011612,
+      "step": 340
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 240.20313262939453,
+      "epoch": 0.5114360704911886,
+      "grad_norm": 0.836838721680916,
+      "kl": 0.41064453125,
+      "learning_rate": 1.1333422523238858e-05,
+      "loss": 0.5716,
+      "reward": 2.467832863330841,
+      "reward_std": 0.6241201423108578,
+      "rewards/accuracy_reward": 0.6302083507180214,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.08598665753379464,
+      "rewards/tag_count_reward": 0.9479166865348816,
+      "step": 341
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 275.5885467529297,
+      "epoch": 0.5129358830146231,
+      "grad_norm": 0.6233336001065312,
+      "kl": 0.3583984375,
+      "learning_rate": 1.1281425484203908e-05,
+      "loss": 0.5216,
+      "reward": 2.375839054584503,
+      "reward_std": 0.570319652557373,
+      "rewards/accuracy_reward": 0.5781250149011612,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.12459501624107361,
+      "rewards/tag_count_reward": 0.9361979365348816,
+      "step": 342
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 270.3385543823242,
+      "epoch": 0.5144356955380578,
+      "grad_norm": 0.781462095769395,
+      "kl": 0.4208984375,
+      "learning_rate": 1.122939319684577e-05,
+      "loss": 0.8224,
+      "reward": 2.3723965883255005,
+      "reward_std": 0.892381027340889,
+      "rewards/accuracy_reward": 0.6406250074505806,
+      "rewards/reasoning_steps_reward": 0.9409722536802292,
+      "rewards/repetition_penalty_reward": -0.1271694442257285,
+      "rewards/tag_count_reward": 0.9179687649011612,
+      "step": 343
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 243.04167556762695,
+      "epoch": 0.5159355080614924,
+      "grad_norm": 0.5806453568608354,
+      "kl": 0.3466796875,
+      "learning_rate": 1.1177327092422761e-05,
+      "loss": 0.682,
+      "reward": 2.4070491790771484,
+      "reward_std": 0.619476206600666,
+      "rewards/accuracy_reward": 0.6145833432674408,
+      "rewards/reasoning_steps_reward": 0.9479166865348816,
+      "rewards/repetition_penalty_reward": -0.1007634038105607,
+      "rewards/tag_count_reward": 0.9453125149011612,
+      "step": 344
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 239.41146850585938,
+      "epoch": 0.5174353205849269,
+      "grad_norm": 2.1866070331433964,
+      "kl": 0.57275390625,
+      "learning_rate": 1.1125228603123408e-05,
+      "loss": 0.6429,
+      "reward": 2.6293810606002808,
+      "reward_std": 0.6588820517063141,
+      "rewards/accuracy_reward": 0.802083358168602,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.09327533282339573,
+      "rewards/tag_count_reward": 0.9570312649011612,
+      "step": 345
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 198.65105056762695,
+      "epoch": 0.5189351331083615,
+      "grad_norm": 0.6514816604889703,
+      "kl": 0.34228515625,
+      "learning_rate": 1.107309916202705e-05,
+      "loss": 0.3713,
+      "reward": 2.694413661956787,
+      "reward_std": 0.3578246496617794,
+      "rewards/accuracy_reward": 0.7864583432674408,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.051680129021406174,
+      "rewards/tag_count_reward": 0.9752604365348816,
+      "step": 346
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 208.5416717529297,
+      "epoch": 0.520434945631796,
+      "grad_norm": 6.6032813165402935,
+      "kl": 0.46484375,
+      "learning_rate": 1.1020940203064425e-05,
+      "loss": 0.3975,
+      "reward": 2.6713653802871704,
+      "reward_std": 0.4076192807406187,
+      "rewards/accuracy_reward": 0.7604167014360428,
+      "rewards/reasoning_steps_reward": 0.9913194626569748,
+      "rewards/repetition_penalty_reward": -0.05563134513795376,
+      "rewards/tag_count_reward": 0.9752604365348816,
+      "step": 347
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 221.29688262939453,
+      "epoch": 0.5219347581552306,
+      "grad_norm": 2.4618917497796233,
+      "kl": 0.431640625,
+      "learning_rate": 1.096875316097822e-05,
+      "loss": 0.519,
+      "reward": 2.731937885284424,
+      "reward_std": 0.5047559663653374,
+      "rewards/accuracy_reward": 0.8593750298023224,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.07491973042488098,
+      "rewards/tag_count_reward": 0.9648437798023224,
+      "step": 348
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 252.98959350585938,
+      "epoch": 0.5234345706786652,
+      "grad_norm": 0.6526608511425668,
+      "kl": 0.38037109375,
+      "learning_rate": 1.0916539471283607e-05,
+      "loss": 0.4489,
+      "reward": 2.611004889011383,
+      "reward_std": 0.6215125024318695,
+      "rewards/accuracy_reward": 0.8229166865348816,
+      "rewards/reasoning_steps_reward": 0.9531250298023224,
+      "rewards/repetition_penalty_reward": -0.10904721543192863,
+      "rewards/tag_count_reward": 0.9440104365348816,
+      "step": 349
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 238.0572967529297,
+      "epoch": 0.5249343832020997,
+      "grad_norm": 1.3490617390001882,
+      "kl": 0.43798828125,
+      "learning_rate": 1.0864300570228757e-05,
+      "loss": 0.5916,
+      "reward": 2.349369764328003,
+      "reward_std": 0.591891311109066,
+      "rewards/accuracy_reward": 0.5625000074505806,
+      "rewards/reasoning_steps_reward": 0.9496527910232544,
+      "rewards/repetition_penalty_reward": -0.11200192756950855,
+      "rewards/tag_count_reward": 0.9492187798023224,
+      "step": 350
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 202.79688262939453,
+      "epoch": 0.5264341957255343,
+      "grad_norm": 0.7377470109327692,
+      "kl": 0.35205078125,
+      "learning_rate": 1.0812037894755336e-05,
+      "loss": 0.4294,
+      "reward": 2.7148218750953674,
+      "reward_std": 0.4056655466556549,
+      "rewards/accuracy_reward": 0.802083358168602,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.06121993251144886,
+      "rewards/tag_count_reward": 0.973958358168602,
+      "step": 351
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 195.13021087646484,
+      "epoch": 0.5279340082489689,
+      "grad_norm": 0.9262254939525653,
+      "kl": 0.3154296875,
+      "learning_rate": 1.0759752882458972e-05,
+      "loss": 0.4753,
+      "reward": 2.716402769088745,
+      "reward_std": 0.4552469737827778,
+      "rewards/accuracy_reward": 0.8072916865348816,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.05963901709765196,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 352
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 222.2343864440918,
+      "epoch": 0.5294338207724034,
+      "grad_norm": 1.2036216767618395,
+      "kl": 0.32861328125,
+      "learning_rate": 1.0707446971549717e-05,
+      "loss": 0.6401,
+      "reward": 2.590088963508606,
+      "reward_std": 0.5779636353254318,
+      "rewards/accuracy_reward": 0.755208358168602,
+      "rewards/reasoning_steps_reward": 0.9704861491918564,
+      "rewards/repetition_penalty_reward": -0.09524106327444315,
+      "rewards/tag_count_reward": 0.9596354365348816,
+      "step": 353
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 203.28646087646484,
+      "epoch": 0.530933633295838,
+      "grad_norm": 0.8356292683714005,
+      "kl": 0.3642578125,
+      "learning_rate": 1.0655121600812482e-05,
+      "loss": 0.6197,
+      "reward": 2.7302719950675964,
+      "reward_std": 0.5387823097407818,
+      "rewards/accuracy_reward": 0.8593750298023224,
+      "rewards/reasoning_steps_reward": 0.9791667014360428,
+      "rewards/repetition_penalty_reward": -0.06920717097818851,
+      "rewards/tag_count_reward": 0.9609375,
+      "step": 354
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 219.63542556762695,
+      "epoch": 0.5324334458192725,
+      "grad_norm": 0.9550640481800935,
+      "kl": 0.4052734375,
+      "learning_rate": 1.0602778209567462e-05,
+      "loss": 0.7535,
+      "reward": 2.541829466819763,
+      "reward_std": 0.7257425487041473,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9548610895872116,
+      "rewards/repetition_penalty_reward": -0.08751083724200726,
+      "rewards/tag_count_reward": 0.9453125298023224,
+      "step": 355
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.01562881469727,
+      "epoch": 0.5339332583427071,
+      "grad_norm": 1.070677353710695,
+      "kl": 0.404296875,
+      "learning_rate": 1.0550418237630547e-05,
+      "loss": 1.0339,
+      "reward": 2.5512691140174866,
+      "reward_std": 0.7947003692388535,
+      "rewards/accuracy_reward": 0.7500000149011612,
+      "rewards/reasoning_steps_reward": 0.9496527910232544,
+      "rewards/repetition_penalty_reward": -0.09499827027320862,
+      "rewards/tag_count_reward": 0.9466145932674408,
+      "step": 356
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 182.71875762939453,
+      "epoch": 0.5354330708661418,
+      "grad_norm": 2.6963525226313068,
+      "kl": 0.4677734375,
+      "learning_rate": 1.0498043125273714e-05,
+      "loss": 0.5703,
+      "reward": 2.6821329593658447,
+      "reward_std": 0.52312882989645,
+      "rewards/accuracy_reward": 0.8020833432674408,
+      "rewards/reasoning_steps_reward": 0.9722222536802292,
+      "rewards/repetition_penalty_reward": -0.05831848084926605,
+      "rewards/tag_count_reward": 0.966145858168602,
+      "step": 357
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 222.8020896911621,
+      "epoch": 0.5369328833895763,
+      "grad_norm": 4.425571727643998,
+      "kl": 0.5517578125,
+      "learning_rate": 1.0445654313185402e-05,
+      "loss": 0.7903,
+      "reward": 2.470693051815033,
+      "reward_std": 0.6807428225874901,
+      "rewards/accuracy_reward": 0.6562500298023224,
+      "rewards/reasoning_steps_reward": 0.96875,
+      "rewards/repetition_penalty_reward": -0.09701530169695616,
+      "rewards/tag_count_reward": 0.942708358168602,
+      "step": 358
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 246.9947967529297,
+      "epoch": 0.5384326959130109,
+      "grad_norm": 2.2749469499741446,
+      "kl": 0.54541015625,
+      "learning_rate": 1.0393253242430898e-05,
+      "loss": 0.766,
+      "reward": 2.610697090625763,
+      "reward_std": 0.7646311447024345,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.954861119389534,
+      "rewards/repetition_penalty_reward": -0.11760163493454456,
+      "rewards/tag_count_reward": 0.9348958432674408,
+      "step": 359
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.45833587646484,
+      "epoch": 0.5399325084364455,
+      "grad_norm": 0.9100444614802594,
+      "kl": 0.39501953125,
+      "learning_rate": 1.0340841354412688e-05,
+      "loss": 0.6344,
+      "reward": 2.6307865977287292,
+      "reward_std": 0.5916828811168671,
+      "rewards/accuracy_reward": 0.8177083432674408,
+      "rewards/reasoning_steps_reward": 0.9513889104127884,
+      "rewards/repetition_penalty_reward": -0.08752944599837065,
+      "rewards/tag_count_reward": 0.9492187798023224,
+      "step": 360
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 202.50521087646484,
+      "epoch": 0.54143232095988,
+      "grad_norm": 1.3542848450477425,
+      "kl": 0.458984375,
+      "learning_rate": 1.0288420090830803e-05,
+      "loss": 0.5782,
+      "reward": 2.5363489985466003,
+      "reward_std": 0.5847525298595428,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.07215808611363173,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 361
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.29688262939453,
+      "epoch": 0.5429321334833146,
+      "grad_norm": 1.0691193581184522,
+      "kl": 0.54150390625,
+      "learning_rate": 1.0235990893643184e-05,
+      "loss": 0.614,
+      "reward": 2.446419894695282,
+      "reward_std": 0.5024325773119926,
+      "rewards/accuracy_reward": 0.6197916939854622,
+      "rewards/reasoning_steps_reward": 0.9878472536802292,
+      "rewards/repetition_penalty_reward": -0.10002116020768881,
+      "rewards/tag_count_reward": 0.9388020932674408,
+      "step": 362
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 239.64063262939453,
+      "epoch": 0.5444319460067492,
+      "grad_norm": 1.0363445445035244,
+      "kl": 0.48095703125,
+      "learning_rate": 1.0183555205025986e-05,
+      "loss": 0.8301,
+      "reward": 2.533776819705963,
+      "reward_std": 0.8274639621376991,
+      "rewards/accuracy_reward": 0.7812500298023224,
+      "rewards/reasoning_steps_reward": 0.9479166865348816,
+      "rewards/repetition_penalty_reward": -0.11335871182382107,
+      "rewards/tag_count_reward": 0.9179687798023224,
+      "step": 363
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 247.2395896911621,
+      "epoch": 0.5459317585301837,
+      "grad_norm": 1.4285370079745647,
+      "kl": 0.4580078125,
+      "learning_rate": 1.0131114467333935e-05,
+      "loss": 0.9421,
+      "reward": 2.5757275223731995,
+      "reward_std": 0.940077006816864,
+      "rewards/accuracy_reward": 0.848958358168602,
+      "rewards/reasoning_steps_reward": 0.9409722685813904,
+      "rewards/repetition_penalty_reward": -0.1282655205577612,
+      "rewards/tag_count_reward": 0.9140625149011612,
+      "step": 364
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 203.9322967529297,
+      "epoch": 0.5474315710536183,
+      "grad_norm": 2.976909353877276,
+      "kl": 0.41650390625,
+      "learning_rate": 1.0078670123060638e-05,
+      "loss": 0.5808,
+      "reward": 2.7071104049682617,
+      "reward_std": 0.5107903629541397,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9878472238779068,
+      "rewards/repetition_penalty_reward": -0.0724035445600748,
+      "rewards/tag_count_reward": 0.9531250149011612,
+      "step": 365
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 241.50000381469727,
+      "epoch": 0.5489313835770528,
+      "grad_norm": 1.5290214581037804,
+      "kl": 0.45263671875,
+      "learning_rate": 1.002622361479891e-05,
+      "loss": 0.7503,
+      "reward": 2.2156243920326233,
+      "reward_std": 0.8056820183992386,
+      "rewards/accuracy_reward": 0.510416679084301,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.11901108920574188,
+      "rewards/tag_count_reward": 0.8658854216337204,
+      "step": 366
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 285.9895935058594,
+      "epoch": 0.5504311961004874,
+      "grad_norm": 1.6300575912920117,
+      "kl": 0.43505859375,
+      "learning_rate": 9.973776385201093e-06,
+      "loss": 0.9551,
+      "reward": 2.446646273136139,
+      "reward_std": 0.9035588353872299,
+      "rewards/accuracy_reward": 0.7604166865348816,
+      "rewards/reasoning_steps_reward": 0.9756944924592972,
+      "rewards/repetition_penalty_reward": -0.16316283866763115,
+      "rewards/tag_count_reward": 0.8736979365348816,
+      "step": 367
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 249.50000762939453,
+      "epoch": 0.551931008623922,
+      "grad_norm": 2.432545958790851,
+      "kl": 0.43017578125,
+      "learning_rate": 9.921329876939365e-06,
+      "loss": 0.6609,
+      "reward": 2.449833869934082,
+      "reward_std": 0.7589901238679886,
+      "rewards/accuracy_reward": 0.6927083656191826,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.12221488542854786,
+      "rewards/tag_count_reward": 0.9088542014360428,
+      "step": 368
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 288.2135543823242,
+      "epoch": 0.5534308211473565,
+      "grad_norm": 1.139092646573938,
+      "kl": 0.4443359375,
+      "learning_rate": 9.868885532666068e-06,
+      "loss": 0.9903,
+      "reward": 2.31951966881752,
+      "reward_std": 1.0058775544166565,
+      "rewards/accuracy_reward": 0.661458358168602,
+      "rewards/reasoning_steps_reward": 0.9409722685813904,
+      "rewards/repetition_penalty_reward": -0.17223386466503143,
+      "rewards/tag_count_reward": 0.8893229514360428,
+      "step": 369
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 237.46355056762695,
+      "epoch": 0.5549306336707911,
+      "grad_norm": 0.9950921053074119,
+      "kl": 0.3984375,
+      "learning_rate": 9.816444794974018e-06,
+      "loss": 0.6193,
+      "reward": 2.5471015572547913,
+      "reward_std": 0.6440750285983086,
+      "rewards/accuracy_reward": 0.7395833730697632,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.10133606940507889,
+      "rewards/tag_count_reward": 0.9401041716337204,
+      "step": 370
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.4583396911621,
+      "epoch": 0.5564304461942258,
+      "grad_norm": 1.440820850213343,
+      "kl": 0.43701171875,
+      "learning_rate": 9.76400910635682e-06,
+      "loss": 0.8709,
+      "reward": 2.2166844606399536,
+      "reward_std": 0.7375850528478622,
+      "rewards/accuracy_reward": 0.4583333432674408,
+      "rewards/reasoning_steps_reward": 0.9461805820465088,
+      "rewards/repetition_penalty_reward": -0.11100666876882315,
+      "rewards/tag_count_reward": 0.9231770932674408,
+      "step": 371
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.1979217529297,
+      "epoch": 0.5579302587176603,
+      "grad_norm": 1.089340536225687,
+      "kl": 0.47412109375,
+      "learning_rate": 9.7115799091692e-06,
+      "loss": 0.4976,
+      "reward": 2.5988621711730957,
+      "reward_std": 0.6001707911491394,
+      "rewards/accuracy_reward": 0.7604166716337204,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.07778726145625114,
+      "rewards/tag_count_reward": 0.9492187649011612,
+      "step": 372
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 204.90625381469727,
+      "epoch": 0.5594300712410949,
+      "grad_norm": 1.0735796637227366,
+      "kl": 0.4130859375,
+      "learning_rate": 9.659158645587319e-06,
+      "loss": 0.6675,
+      "reward": 2.7782450914382935,
+      "reward_std": 0.5766544118523598,
+      "rewards/accuracy_reward": 0.9062500149011612,
+      "rewards/reasoning_steps_reward": 0.9756944328546524,
+      "rewards/repetition_penalty_reward": -0.06984523870050907,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 373
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.63021850585938,
+      "epoch": 0.5609298837645295,
+      "grad_norm": 0.6707545410048498,
+      "kl": 0.36767578125,
+      "learning_rate": 9.606746757569107e-06,
+      "loss": 0.6822,
+      "reward": 2.6383379101753235,
+      "reward_std": 0.7204165309667587,
+      "rewards/accuracy_reward": 0.8281250149011612,
+      "rewards/reasoning_steps_reward": 0.9618055820465088,
+      "rewards/repetition_penalty_reward": -0.09560317918658257,
+      "rewards/tag_count_reward": 0.9440104216337204,
+      "step": 374
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 209.52605056762695,
+      "epoch": 0.562429696287964,
+      "grad_norm": 0.8820593625713022,
+      "kl": 0.42138671875,
+      "learning_rate": 9.554345686814601e-06,
+      "loss": 0.4017,
+      "reward": 2.5131508708000183,
+      "reward_std": 0.4276548661291599,
+      "rewards/accuracy_reward": 0.6302083507180214,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06410606950521469,
+      "rewards/tag_count_reward": 0.9609375298023224,
+      "step": 375
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 200.4479217529297,
+      "epoch": 0.5639295088113986,
+      "grad_norm": 0.6991131907883488,
+      "kl": 0.3740234375,
+      "learning_rate": 9.501956874726289e-06,
+      "loss": 0.4624,
+      "reward": 2.7490947246551514,
+      "reward_std": 0.4287390150129795,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.05255457712337375,
+      "rewards/tag_count_reward": 0.9648437798023224,
+      "step": 376
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.77605056762695,
+      "epoch": 0.5654293213348331,
+      "grad_norm": 1.725949820619147,
+      "kl": 0.44384765625,
+      "learning_rate": 9.449581762369454e-06,
+      "loss": 0.5851,
+      "reward": 2.348751664161682,
+      "reward_std": 0.5454598441720009,
+      "rewards/accuracy_reward": 0.494791679084301,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.07051914185285568,
+      "rewards/tag_count_reward": 0.9557291865348816,
+      "step": 377
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 191.84896850585938,
+      "epoch": 0.5669291338582677,
+      "grad_norm": 0.7636716518401006,
+      "kl": 0.38134765625,
+      "learning_rate": 9.39722179043254e-06,
+      "loss": 0.238,
+      "reward": 2.7479037046432495,
+      "reward_std": 0.232671735342592,
+      "rewards/accuracy_reward": 0.8177083432674408,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.04636712512001395,
+      "rewards/tag_count_reward": 0.9869791865348816,
+      "step": 378
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 190.34375381469727,
+      "epoch": 0.5684289463817023,
+      "grad_norm": 0.832309114343419,
+      "kl": 0.37158203125,
+      "learning_rate": 9.344878399187521e-06,
+      "loss": 0.2999,
+      "reward": 2.6157588362693787,
+      "reward_std": 0.41164325177669525,
+      "rewards/accuracy_reward": 0.7187500223517418,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.05872038099914789,
+      "rewards/tag_count_reward": 0.9765625149011612,
+      "step": 379
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 232.17187881469727,
+      "epoch": 0.5699287589051368,
+      "grad_norm": 1.2211358696196093,
+      "kl": 0.47216796875,
+      "learning_rate": 9.292553028450286e-06,
+      "loss": 0.5705,
+      "reward": 2.6755369901657104,
+      "reward_std": 0.49481815844774246,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9843750298023224,
+      "rewards/repetition_penalty_reward": -0.09529644902795553,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 380
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 185.0364646911621,
+      "epoch": 0.5714285714285714,
+      "grad_norm": 2.076589624203147,
+      "kl": 0.41943359375,
+      "learning_rate": 9.240247117541031e-06,
+      "loss": 0.3464,
+      "reward": 2.6886950731277466,
+      "reward_std": 0.3029524376615882,
+      "rewards/accuracy_reward": 0.760416679084301,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.040471707470715046,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 381
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 193.8385467529297,
+      "epoch": 0.572928383952006,
+      "grad_norm": 0.7452755017756123,
+      "kl": 0.33544921875,
+      "learning_rate": 9.187962105244667e-06,
+      "loss": 0.4188,
+      "reward": 2.848473310470581,
+      "reward_std": 0.38677336275577545,
+      "rewards/accuracy_reward": 0.9427083432674408,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.05083235865458846,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 382
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 213.88021087646484,
+      "epoch": 0.5744281964754405,
+      "grad_norm": 0.8123392096930626,
+      "kl": 0.357421875,
+      "learning_rate": 9.135699429771245e-06,
+      "loss": 0.4282,
+      "reward": 2.7217193841934204,
+      "reward_std": 0.32535652443766594,
+      "rewards/accuracy_reward": 0.8177083432674408,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.0608327342197299,
+      "rewards/tag_count_reward": 0.970052108168602,
+      "step": 383
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.7864646911621,
+      "epoch": 0.5759280089988752,
+      "grad_norm": 0.7045340047560704,
+      "kl": 0.28662109375,
+      "learning_rate": 9.083460528716396e-06,
+      "loss": 0.3565,
+      "reward": 2.539238929748535,
+      "reward_std": 0.3681572899222374,
+      "rewards/accuracy_reward": 0.6093750149011612,
+      "rewards/reasoning_steps_reward": 0.9982638955116272,
+      "rewards/repetition_penalty_reward": -0.05537911970168352,
+      "rewards/tag_count_reward": 0.9869791865348816,
+      "step": 384
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 229.43750381469727,
+      "epoch": 0.5774278215223098,
+      "grad_norm": 1.429073793360656,
+      "kl": 0.4013671875,
+      "learning_rate": 9.031246839021783e-06,
+      "loss": 0.5535,
+      "reward": 2.4284881353378296,
+      "reward_std": 0.5695413202047348,
+      "rewards/accuracy_reward": 0.557291679084301,
+      "rewards/reasoning_steps_reward": 0.9791667014360428,
+      "rewards/repetition_penalty_reward": -0.07541824877262115,
+      "rewards/tag_count_reward": 0.9674479365348816,
+      "step": 385
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.5416717529297,
+      "epoch": 0.5789276340457443,
+      "grad_norm": 0.7930980550389279,
+      "kl": 0.396484375,
+      "learning_rate": 8.979059796935578e-06,
+      "loss": 0.3136,
+      "reward": 2.754866659641266,
+      "reward_std": 0.33939819782972336,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.056331380270421505,
+      "rewards/tag_count_reward": 0.977864608168602,
+      "step": 386
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 220.7135467529297,
+      "epoch": 0.5804274465691789,
+      "grad_norm": 2.4297195424195097,
+      "kl": 0.35107421875,
+      "learning_rate": 8.926900837972953e-06,
+      "loss": 0.5168,
+      "reward": 2.7247214913368225,
+      "reward_std": 0.5589673742651939,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9791666567325592,
+      "rewards/repetition_penalty_reward": -0.0643411623314023,
+      "rewards/tag_count_reward": 0.9713541865348816,
+      "step": 387
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 220.0885467529297,
+      "epoch": 0.5819272590926134,
+      "grad_norm": 1.0129393824300377,
+      "kl": 0.34033203125,
+      "learning_rate": 8.874771396876597e-06,
+      "loss": 0.6594,
+      "reward": 2.5896897315979004,
+      "reward_std": 0.6142508238554001,
+      "rewards/accuracy_reward": 0.723958358168602,
+      "rewards/reasoning_steps_reward": 0.9791667014360428,
+      "rewards/repetition_penalty_reward": -0.07307065650820732,
+      "rewards/tag_count_reward": 0.9596354365348816,
+      "step": 388
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 245.2187614440918,
+      "epoch": 0.583427071616048,
+      "grad_norm": 0.6967111778457452,
+      "kl": 0.333984375,
+      "learning_rate": 8.822672907577244e-06,
+      "loss": 0.7214,
+      "reward": 2.621681034564972,
+      "reward_std": 0.8044816702604294,
+      "rewards/accuracy_reward": 0.8229166865348816,
+      "rewards/reasoning_steps_reward": 0.9548611044883728,
+      "rewards/repetition_penalty_reward": -0.09620102681219578,
+      "rewards/tag_count_reward": 0.9401041716337204,
+      "step": 389
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 198.38541793823242,
+      "epoch": 0.5849268841394826,
+      "grad_norm": 0.8231881380174958,
+      "kl": 0.31884765625,
+      "learning_rate": 8.770606803154235e-06,
+      "loss": 0.2053,
+      "reward": 2.5868722200393677,
+      "reward_std": 0.4082149714231491,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.9878472238779068,
+      "rewards/repetition_penalty_reward": -0.03899595024995506,
+      "rewards/tag_count_reward": 0.9817708432674408,
+      "step": 390
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 243.2395896911621,
+      "epoch": 0.5864266966629171,
+      "grad_norm": 0.7279318824033988,
+      "kl": 0.353515625,
+      "learning_rate": 8.718574515796099e-06,
+      "loss": 0.6324,
+      "reward": 2.612482249736786,
+      "reward_std": 0.6807690560817719,
+      "rewards/accuracy_reward": 0.8072916865348816,
+      "rewards/reasoning_steps_reward": 0.9548611491918564,
+      "rewards/repetition_penalty_reward": -0.09237884357571602,
+      "rewards/tag_count_reward": 0.942708358168602,
+      "step": 391
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 276.40625762939453,
+      "epoch": 0.5879265091863517,
+      "grad_norm": 1.6569386224229496,
+      "kl": 0.43408203125,
+      "learning_rate": 8.666577476761147e-06,
+      "loss": 0.9446,
+      "reward": 2.3660694360733032,
+      "reward_std": 1.0445173233747482,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.9288194477558136,
+      "rewards/repetition_penalty_reward": -0.1317605171352625,
+      "rewards/tag_count_reward": 0.9023437649011612,
+      "step": 392
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.8489685058594,
+      "epoch": 0.5894263217097863,
+      "grad_norm": 1.3890685202023243,
+      "kl": 0.42333984375,
+      "learning_rate": 8.61461711633812e-06,
+      "loss": 0.7734,
+      "reward": 2.335377037525177,
+      "reward_std": 1.022938460111618,
+      "rewards/accuracy_reward": 0.6875000298023224,
+      "rewards/reasoning_steps_reward": 0.925347238779068,
+      "rewards/repetition_penalty_reward": -0.16418905928730965,
+      "rewards/tag_count_reward": 0.8867187649011612,
+      "step": 393
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 332.44792556762695,
+      "epoch": 0.5909261342332208,
+      "grad_norm": 1.758344512665889,
+      "kl": 0.47607421875,
+      "learning_rate": 8.562694863806833e-06,
+      "loss": 0.7599,
+      "reward": 2.0866143703460693,
+      "reward_std": 1.0357710719108582,
+      "rewards/accuracy_reward": 0.5000000223517418,
+      "rewards/reasoning_steps_reward": 0.9027778208255768,
+      "rewards/repetition_penalty_reward": -0.19246553257107735,
+      "rewards/tag_count_reward": 0.8763020932674408,
+      "step": 394
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.70314025878906,
+      "epoch": 0.5924259467566554,
+      "grad_norm": 0.7342927358013376,
+      "kl": 0.3798828125,
+      "learning_rate": 8.510812147398857e-06,
+      "loss": 0.6993,
+      "reward": 2.433952748775482,
+      "reward_std": 0.7940848171710968,
+      "rewards/accuracy_reward": 0.661458358168602,
+      "rewards/reasoning_steps_reward": 0.9461806118488312,
+      "rewards/repetition_penalty_reward": -0.10988406464457512,
+      "rewards/tag_count_reward": 0.9361979216337204,
+      "step": 395
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 248.14063262939453,
+      "epoch": 0.59392575928009,
+      "grad_norm": 0.7272681746221773,
+      "kl": 0.40625,
+      "learning_rate": 8.458970394258244e-06,
+      "loss": 0.6893,
+      "reward": 2.5480750799179077,
+      "reward_std": 0.6472647786140442,
+      "rewards/accuracy_reward": 0.7343750149011612,
+      "rewards/reasoning_steps_reward": 0.960069477558136,
+      "rewards/repetition_penalty_reward": -0.09558827057480812,
+      "rewards/tag_count_reward": 0.9492187798023224,
+      "step": 396
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 280.07813262939453,
+      "epoch": 0.5954255718035245,
+      "grad_norm": 1.774304333205819,
+      "kl": 0.43408203125,
+      "learning_rate": 8.407171030402263e-06,
+      "loss": 0.9403,
+      "reward": 2.3916409015655518,
+      "reward_std": 0.9467423260211945,
+      "rewards/accuracy_reward": 0.6979167014360428,
+      "rewards/reasoning_steps_reward": 0.9218750149011612,
+      "rewards/repetition_penalty_reward": -0.14221329241991043,
+      "rewards/tag_count_reward": 0.9140625298023224,
+      "step": 397
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.25000762939453,
+      "epoch": 0.5969253843269592,
+      "grad_norm": 0.757311691683406,
+      "kl": 0.38232421875,
+      "learning_rate": 8.355415480682176e-06,
+      "loss": 0.5822,
+      "reward": 2.5999165177345276,
+      "reward_std": 0.6902973502874374,
+      "rewards/accuracy_reward": 0.7968750149011612,
+      "rewards/reasoning_steps_reward": 0.9670139402151108,
+      "rewards/repetition_penalty_reward": -0.10017046704888344,
+      "rewards/tag_count_reward": 0.9361979365348816,
+      "step": 398
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 244.0260467529297,
+      "epoch": 0.5984251968503937,
+      "grad_norm": 0.8003986775615329,
+      "kl": 0.32861328125,
+      "learning_rate": 8.303705168744042e-06,
+      "loss": 0.8742,
+      "reward": 2.4283345341682434,
+      "reward_std": 0.7666500955820084,
+      "rewards/accuracy_reward": 0.6302083432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.1033495794981718,
+      "rewards/tag_count_reward": 0.9361979365348816,
+      "step": 399
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 283.21355056762695,
+      "epoch": 0.5999250093738283,
+      "grad_norm": 1.0199487663866538,
+      "kl": 0.4130859375,
+      "learning_rate": 8.252041516989565e-06,
+      "loss": 0.9242,
+      "reward": 2.414180636405945,
+      "reward_std": 0.9542115926742554,
+      "rewards/accuracy_reward": 0.703125,
+      "rewards/reasoning_steps_reward": 0.9444444626569748,
+      "rewards/repetition_penalty_reward": -0.14614922180771828,
+      "rewards/tag_count_reward": 0.9127604216337204,
+      "step": 400
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 243.81771087646484,
+      "epoch": 0.6014248218972629,
+      "grad_norm": 1.0978785601748482,
+      "kl": 0.39208984375,
+      "learning_rate": 8.200425946536956e-06,
+      "loss": 0.8053,
+      "reward": 2.522300899028778,
+      "reward_std": 0.8708243519067764,
+      "rewards/accuracy_reward": 0.7604166865348816,
+      "rewards/reasoning_steps_reward": 0.9461805820465088,
+      "rewards/repetition_penalty_reward": -0.11137974262237549,
+      "rewards/tag_count_reward": 0.927083358168602,
+      "step": 401
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 244.26562881469727,
+      "epoch": 0.6029246344206974,
+      "grad_norm": 0.842070280135897,
+      "kl": 0.31103515625,
+      "learning_rate": 8.148859877181849e-06,
+      "loss": 0.7132,
+      "reward": 2.6524030566215515,
+      "reward_std": 0.7079126834869385,
+      "rewards/accuracy_reward": 0.848958358168602,
+      "rewards/reasoning_steps_reward": 0.9635417014360428,
+      "rewards/repetition_penalty_reward": -0.10540947876870632,
+      "rewards/tag_count_reward": 0.9453125149011612,
+      "step": 402
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 246.54687881469727,
+      "epoch": 0.604424446944132,
+      "grad_norm": 9.414923410713605,
+      "kl": 0.35693359375,
+      "learning_rate": 8.097344727358247e-06,
+      "loss": 0.6565,
+      "reward": 2.452816426753998,
+      "reward_std": 0.8280535340309143,
+      "rewards/accuracy_reward": 0.6770833507180214,
+      "rewards/reasoning_steps_reward": 0.9375000298023224,
+      "rewards/repetition_penalty_reward": -0.10447533056139946,
+      "rewards/tag_count_reward": 0.942708358168602,
+      "step": 403
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 212.87500381469727,
+      "epoch": 0.6059242594675666,
+      "grad_norm": 1.1724536798403964,
+      "kl": 0.390625,
+      "learning_rate": 8.045881914099503e-06,
+      "loss": 0.5066,
+      "reward": 2.6050479412078857,
+      "reward_std": 0.6001375466585159,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.973958358168602,
+      "rewards/repetition_penalty_reward": -0.06943127233535051,
+      "rewards/tag_count_reward": 0.9557291865348816,
+      "step": 404
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 255.69792938232422,
+      "epoch": 0.6074240719910011,
+      "grad_norm": 0.8089367863487846,
+      "kl": 0.3916015625,
+      "learning_rate": 7.99447285299934e-06,
+      "loss": 0.8552,
+      "reward": 2.4925881028175354,
+      "reward_std": 0.8843167871236801,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.942708358168602,
+      "rewards/repetition_penalty_reward": -0.12069321796298027,
+      "rewards/tag_count_reward": 0.9257812649011612,
+      "step": 405
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 245.58334732055664,
+      "epoch": 0.6089238845144357,
+      "grad_norm": 67.41761784039255,
+      "kl": 2.849609375,
+      "learning_rate": 7.943118958172917e-06,
+      "loss": 0.7802,
+      "reward": 2.5025484561920166,
+      "reward_std": 0.8601765781641006,
+      "rewards/accuracy_reward": 0.755208358168602,
+      "rewards/reasoning_steps_reward": 0.9340278208255768,
+      "rewards/repetition_penalty_reward": -0.11377106420695782,
+      "rewards/tag_count_reward": 0.927083358168602,
+      "step": 406
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 252.80209350585938,
+      "epoch": 0.6104236970378702,
+      "grad_norm": 50.58089501980667,
+      "kl": 3.66796875,
+      "learning_rate": 7.891821642217926e-06,
+      "loss": 0.8609,
+      "reward": 2.4164522886276245,
+      "reward_std": 0.693310096859932,
+      "rewards/accuracy_reward": 0.6354166716337204,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.1191379763185978,
+      "rewards/tag_count_reward": 0.9348958432674408,
+      "step": 407
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 284.53125381469727,
+      "epoch": 0.6119235095613048,
+      "grad_norm": 1.7854042705861606,
+      "kl": 0.47021484375,
+      "learning_rate": 7.840582316175737e-06,
+      "loss": 0.6802,
+      "reward": 2.4898502826690674,
+      "reward_std": 0.7629896551370621,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.958333358168602,
+      "rewards/repetition_penalty_reward": -0.13645190559327602,
+      "rewards/tag_count_reward": 0.9231770932674408,
+      "step": 408
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.7708435058594,
+      "epoch": 0.6134233220847394,
+      "grad_norm": 0.6181557051707668,
+      "kl": 0.3818359375,
+      "learning_rate": 7.789402389492582e-06,
+      "loss": 0.7549,
+      "reward": 2.4589642882347107,
+      "reward_std": 0.831140786409378,
+      "rewards/accuracy_reward": 0.6979166939854622,
+      "rewards/reasoning_steps_reward": 0.9548611342906952,
+      "rewards/repetition_penalty_reward": -0.13001148030161858,
+      "rewards/tag_count_reward": 0.9361979514360428,
+      "step": 409
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.5104293823242,
+      "epoch": 0.6149231346081739,
+      "grad_norm": 0.9617852079175642,
+      "kl": 0.36083984375,
+      "learning_rate": 7.738283269980798e-06,
+      "loss": 0.7348,
+      "reward": 2.555156171321869,
+      "reward_std": 0.8534037470817566,
+      "rewards/accuracy_reward": 0.8072916865348816,
+      "rewards/reasoning_steps_reward": 0.9444444477558136,
+      "rewards/repetition_penalty_reward": -0.11845514364540577,
+      "rewards/tag_count_reward": 0.9218750149011612,
+      "step": 410
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 219.22916793823242,
+      "epoch": 0.6164229471316085,
+      "grad_norm": 0.7921173872233128,
+      "kl": 0.29833984375,
+      "learning_rate": 7.687226363780084e-06,
+      "loss": 0.5226,
+      "reward": 2.4889369010925293,
+      "reward_std": 0.5284189339727163,
+      "rewards/accuracy_reward": 0.6406250074505806,
+      "rewards/reasoning_steps_reward": 0.9722222238779068,
+      "rewards/repetition_penalty_reward": -0.08094170223921537,
+      "rewards/tag_count_reward": 0.9570312649011612,
+      "step": 411
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 205.95312881469727,
+      "epoch": 0.6179227596550432,
+      "grad_norm": 0.6969647943689223,
+      "kl": 0.3193359375,
+      "learning_rate": 7.636233075318824e-06,
+      "loss": 0.3959,
+      "reward": 2.670749843120575,
+      "reward_std": 0.4937814176082611,
+      "rewards/accuracy_reward": 0.7968750298023224,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.06666341191157699,
+      "rewards/tag_count_reward": 0.9648437798023224,
+      "step": 412
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 234.8854217529297,
+      "epoch": 0.6194225721784777,
+      "grad_norm": 0.9557907221459236,
+      "kl": 0.310546875,
+      "learning_rate": 7.585304807275473e-06,
+      "loss": 0.571,
+      "reward": 2.531363546848297,
+      "reward_std": 0.6936883656308055,
+      "rewards/accuracy_reward": 0.7343750149011612,
+      "rewards/reasoning_steps_reward": 0.9427083283662796,
+      "rewards/repetition_penalty_reward": -0.0910323103889823,
+      "rewards/tag_count_reward": 0.9453125298023224,
+      "step": 413
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 199.11458587646484,
+      "epoch": 0.6209223847019123,
+      "grad_norm": 0.8453338172094219,
+      "kl": 0.28857421875,
+      "learning_rate": 7.534442960539956e-06,
+      "loss": 0.3905,
+      "reward": 2.7355194687843323,
+      "reward_std": 0.49409135431051254,
+      "rewards/accuracy_reward": 0.8385416716337204,
+      "rewards/reasoning_steps_reward": 0.9809028059244156,
+      "rewards/repetition_penalty_reward": -0.05527916317805648,
+      "rewards/tag_count_reward": 0.9713542014360428,
+      "step": 414
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 182.48437881469727,
+      "epoch": 0.6224221972253469,
+      "grad_norm": 0.7857191402885518,
+      "kl": 0.303955078125,
+      "learning_rate": 7.483648934175138e-06,
+      "loss": 0.1902,
+      "reward": 2.6866626739501953,
+      "reward_std": 0.28315746411681175,
+      "rewards/accuracy_reward": 0.7395833432674408,
+      "rewards/reasoning_steps_reward": 1.0,
+      "rewards/repetition_penalty_reward": -0.042504156939685345,
+      "rewards/tag_count_reward": 0.9895833432674408,
+      "step": 415
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 192.37500762939453,
+      "epoch": 0.6239220097487814,
+      "grad_norm": 1.0846986274613384,
+      "kl": 0.35107421875,
+      "learning_rate": 7.432924125378345e-06,
+      "loss": 0.3789,
+      "reward": 2.4783818125724792,
+      "reward_std": 0.5089812204241753,
+      "rewards/accuracy_reward": 0.5937500223517418,
+      "rewards/reasoning_steps_reward": 0.9809028208255768,
+      "rewards/repetition_penalty_reward": -0.0624168599024415,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 416
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 238.28125762939453,
+      "epoch": 0.625421822272216,
+      "grad_norm": 0.6289618055873664,
+      "kl": 0.32958984375,
+      "learning_rate": 7.382269929442925e-06,
+      "loss": 0.6525,
+      "reward": 2.729068160057068,
+      "reward_std": 0.6565304845571518,
+      "rewards/accuracy_reward": 0.9062500149011612,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.09384858049452305,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 417
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 229.7187614440918,
+      "epoch": 0.6269216347956506,
+      "grad_norm": 2.2511596836127628,
+      "kl": 0.37158203125,
+      "learning_rate": 7.331687739719868e-06,
+      "loss": 0.6599,
+      "reward": 2.739536941051483,
+      "reward_std": 0.7113066837191582,
+      "rewards/accuracy_reward": 0.9062500149011612,
+      "rewards/reasoning_steps_reward": 0.9652778059244156,
+      "rewards/repetition_penalty_reward": -0.08251171908341348,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 418
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.0260467529297,
+      "epoch": 0.6284214473190851,
+      "grad_norm": 0.7481899237512065,
+      "kl": 0.3017578125,
+      "learning_rate": 7.281178947579484e-06,
+      "loss": 0.4911,
+      "reward": 2.677112579345703,
+      "reward_std": 0.500836968421936,
+      "rewards/accuracy_reward": 0.802083358168602,
+      "rewards/reasoning_steps_reward": 0.9843750447034836,
+      "rewards/repetition_penalty_reward": -0.07549162488430738,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 419
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.0572967529297,
+      "epoch": 0.6299212598425197,
+      "grad_norm": 2.4166560990937582,
+      "kl": 0.6650390625,
+      "learning_rate": 7.230744942373125e-06,
+      "loss": 0.5609,
+      "reward": 2.6925852298736572,
+      "reward_std": 0.5387180671095848,
+      "rewards/accuracy_reward": 0.8333333432674408,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.08041842561215162,
+      "rewards/tag_count_reward": 0.9570312649011612,
+      "step": 420
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 204.2083396911621,
+      "epoch": 0.6314210723659542,
+      "grad_norm": 3.5367701148683515,
+      "kl": 0.3935546875,
+      "learning_rate": 7.1803871113949675e-06,
+      "loss": 0.5299,
+      "reward": 2.5729278326034546,
+      "reward_std": 0.5818023979663849,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.07290551625192165,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 421
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 184.77083587646484,
+      "epoch": 0.6329208848893888,
+      "grad_norm": 0.81390385216287,
+      "kl": 0.33544921875,
+      "learning_rate": 7.13010683984386e-06,
+      "loss": 0.4888,
+      "reward": 2.656929612159729,
+      "reward_std": 0.4153623729944229,
+      "rewards/accuracy_reward": 0.739583358168602,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.04923371039330959,
+      "rewards/tag_count_reward": 0.9804687649011612,
+      "step": 422
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 219.2291717529297,
+      "epoch": 0.6344206974128234,
+      "grad_norm": 0.7193436114614246,
+      "kl": 0.322998046875,
+      "learning_rate": 7.07990551078521e-06,
+      "loss": 0.5418,
+      "reward": 2.480729579925537,
+      "reward_std": 0.5471135228872299,
+      "rewards/accuracy_reward": 0.614583358168602,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.07048568688333035,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 423
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.1458396911621,
+      "epoch": 0.6359205099362579,
+      "grad_norm": 1.2355026811447927,
+      "kl": 0.39794921875,
+      "learning_rate": 7.029784505112948e-06,
+      "loss": 0.6063,
+      "reward": 2.551844298839569,
+      "reward_std": 0.5733724534511566,
+      "rewards/accuracy_reward": 0.6979166865348816,
+      "rewards/reasoning_steps_reward": 0.9687500298023224,
+      "rewards/repetition_penalty_reward": -0.07836401462554932,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 424
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 185.15625762939453,
+      "epoch": 0.6374203224596925,
+      "grad_norm": 0.7168826565129065,
+      "kl": 0.306640625,
+      "learning_rate": 6.979745201511531e-06,
+      "loss": 0.3983,
+      "reward": 2.7733737230300903,
+      "reward_std": 0.3679837482050061,
+      "rewards/accuracy_reward": 0.8750000149011612,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0534491918515414,
+      "rewards/tag_count_reward": 0.97265625,
+      "step": 425
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 189.7760467529297,
+      "epoch": 0.6389201349831272,
+      "grad_norm": 1.2687322158065155,
+      "kl": 0.318359375,
+      "learning_rate": 6.929788976418044e-06,
+      "loss": 0.6424,
+      "reward": 2.7650052905082703,
+      "reward_std": 0.5291210561990738,
+      "rewards/accuracy_reward": 0.8750000298023224,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.053571161814033985,
+      "rewards/tag_count_reward": 0.966145858168602,
+      "step": 426
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.81250762939453,
+      "epoch": 0.6404199475065617,
+      "grad_norm": 1.6352357141251739,
+      "kl": 0.33349609375,
+      "learning_rate": 6.879917203984306e-06,
+      "loss": 0.6536,
+      "reward": 2.448298752307892,
+      "reward_std": 0.5732715502381325,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.07210064399987459,
+      "rewards/tag_count_reward": 0.9648437649011612,
+      "step": 427
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 213.97917556762695,
+      "epoch": 0.6419197600299963,
+      "grad_norm": 1.886886621974555,
+      "kl": 0.58642578125,
+      "learning_rate": 6.830131256039094e-06,
+      "loss": 0.5201,
+      "reward": 2.543819785118103,
+      "reward_std": 0.5805186182260513,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.07814204692840576,
+      "rewards/tag_count_reward": 0.9622395932674408,
+      "step": 428
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.07813262939453,
+      "epoch": 0.6434195725534309,
+      "grad_norm": 1.0666842338328122,
+      "kl": 0.36376953125,
+      "learning_rate": 6.7804325020504e-06,
+      "loss": 0.6236,
+      "reward": 2.523444414138794,
+      "reward_std": 0.4915116261690855,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.07421195041388273,
+      "rewards/tag_count_reward": 0.962239608168602,
+      "step": 429
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 211.85937881469727,
+      "epoch": 0.6449193850768654,
+      "grad_norm": 1.060449887686782,
+      "kl": 0.2978515625,
+      "learning_rate": 6.730822309087756e-06,
+      "loss": 0.5033,
+      "reward": 2.6772631406784058,
+      "reward_std": 0.48493412137031555,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.06666056625545025,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 430
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 259.1614646911621,
+      "epoch": 0.6464191976003,
+      "grad_norm": 1.3061723100608225,
+      "kl": 0.361328125,
+      "learning_rate": 6.6813020417846456e-06,
+      "loss": 0.7871,
+      "reward": 2.4965052604675293,
+      "reward_std": 0.7866370305418968,
+      "rewards/accuracy_reward": 0.723958358168602,
+      "rewards/reasoning_steps_reward": 0.9618055522441864,
+      "rewards/repetition_penalty_reward": -0.12024837546050549,
+      "rewards/tag_count_reward": 0.930989608168602,
+      "step": 431
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 285.40105056762695,
+      "epoch": 0.6479190101237345,
+      "grad_norm": 0.7888584050372874,
+      "kl": 0.3642578125,
+      "learning_rate": 6.6318730623009465e-06,
+      "loss": 0.8798,
+      "reward": 2.5130550861358643,
+      "reward_std": 1.0070786774158478,
+      "rewards/accuracy_reward": 0.8229166865348816,
+      "rewards/reasoning_steps_reward": 0.942708358168602,
+      "rewards/repetition_penalty_reward": -0.14840331301093102,
+      "rewards/tag_count_reward": 0.8958333432674408,
+      "step": 432
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.69272232055664,
+      "epoch": 0.6494188226471691,
+      "grad_norm": 2.6122932218321706,
+      "kl": 0.51123046875,
+      "learning_rate": 6.582536730285476e-06,
+      "loss": 0.847,
+      "reward": 2.5346211194992065,
+      "reward_std": 0.9135303720831871,
+      "rewards/accuracy_reward": 0.817708358168602,
+      "rewards/reasoning_steps_reward": 0.9548611491918564,
+      "rewards/repetition_penalty_reward": -0.1350838476791978,
+      "rewards/tag_count_reward": 0.8971354365348816,
+      "step": 433
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.45313262939453,
+      "epoch": 0.6509186351706037,
+      "grad_norm": 0.8518186687660756,
+      "kl": 0.46337890625,
+      "learning_rate": 6.5332944028385885e-06,
+      "loss": 0.5872,
+      "reward": 2.459271728992462,
+      "reward_std": 0.7705230340361595,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.9479167014360428,
+      "rewards/repetition_penalty_reward": -0.14489501249045134,
+      "rewards/tag_count_reward": 0.9114583432674408,
+      "step": 434
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 301.73438262939453,
+      "epoch": 0.6524184476940382,
+      "grad_norm": 0.9613160246102254,
+      "kl": 0.443359375,
+      "learning_rate": 6.484147434474837e-06,
+      "loss": 0.9578,
+      "reward": 2.159851849079132,
+      "reward_std": 0.9833470582962036,
+      "rewards/accuracy_reward": 0.520833358168602,
+      "rewards/reasoning_steps_reward": 0.942708358168602,
+      "rewards/repetition_penalty_reward": -0.16957524791359901,
+      "rewards/tag_count_reward": 0.8658854365348816,
+      "step": 435
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 305.9010467529297,
+      "epoch": 0.6539182602174728,
+      "grad_norm": 0.8755106034657409,
+      "kl": 0.4609375,
+      "learning_rate": 6.435097177085728e-06,
+      "loss": 0.8985,
+      "reward": 2.2428812980651855,
+      "reward_std": 0.9718329310417175,
+      "rewards/accuracy_reward": 0.6041666865348816,
+      "rewards/reasoning_steps_reward": 0.9444445073604584,
+      "rewards/repetition_penalty_reward": -0.1755215786397457,
+      "rewards/tag_count_reward": 0.8697916865348816,
+      "step": 436
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 252.41667556762695,
+      "epoch": 0.6554180727409074,
+      "grad_norm": 0.7422509830131178,
+      "kl": 0.373046875,
+      "learning_rate": 6.386144979902527e-06,
+      "loss": 0.703,
+      "reward": 2.3507115840911865,
+      "reward_std": 0.7452640384435654,
+      "rewards/accuracy_reward": 0.598958358168602,
+      "rewards/reasoning_steps_reward": 0.9548611491918564,
+      "rewards/repetition_penalty_reward": -0.11196202971041203,
+      "rewards/tag_count_reward": 0.9088541865348816,
+      "step": 437
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 256.5208435058594,
+      "epoch": 0.6569178852643419,
+      "grad_norm": 0.7991022183328756,
+      "kl": 0.361328125,
+      "learning_rate": 6.337292189459139e-06,
+      "loss": 0.7431,
+      "reward": 2.456838309764862,
+      "reward_std": 0.9040014296770096,
+      "rewards/accuracy_reward": 0.7239583432674408,
+      "rewards/reasoning_steps_reward": 0.9496528059244156,
+      "rewards/repetition_penalty_reward": -0.12041868269443512,
+      "rewards/tag_count_reward": 0.9036458432674408,
+      "step": 438
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 231.63542938232422,
+      "epoch": 0.6584176977877765,
+      "grad_norm": 1.3902018588463414,
+      "kl": 0.35986328125,
+      "learning_rate": 6.2885401495550826e-06,
+      "loss": 0.7122,
+      "reward": 2.5194268226623535,
+      "reward_std": 0.7179795950651169,
+      "rewards/accuracy_reward": 0.729166679084301,
+      "rewards/reasoning_steps_reward": 0.9670138955116272,
+      "rewards/repetition_penalty_reward": -0.09732666984200478,
+      "rewards/tag_count_reward": 0.9205729365348816,
+      "step": 439
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.578125,
+      "epoch": 0.6599175103112112,
+      "grad_norm": 0.9152954380082454,
+      "kl": 0.36083984375,
+      "learning_rate": 6.239890201218517e-06,
+      "loss": 0.492,
+      "reward": 2.469900608062744,
+      "reward_std": 0.6492541283369064,
+      "rewards/accuracy_reward": 0.651041679084301,
+      "rewards/reasoning_steps_reward": 0.9600694477558136,
+      "rewards/repetition_penalty_reward": -0.07220023218542337,
+      "rewards/tag_count_reward": 0.9309895932674408,
+      "step": 440
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.1458396911621,
+      "epoch": 0.6614173228346457,
+      "grad_norm": 1.5238075521377912,
+      "kl": 0.3486328125,
+      "learning_rate": 6.191343682669357e-06,
+      "loss": 0.6152,
+      "reward": 2.59203839302063,
+      "reward_std": 0.5786980837583542,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.08374295756220818,
+      "rewards/tag_count_reward": 0.9414062649011612,
+      "step": 441
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.8854217529297,
+      "epoch": 0.6629171353580803,
+      "grad_norm": 1.014568496944409,
+      "kl": 0.3203125,
+      "learning_rate": 6.142901929282459e-06,
+      "loss": 0.5185,
+      "reward": 2.4907559752464294,
+      "reward_std": 0.527790479362011,
+      "rewards/accuracy_reward": 0.6093750074505806,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.06783778499811888,
+      "rewards/tag_count_reward": 0.9648437649011612,
+      "step": 442
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 213.52083587646484,
+      "epoch": 0.6644169478815148,
+      "grad_norm": 0.8930401982873789,
+      "kl": 0.3203125,
+      "learning_rate": 6.094566273550899e-06,
+      "loss": 0.554,
+      "reward": 2.7469860911369324,
+      "reward_std": 0.5041925981640816,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9930555671453476,
+      "rewards/repetition_penalty_reward": -0.06507986783981323,
+      "rewards/tag_count_reward": 0.9648437798023224,
+      "step": 443
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 223.73438262939453,
+      "epoch": 0.6659167604049494,
+      "grad_norm": 1.588063897389527,
+      "kl": 0.3759765625,
+      "learning_rate": 6.046338045049307e-06,
+      "loss": 0.3918,
+      "reward": 2.5898342728614807,
+      "reward_std": 0.5453440099954605,
+      "rewards/accuracy_reward": 0.723958358168602,
+      "rewards/reasoning_steps_reward": 0.9809027910232544,
+      "rewards/repetition_penalty_reward": -0.07075604610145092,
+      "rewards/tag_count_reward": 0.9557291865348816,
+      "step": 444
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 211.0677146911621,
+      "epoch": 0.667416572928384,
+      "grad_norm": 0.7683663208645458,
+      "kl": 0.33544921875,
+      "learning_rate": 5.998218570397298e-06,
+      "loss": 0.5318,
+      "reward": 2.6227740049362183,
+      "reward_std": 0.56145179271698,
+      "rewards/accuracy_reward": 0.7604166865348816,
+      "rewards/reasoning_steps_reward": 0.970486119389534,
+      "rewards/repetition_penalty_reward": -0.06646222807466984,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 445
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.59375381469727,
+      "epoch": 0.6689163854518185,
+      "grad_norm": 2.006829340741143,
+      "kl": 0.36865234375,
+      "learning_rate": 5.950209173222985e-06,
+      "loss": 0.4647,
+      "reward": 2.7126694917678833,
+      "reward_std": 0.5051566585898399,
+      "rewards/accuracy_reward": 0.8281250298023224,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.07031677477061749,
+      "rewards/tag_count_reward": 0.9687500149011612,
+      "step": 446
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 233.4062614440918,
+      "epoch": 0.6704161979752531,
+      "grad_norm": 0.8834971529763043,
+      "kl": 0.36572265625,
+      "learning_rate": 5.902311174126565e-06,
+      "loss": 0.4519,
+      "reward": 2.7030810117721558,
+      "reward_std": 0.49621870182454586,
+      "rewards/accuracy_reward": 0.8645833432674408,
+      "rewards/reasoning_steps_reward": 0.967013880610466,
+      "rewards/repetition_penalty_reward": -0.08033910719677806,
+      "rewards/tag_count_reward": 0.9518229216337204,
+      "step": 447
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 219.4427146911621,
+      "epoch": 0.6719160104986877,
+      "grad_norm": 0.98109027161852,
+      "kl": 0.359375,
+      "learning_rate": 5.854525890643996e-06,
+      "loss": 0.3661,
+      "reward": 2.567684829235077,
+      "reward_std": 0.5093120224773884,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9756944477558136,
+      "rewards/repetition_penalty_reward": -0.060353430919349194,
+      "rewards/tag_count_reward": 0.9648437649011612,
+      "step": 448
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 203.7916717529297,
+      "epoch": 0.6734158230221222,
+      "grad_norm": 0.8311384203525323,
+      "kl": 0.34130859375,
+      "learning_rate": 5.806854637210752e-06,
+      "loss": 0.4174,
+      "reward": 2.7751079201698303,
+      "reward_std": 0.3471484985202551,
+      "rewards/accuracy_reward": 0.8750000149011612,
+      "rewards/reasoning_steps_reward": 0.9826389253139496,
+      "rewards/repetition_penalty_reward": -0.0551872905343771,
+      "rewards/tag_count_reward": 0.9726562798023224,
+      "step": 449
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 197.1354217529297,
+      "epoch": 0.6749156355455568,
+      "grad_norm": 0.8629385750603632,
+      "kl": 0.341796875,
+      "learning_rate": 5.759298725125671e-06,
+      "loss": 0.2358,
+      "reward": 2.6736323833465576,
+      "reward_std": 0.344327449798584,
+      "rewards/accuracy_reward": 0.755208358168602,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.04251353442668915,
+      "rewards/tag_count_reward": 0.9765625298023224,
+      "step": 450
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 187.3958396911621,
+      "epoch": 0.6764154480689913,
+      "grad_norm": 0.9794837217183727,
+      "kl": 0.3740234375,
+      "learning_rate": 5.711859462514883e-06,
+      "loss": 0.4027,
+      "reward": 2.6471253037452698,
+      "reward_std": 0.46060725301504135,
+      "rewards/accuracy_reward": 0.7500000149011612,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.049923318438231945,
+      "rewards/tag_count_reward": 0.9609375298023224,
+      "step": 451
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 224.30730056762695,
+      "epoch": 0.6779152605924259,
+      "grad_norm": 1.191297396076842,
+      "kl": 0.4599609375,
+      "learning_rate": 5.664538154295827e-06,
+      "loss": 0.7305,
+      "reward": 2.573324501514435,
+      "reward_std": 0.6858862191438675,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.970486119389534,
+      "rewards/repetition_penalty_reward": -0.0768492091447115,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 452
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.9322967529297,
+      "epoch": 0.6794150731158605,
+      "grad_norm": 1.4404111173702936,
+      "kl": 0.43310546875,
+      "learning_rate": 5.617336102141356e-06,
+      "loss": 0.5628,
+      "reward": 2.5157384872436523,
+      "reward_std": 0.43997257202863693,
+      "rewards/accuracy_reward": 0.630208358168602,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06542493868619204,
+      "rewards/tag_count_reward": 0.9648437649011612,
+      "step": 453
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.62500381469727,
+      "epoch": 0.6809148856392951,
+      "grad_norm": 1.959555679611255,
+      "kl": 0.46826171875,
+      "learning_rate": 5.570254604443929e-06,
+      "loss": 0.7598,
+      "reward": 2.588327646255493,
+      "reward_std": 0.7168947905302048,
+      "rewards/accuracy_reward": 0.765625,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.08484961278736591,
+      "rewards/tag_count_reward": 0.9440104365348816,
+      "step": 454
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 232.7395896911621,
+      "epoch": 0.6824146981627297,
+      "grad_norm": 148.69573978924745,
+      "kl": 3.85595703125,
+      "learning_rate": 5.5232949562799055e-06,
+      "loss": 1.3481,
+      "reward": 2.4738574028015137,
+      "reward_std": 0.779233306646347,
+      "rewards/accuracy_reward": 0.6770833535119891,
+      "rewards/reasoning_steps_reward": 0.9618055522441864,
+      "rewards/repetition_penalty_reward": -0.09862528461962938,
+      "rewards/tag_count_reward": 0.9335937798023224,
+      "step": 455
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.4583396911621,
+      "epoch": 0.6839145106861643,
+      "grad_norm": 1.9450934585599418,
+      "kl": 0.41796875,
+      "learning_rate": 5.4764584493739095e-06,
+      "loss": 0.7255,
+      "reward": 2.364755541086197,
+      "reward_std": 0.6270653009414673,
+      "rewards/accuracy_reward": 0.5312500149011612,
+      "rewards/reasoning_steps_reward": 0.9809027761220932,
+      "rewards/repetition_penalty_reward": -0.09010571241378784,
+      "rewards/tag_count_reward": 0.942708358168602,
+      "step": 456
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.8697967529297,
+      "epoch": 0.6854143232095988,
+      "grad_norm": 1.9127619691878879,
+      "kl": 0.4794921875,
+      "learning_rate": 5.429746372063309e-06,
+      "loss": 0.8757,
+      "reward": 2.4884825348854065,
+      "reward_std": 0.784279853105545,
+      "rewards/accuracy_reward": 0.6770833432674408,
+      "rewards/reasoning_steps_reward": 0.963541716337204,
+      "rewards/repetition_penalty_reward": -0.09224662370979786,
+      "rewards/tag_count_reward": 0.9401041865348816,
+      "step": 457
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.8489646911621,
+      "epoch": 0.6869141357330334,
+      "grad_norm": 1.2094457444337063,
+      "kl": 0.4833984375,
+      "learning_rate": 5.3831600092627704e-06,
+      "loss": 0.714,
+      "reward": 2.73754620552063,
+      "reward_std": 0.6269456818699837,
+      "rewards/accuracy_reward": 0.895833358168602,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.08667258732020855,
+      "rewards/tag_count_reward": 0.9492187649011612,
+      "step": 458
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 195.3385467529297,
+      "epoch": 0.688413948256468,
+      "grad_norm": 1.970689168479106,
+      "kl": 0.4453125,
+      "learning_rate": 5.336700642428913e-06,
+      "loss": 0.5537,
+      "reward": 2.5139536261558533,
+      "reward_std": 0.5696927979588509,
+      "rewards/accuracy_reward": 0.6458333358168602,
+      "rewards/reasoning_steps_reward": 0.9722222536802292,
+      "rewards/repetition_penalty_reward": -0.06634163623675704,
+      "rewards/tag_count_reward": 0.9622395932674408,
+      "step": 459
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 254.3125114440918,
+      "epoch": 0.6899137607799025,
+      "grad_norm": 0.7625934263204471,
+      "kl": 0.36962890625,
+      "learning_rate": 5.290369549525066e-06,
+      "loss": 0.6621,
+      "reward": 2.52715528011322,
+      "reward_std": 0.7988520860671997,
+      "rewards/accuracy_reward": 0.7760416865348816,
+      "rewards/reasoning_steps_reward": 0.9461805820465088,
+      "rewards/repetition_penalty_reward": -0.12345251068472862,
+      "rewards/tag_count_reward": 0.9283854365348816,
+      "step": 460
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 195.43750762939453,
+      "epoch": 0.6914135733033371,
+      "grad_norm": 0.989857600799216,
+      "kl": 0.39501953125,
+      "learning_rate": 5.2441680049861125e-06,
+      "loss": 0.6605,
+      "reward": 2.4125567078590393,
+      "reward_std": 0.6474236696958542,
+      "rewards/accuracy_reward": 0.5677083432674408,
+      "rewards/reasoning_steps_reward": 0.9670139253139496,
+      "rewards/repetition_penalty_reward": -0.0713843759149313,
+      "rewards/tag_count_reward": 0.9492187649011612,
+      "step": 461
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.6875114440918,
+      "epoch": 0.6929133858267716,
+      "grad_norm": 0.8986450074706076,
+      "kl": 0.3359375,
+      "learning_rate": 5.198097279683434e-06,
+      "loss": 0.604,
+      "reward": 2.4094382524490356,
+      "reward_std": 0.6442625038325787,
+      "rewards/accuracy_reward": 0.614583358168602,
+      "rewards/reasoning_steps_reward": 0.9513889104127884,
+      "rewards/repetition_penalty_reward": -0.09273206302896142,
+      "rewards/tag_count_reward": 0.9361979365348816,
+      "step": 462
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 200.44271087646484,
+      "epoch": 0.6944131983502062,
+      "grad_norm": 606.4249596937021,
+      "kl": 24.65576171875,
+      "learning_rate": 5.152158640889947e-06,
+      "loss": 1.5527,
+      "reward": 2.6447129249572754,
+      "reward_std": 0.4999554455280304,
+      "rewards/accuracy_reward": 0.770833358168602,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.06882883794605732,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 463
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 219.21875762939453,
+      "epoch": 0.6959130108736408,
+      "grad_norm": 1.5220309197508348,
+      "kl": 0.36181640625,
+      "learning_rate": 5.106353352245254e-06,
+      "loss": 0.5723,
+      "reward": 2.6008208990097046,
+      "reward_std": 0.5897117927670479,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.08667923882603645,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 464
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 181.75521087646484,
+      "epoch": 0.6974128233970753,
+      "grad_norm": 1.3974522149419788,
+      "kl": 0.3447265625,
+      "learning_rate": 5.060682673720878e-06,
+      "loss": 0.5287,
+      "reward": 2.4272356927394867,
+      "reward_std": 0.3737209364771843,
+      "rewards/accuracy_reward": 0.5156250251457095,
+      "rewards/reasoning_steps_reward": 0.9878472536802292,
+      "rewards/repetition_penalty_reward": -0.05019487999379635,
+      "rewards/tag_count_reward": 0.973958358168602,
+      "step": 465
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 181.8541717529297,
+      "epoch": 0.6989126359205099,
+      "grad_norm": 1.6330449051409097,
+      "kl": 0.28955078125,
+      "learning_rate": 5.015147861585603e-06,
+      "loss": 0.5821,
+      "reward": 2.5756843090057373,
+      "reward_std": 0.48887188360095024,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.05495820567011833,
+      "rewards/tag_count_reward": 0.9674479216337204,
+      "step": 466
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.62500381469727,
+      "epoch": 0.7004124484439445,
+      "grad_norm": 1.0551997003189855,
+      "kl": 0.2900390625,
+      "learning_rate": 4.969750168370924e-06,
+      "loss": 0.6587,
+      "reward": 2.587502121925354,
+      "reward_std": 0.6842982918024063,
+      "rewards/accuracy_reward": 0.7812500149011612,
+      "rewards/reasoning_steps_reward": 0.9565972238779068,
+      "rewards/repetition_penalty_reward": -0.09175141900777817,
+      "rewards/tag_count_reward": 0.9414062649011612,
+      "step": 467
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.6979217529297,
+      "epoch": 0.7019122609673791,
+      "grad_norm": 1.172638919017115,
+      "kl": 0.3447265625,
+      "learning_rate": 4.924490842836584e-06,
+      "loss": 0.6767,
+      "reward": 2.4782765209674835,
+      "reward_std": 0.6802399158477783,
+      "rewards/accuracy_reward": 0.6510416939854622,
+      "rewards/reasoning_steps_reward": 0.96875,
+      "rewards/repetition_penalty_reward": -0.08552560955286026,
+      "rewards/tag_count_reward": 0.9440104365348816,
+      "step": 468
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 246.8020896911621,
+      "epoch": 0.7034120734908137,
+      "grad_norm": 1.208509544801629,
+      "kl": 0.30029296875,
+      "learning_rate": 4.879371129936233e-06,
+      "loss": 0.7641,
+      "reward": 2.485842287540436,
+      "reward_std": 0.7981886714696884,
+      "rewards/accuracy_reward": 0.6979166865348816,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.11224810406565666,
+      "rewards/tag_count_reward": 0.9348958432674408,
+      "step": 469
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.69271087646484,
+      "epoch": 0.7049118860142483,
+      "grad_norm": 1.2784121913501234,
+      "kl": 0.337890625,
+      "learning_rate": 4.834392270783183e-06,
+      "loss": 0.8845,
+      "reward": 2.6402639150619507,
+      "reward_std": 0.8152274489402771,
+      "rewards/accuracy_reward": 0.8437500298023224,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.09411128051578999,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 470
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.55209350585938,
+      "epoch": 0.7064116985376828,
+      "grad_norm": 0.9623091877361314,
+      "kl": 0.330078125,
+      "learning_rate": 4.789555502616258e-06,
+      "loss": 0.7554,
+      "reward": 2.4469590187072754,
+      "reward_std": 0.7040076702833176,
+      "rewards/accuracy_reward": 0.6458333507180214,
+      "rewards/reasoning_steps_reward": 0.9565972238779068,
+      "rewards/repetition_penalty_reward": -0.09427369153127074,
+      "rewards/tag_count_reward": 0.9388020932674408,
+      "step": 471
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 178.8333396911621,
+      "epoch": 0.7079115110611174,
+      "grad_norm": 1.0260244805761438,
+      "kl": 0.296875,
+      "learning_rate": 4.744862058765776e-06,
+      "loss": 0.2163,
+      "reward": 2.727869689464569,
+      "reward_std": 0.36412858217954636,
+      "rewards/accuracy_reward": 0.8072916865348816,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.036019228398799896,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 472
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 226.10937881469727,
+      "epoch": 0.709411323584552,
+      "grad_norm": 0.9517703362898857,
+      "kl": 0.3935546875,
+      "learning_rate": 4.700313168619608e-06,
+      "loss": 0.8129,
+      "reward": 2.543964922428131,
+      "reward_std": 0.6564487293362617,
+      "rewards/accuracy_reward": 0.7500000149011612,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.1053406372666359,
+      "rewards/tag_count_reward": 0.9322916716337204,
+      "step": 473
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 237.50521087646484,
+      "epoch": 0.7109111361079865,
+      "grad_norm": 1.006357822901466,
+      "kl": 0.4013671875,
+      "learning_rate": 4.655910057589377e-06,
+      "loss": 0.9248,
+      "reward": 2.5268329977989197,
+      "reward_std": 0.9078202545642853,
+      "rewards/accuracy_reward": 0.7656250149011612,
+      "rewards/reasoning_steps_reward": 0.9374999850988388,
+      "rewards/repetition_penalty_reward": -0.10207330994307995,
+      "rewards/tag_count_reward": 0.9257812649011612,
+      "step": 474
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.26562881469727,
+      "epoch": 0.7124109486314211,
+      "grad_norm": 1.4034228060868457,
+      "kl": 0.416015625,
+      "learning_rate": 4.611653947076732e-06,
+      "loss": 0.7093,
+      "reward": 2.54375296831131,
+      "reward_std": 0.6568827331066132,
+      "rewards/accuracy_reward": 0.723958358168602,
+      "rewards/reasoning_steps_reward": 0.9687500447034836,
+      "rewards/repetition_penalty_reward": -0.09166385605931282,
+      "rewards/tag_count_reward": 0.942708358168602,
+      "step": 475
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 221.40625762939453,
+      "epoch": 0.7139107611548556,
+      "grad_norm": 2.3545088822080387,
+      "kl": 0.49462890625,
+      "learning_rate": 4.567546054439777e-06,
+      "loss": 0.7007,
+      "reward": 2.5734556913375854,
+      "reward_std": 0.699681967496872,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9670138657093048,
+      "rewards/repetition_penalty_reward": -0.08626659773290157,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 476
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 227.00000762939453,
+      "epoch": 0.7154105736782902,
+      "grad_norm": 1.5468969267610364,
+      "kl": 0.47802734375,
+      "learning_rate": 4.523587592959557e-06,
+      "loss": 0.7472,
+      "reward": 2.545016586780548,
+      "reward_std": 0.7152352035045624,
+      "rewards/accuracy_reward": 0.7135416865348816,
+      "rewards/reasoning_steps_reward": 0.9722222536802292,
+      "rewards/repetition_penalty_reward": -0.09126834943890572,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 477
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 199.4739646911621,
+      "epoch": 0.7169103862017248,
+      "grad_norm": 2.851573423423965,
+      "kl": 0.419921875,
+      "learning_rate": 4.479779771806699e-06,
+      "loss": 0.4624,
+      "reward": 2.572330951690674,
+      "reward_std": 0.5128215774893761,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.06265182606875896,
+      "rewards/tag_count_reward": 0.970052108168602,
+      "step": 478
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 204.9322967529297,
+      "epoch": 0.7184101987251593,
+      "grad_norm": 0.9372143757391351,
+      "kl": 0.325927734375,
+      "learning_rate": 4.436123796008149e-06,
+      "loss": 0.555,
+      "reward": 2.526001214981079,
+      "reward_std": 0.5274576209485531,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.9809028059244156,
+      "rewards/repetition_penalty_reward": -0.07208916172385216,
+      "rewards/tag_count_reward": 0.9609375149011612,
+      "step": 479
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 179.66146087646484,
+      "epoch": 0.7199100112485939,
+      "grad_norm": 1.4178034846859082,
+      "kl": 0.3603515625,
+      "learning_rate": 4.392620866414026e-06,
+      "loss": 0.385,
+      "reward": 2.7349607348442078,
+      "reward_std": 0.3730153478682041,
+      "rewards/accuracy_reward": 0.8072916865348816,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.04194910801015794,
+      "rewards/tag_count_reward": 0.9817708432674408,
+      "step": 480
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 208.9270896911621,
+      "epoch": 0.7214098237720284,
+      "grad_norm": 0.7783089792228611,
+      "kl": 0.318359375,
+      "learning_rate": 4.349272179664586e-06,
+      "loss": 0.4811,
+      "reward": 2.5647249817848206,
+      "reward_std": 0.525773711502552,
+      "rewards/accuracy_reward": 0.6822916716337204,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.06808762066066265,
+      "rewards/tag_count_reward": 0.9713541865348816,
+      "step": 481
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 211.82812881469727,
+      "epoch": 0.7229096362954631,
+      "grad_norm": 0.9640643886143824,
+      "kl": 0.37255859375,
+      "learning_rate": 4.3060789281573135e-06,
+      "loss": 0.5915,
+      "reward": 2.5671426653862,
+      "reward_std": 0.6002469211816788,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.07955875806510448,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 482
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.14583587646484,
+      "epoch": 0.7244094488188977,
+      "grad_norm": 2.303908104951783,
+      "kl": 0.44384765625,
+      "learning_rate": 4.263042300014112e-06,
+      "loss": 0.7262,
+      "reward": 2.679473400115967,
+      "reward_std": 0.6004434674978256,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.07530095893889666,
+      "rewards/tag_count_reward": 0.9492187798023224,
+      "step": 483
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 221.8177146911621,
+      "epoch": 0.7259092613423322,
+      "grad_norm": 1.4472721026550845,
+      "kl": 0.38330078125,
+      "learning_rate": 4.220163479048632e-06,
+      "loss": 0.5802,
+      "reward": 2.5944811701774597,
+      "reward_std": 0.6224361211061478,
+      "rewards/accuracy_reward": 0.7604166865348816,
+      "rewards/reasoning_steps_reward": 0.9704861491918564,
+      "rewards/repetition_penalty_reward": -0.08564047096297145,
+      "rewards/tag_count_reward": 0.9492187798023224,
+      "step": 484
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.98437881469727,
+      "epoch": 0.7274090738657668,
+      "grad_norm": 1.352373192683157,
+      "kl": 0.4716796875,
+      "learning_rate": 4.177443644733699e-06,
+      "loss": 0.8147,
+      "reward": 2.6568827629089355,
+      "reward_std": 0.7592978328466415,
+      "rewards/accuracy_reward": 0.848958358168602,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.09875956550240517,
+      "rewards/tag_count_reward": 0.9414062649011612,
+      "step": 485
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 245.1354217529297,
+      "epoch": 0.7289088863892014,
+      "grad_norm": 1.4809254865523336,
+      "kl": 0.44677734375,
+      "learning_rate": 4.134883972168877e-06,
+      "loss": 0.7938,
+      "reward": 2.593150496482849,
+      "reward_std": 0.7460722476243973,
+      "rewards/accuracy_reward": 0.7968750149011612,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.10954048298299313,
+      "rewards/tag_count_reward": 0.9283854365348816,
+      "step": 486
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 200.30729293823242,
+      "epoch": 0.7304086989126359,
+      "grad_norm": 0.9850557511849055,
+      "kl": 0.349609375,
+      "learning_rate": 4.092485632048142e-06,
+      "loss": 0.6449,
+      "reward": 2.609771966934204,
+      "reward_std": 0.45559458062052727,
+      "rewards/accuracy_reward": 0.739583358168602,
+      "rewards/reasoning_steps_reward": 0.9756944477558136,
+      "rewards/repetition_penalty_reward": -0.06774546951055527,
+      "rewards/tag_count_reward": 0.9622395932674408,
+      "step": 487
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.54688262939453,
+      "epoch": 0.7319085114360705,
+      "grad_norm": 0.7287131874968185,
+      "kl": 0.322265625,
+      "learning_rate": 4.050249790627675e-06,
+      "loss": 0.582,
+      "reward": 2.571848511695862,
+      "reward_std": 0.6445990055799484,
+      "rewards/accuracy_reward": 0.739583358168602,
+      "rewards/reasoning_steps_reward": 0.9618055671453476,
+      "rewards/repetition_penalty_reward": -0.0800613546743989,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 488
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 202.48437881469727,
+      "epoch": 0.7334083239595051,
+      "grad_norm": 0.8704961241862007,
+      "kl": 0.37255859375,
+      "learning_rate": 4.008177609693791e-06,
+      "loss": 0.6146,
+      "reward": 2.626339375972748,
+      "reward_std": 0.568084180355072,
+      "rewards/accuracy_reward": 0.770833358168602,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.07201146148145199,
+      "rewards/tag_count_reward": 0.9570312798023224,
+      "step": 489
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.3593864440918,
+      "epoch": 0.7349081364829396,
+      "grad_norm": 0.9183454896836771,
+      "kl": 0.42041015625,
+      "learning_rate": 3.966270246530975e-06,
+      "loss": 0.5819,
+      "reward": 2.5615740418434143,
+      "reward_std": 0.6961846798658371,
+      "rewards/accuracy_reward": 0.7447916716337204,
+      "rewards/reasoning_steps_reward": 0.9496527910232544,
+      "rewards/repetition_penalty_reward": -0.08208927698433399,
+      "rewards/tag_count_reward": 0.9492187798023224,
+      "step": 490
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 228.77084350585938,
+      "epoch": 0.7364079490063742,
+      "grad_norm": 0.7687763462508892,
+      "kl": 0.3623046875,
+      "learning_rate": 3.924528853890046e-06,
+      "loss": 0.5661,
+      "reward": 2.5703277587890625,
+      "reward_std": 0.6913676261901855,
+      "rewards/accuracy_reward": 0.739583358168602,
+      "rewards/reasoning_steps_reward": 0.9600694477558136,
+      "rewards/repetition_penalty_reward": -0.08375214599072933,
+      "rewards/tag_count_reward": 0.9544270932674408,
+      "step": 491
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 187.3229217529297,
+      "epoch": 0.7379077615298087,
+      "grad_norm": 0.751985130971036,
+      "kl": 0.44140625,
+      "learning_rate": 3.882954579956455e-06,
+      "loss": 0.3035,
+      "reward": 2.8052607774734497,
+      "reward_std": 0.39791389554739,
+      "rewards/accuracy_reward": 0.8906250298023224,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.04239548835903406,
+      "rewards/tag_count_reward": 0.9778645932674408,
+      "step": 492
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.6979217529297,
+      "epoch": 0.7394075740532433,
+      "grad_norm": 0.7659539513605523,
+      "kl": 0.29736328125,
+      "learning_rate": 3.841548568318706e-06,
+      "loss": 0.4617,
+      "reward": 2.7041468620300293,
+      "reward_std": 0.5456740781664848,
+      "rewards/accuracy_reward": 0.8281250149011612,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.060610027983784676,
+      "rewards/tag_count_reward": 0.9609375149011612,
+      "step": 493
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 212.7395896911621,
+      "epoch": 0.7409073865766779,
+      "grad_norm": 0.8837591427962717,
+      "kl": 0.31787109375,
+      "learning_rate": 3.8003119579368806e-06,
+      "loss": 0.838,
+      "reward": 2.5719869136810303,
+      "reward_std": 0.6782498955726624,
+      "rewards/accuracy_reward": 0.7187500149011612,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.07471455447375774,
+      "rewards/tag_count_reward": 0.9609375149011612,
+      "step": 494
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 194.1666717529297,
+      "epoch": 0.7424071991001124,
+      "grad_norm": 1.4368157135686637,
+      "kl": 0.29443359375,
+      "learning_rate": 3.7592458831113256e-06,
+      "loss": 0.4529,
+      "reward": 2.4826937317848206,
+      "reward_std": 0.47983065992593765,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9913194626569748,
+      "rewards/repetition_penalty_reward": -0.05810481309890747,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 495
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 202.92187881469727,
+      "epoch": 0.7439070116235471,
+      "grad_norm": 1.119059585365707,
+      "kl": 0.35302734375,
+      "learning_rate": 3.718351473451448e-06,
+      "loss": 0.6506,
+      "reward": 2.689814329147339,
+      "reward_std": 0.5917665362358093,
+      "rewards/accuracy_reward": 0.8385417014360428,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.06799825746566057,
+      "rewards/tag_count_reward": 0.9609375149011612,
+      "step": 496
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.27083587646484,
+      "epoch": 0.7454068241469817,
+      "grad_norm": 10.12905143841689,
+      "kl": 0.67431640625,
+      "learning_rate": 3.6776298538446307e-06,
+      "loss": 0.7213,
+      "reward": 2.456771969795227,
+      "reward_std": 0.6756933778524399,
+      "rewards/accuracy_reward": 0.6093750149011612,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.07751636672765017,
+      "rewards/tag_count_reward": 0.9544270932674408,
+      "step": 497
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 205.1510467529297,
+      "epoch": 0.7469066366704162,
+      "grad_norm": 0.9738907873033192,
+      "kl": 0.33203125,
+      "learning_rate": 3.6370821444253112e-06,
+      "loss": 0.3742,
+      "reward": 2.6159667372703552,
+      "reward_std": 0.45941271260380745,
+      "rewards/accuracy_reward": 0.7187500149011612,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.05938061675988138,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 498
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 199.4427146911621,
+      "epoch": 0.7484064491938508,
+      "grad_norm": 2.4109683634611785,
+      "kl": 0.46923828125,
+      "learning_rate": 3.5967094605441545e-06,
+      "loss": 0.5533,
+      "reward": 2.717509090900421,
+      "reward_std": 0.5818516314029694,
+      "rewards/accuracy_reward": 0.8177083432674408,
+      "rewards/reasoning_steps_reward": 0.986111119389534,
+      "rewards/repetition_penalty_reward": -0.06026873830705881,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 499
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 219.06250762939453,
+      "epoch": 0.7499062617172854,
+      "grad_norm": 0.7913869745911991,
+      "kl": 0.34765625,
+      "learning_rate": 3.5565129127373765e-06,
+      "loss": 0.6764,
+      "reward": 2.615268588066101,
+      "reward_std": 0.6076765581965446,
+      "rewards/accuracy_reward": 0.7812500298023224,
+      "rewards/reasoning_steps_reward": 0.96875,
+      "rewards/repetition_penalty_reward": -0.08134600473567843,
+      "rewards/tag_count_reward": 0.946614608168602,
+      "step": 500
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 196.53646087646484,
+      "epoch": 0.7514060742407199,
+      "grad_norm": 1.5573709839742271,
+      "kl": 0.47705078125,
+      "learning_rate": 3.5164936066961984e-06,
+      "loss": 0.3202,
+      "reward": 2.780966341495514,
+      "reward_std": 0.427102904766798,
+      "rewards/accuracy_reward": 0.895833358168602,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.057141443248838186,
+      "rewards/tag_count_reward": 0.9596354365348816,
+      "step": 501
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.50521087646484,
+      "epoch": 0.7529058867641545,
+      "grad_norm": 0.7528789099140956,
+      "kl": 0.30615234375,
+      "learning_rate": 3.476652643236431e-06,
+      "loss": 0.5839,
+      "reward": 2.4664222598075867,
+      "reward_std": 0.6568407695740461,
+      "rewards/accuracy_reward": 0.6302083507180214,
+      "rewards/reasoning_steps_reward": 0.958333358168602,
+      "rewards/repetition_penalty_reward": -0.07394242100417614,
+      "rewards/tag_count_reward": 0.9518229365348816,
+      "step": 502
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 216.70833587646484,
+      "epoch": 0.754405699287589,
+      "grad_norm": 1.1949507672459776,
+      "kl": 0.37646484375,
+      "learning_rate": 3.436991118268195e-06,
+      "loss": 0.6479,
+      "reward": 2.507534086704254,
+      "reward_std": 0.5748837292194366,
+      "rewards/accuracy_reward": 0.6510416865348816,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.07536534033715725,
+      "rewards/tag_count_reward": 0.9544270932674408,
+      "step": 503
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 246.4375114440918,
+      "epoch": 0.7559055118110236,
+      "grad_norm": 0.79103427108142,
+      "kl": 0.3798828125,
+      "learning_rate": 3.3975101227657726e-06,
+      "loss": 0.7612,
+      "reward": 2.440228283405304,
+      "reward_std": 0.8982982710003853,
+      "rewards/accuracy_reward": 0.677083358168602,
+      "rewards/reasoning_steps_reward": 0.9409722536802292,
+      "rewards/repetition_penalty_reward": -0.10230656852945685,
+      "rewards/tag_count_reward": 0.9244791716337204,
+      "step": 504
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.0208396911621,
+      "epoch": 0.7574053243344582,
+      "grad_norm": 1.382653144301306,
+      "kl": 0.349609375,
+      "learning_rate": 3.3582107427376044e-06,
+      "loss": 0.724,
+      "reward": 2.6514702439308167,
+      "reward_std": 0.7024101763963699,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/reasoning_steps_reward": 0.970486119389534,
+      "rewards/repetition_penalty_reward": -0.08333875052630901,
+      "rewards/tag_count_reward": 0.9518229216337204,
+      "step": 505
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 227.0677146911621,
+      "epoch": 0.7589051368578927,
+      "grad_norm": 1.0534398553331488,
+      "kl": 0.42431640625,
+      "learning_rate": 3.3190940591964094e-06,
+      "loss": 0.7011,
+      "reward": 2.410764992237091,
+      "reward_std": 0.7002580761909485,
+      "rewards/accuracy_reward": 0.5989583507180214,
+      "rewards/reasoning_steps_reward": 0.9531250149011612,
+      "rewards/repetition_penalty_reward": -0.08793305046856403,
+      "rewards/tag_count_reward": 0.946614608168602,
+      "step": 506
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 243.32813262939453,
+      "epoch": 0.7604049493813273,
+      "grad_norm": 2.499315939505216,
+      "kl": 0.5712890625,
+      "learning_rate": 3.2801611481294538e-06,
+      "loss": 0.6833,
+      "reward": 2.68420547246933,
+      "reward_std": 0.7678481340408325,
+      "rewards/accuracy_reward": 0.895833358168602,
+      "rewards/reasoning_steps_reward": 0.9496528059244156,
+      "rewards/repetition_penalty_reward": -0.10268685221672058,
+      "rewards/tag_count_reward": 0.9414062649011612,
+      "step": 507
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 231.5833396911621,
+      "epoch": 0.7619047619047619,
+      "grad_norm": 1.6168052609374044,
+      "kl": 0.4296875,
+      "learning_rate": 3.2414130804689492e-06,
+      "loss": 0.7108,
+      "reward": 2.642342746257782,
+      "reward_std": 0.765022836625576,
+      "rewards/accuracy_reward": 0.8229167014360428,
+      "rewards/reasoning_steps_reward": 0.9548611342906952,
+      "rewards/repetition_penalty_reward": -0.08725807629525661,
+      "rewards/tag_count_reward": 0.9518229216337204,
+      "step": 508
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 226.42188262939453,
+      "epoch": 0.7634045744281964,
+      "grad_norm": 1.6205129541399188,
+      "kl": 0.41552734375,
+      "learning_rate": 3.202850922062607e-06,
+      "loss": 0.7307,
+      "reward": 2.597358763217926,
+      "reward_std": 0.8237985223531723,
+      "rewards/accuracy_reward": 0.786458358168602,
+      "rewards/reasoning_steps_reward": 0.949652835726738,
+      "rewards/repetition_penalty_reward": -0.08797122351825237,
+      "rewards/tag_count_reward": 0.9492187649011612,
+      "step": 509
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.37501525878906,
+      "epoch": 0.7649043869516311,
+      "grad_norm": 1.0752711179781738,
+      "kl": 0.4208984375,
+      "learning_rate": 3.1644757336443023e-06,
+      "loss": 0.748,
+      "reward": 2.258354067802429,
+      "reward_std": 0.7389141619205475,
+      "rewards/accuracy_reward": 0.5260416865348816,
+      "rewards/reasoning_steps_reward": 0.960069477558136,
+      "rewards/repetition_penalty_reward": -0.14312171563506126,
+      "rewards/tag_count_reward": 0.915364608168602,
+      "step": 510
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 233.68230056762695,
+      "epoch": 0.7664041994750657,
+      "grad_norm": 1.0295291505864663,
+      "kl": 0.3603515625,
+      "learning_rate": 3.126288570804906e-06,
+      "loss": 0.813,
+      "reward": 2.466791570186615,
+      "reward_std": 0.8091250509023666,
+      "rewards/accuracy_reward": 0.677083358168602,
+      "rewards/reasoning_steps_reward": 0.9479167014360428,
+      "rewards/repetition_penalty_reward": -0.09440644644200802,
+      "rewards/tag_count_reward": 0.9361979365348816,
+      "step": 511
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 257.09375762939453,
+      "epoch": 0.7679040119985002,
+      "grad_norm": 1.1762814130428787,
+      "kl": 0.37109375,
+      "learning_rate": 3.0882904839632476e-06,
+      "loss": 0.7208,
+      "reward": 2.3307202458381653,
+      "reward_std": 0.7969870269298553,
+      "rewards/accuracy_reward": 0.5781250298023224,
+      "rewards/reasoning_steps_reward": 0.9496527761220932,
+      "rewards/repetition_penalty_reward": -0.12023470550775528,
+      "rewards/tag_count_reward": 0.9231771230697632,
+      "step": 512
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 216.29688262939453,
+      "epoch": 0.7694038245219348,
+      "grad_norm": 6.576307135246261,
+      "kl": 0.46337890625,
+      "learning_rate": 3.050482518337221e-06,
+      "loss": 0.5806,
+      "reward": 2.5199908018112183,
+      "reward_std": 0.5410640314221382,
+      "rewards/accuracy_reward": 0.6927083432674408,
+      "rewards/reasoning_steps_reward": 0.9670138955116272,
+      "rewards/repetition_penalty_reward": -0.08113771304488182,
+      "rewards/tag_count_reward": 0.9414062798023224,
+      "step": 513
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 244.30208587646484,
+      "epoch": 0.7709036370453693,
+      "grad_norm": 0.8493543704484349,
+      "kl": 0.4072265625,
+      "learning_rate": 3.012865713915033e-06,
+      "loss": 0.8002,
+      "reward": 2.496680438518524,
+      "reward_std": 0.8196369558572769,
+      "rewards/accuracy_reward": 0.7343750149011612,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.11226061917841434,
+      "rewards/tag_count_reward": 0.923177108168602,
+      "step": 514
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 228.02605056762695,
+      "epoch": 0.7724034495688039,
+      "grad_norm": 0.782983798376921,
+      "kl": 0.50830078125,
+      "learning_rate": 2.9754411054265966e-06,
+      "loss": 0.7341,
+      "reward": 2.4324201941490173,
+      "reward_std": 0.795376256108284,
+      "rewards/accuracy_reward": 0.6614583507180214,
+      "rewards/reasoning_steps_reward": 0.9340277910232544,
+      "rewards/repetition_penalty_reward": -0.09665969014167786,
+      "rewards/tag_count_reward": 0.9335937798023224,
+      "step": 515
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.140625,
+      "epoch": 0.7739032620922385,
+      "grad_norm": 1.2353929587709405,
+      "kl": 0.3515625,
+      "learning_rate": 2.9382097223150675e-06,
+      "loss": 0.6533,
+      "reward": 2.5715506076812744,
+      "reward_std": 0.6366243287920952,
+      "rewards/accuracy_reward": 0.7239583432674408,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.07341482397168875,
+      "rewards/tag_count_reward": 0.9557291716337204,
+      "step": 516
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 258.67188262939453,
+      "epoch": 0.775403074615673,
+      "grad_norm": 0.8839020653845002,
+      "kl": 0.35693359375,
+      "learning_rate": 2.9011725887085286e-06,
+      "loss": 0.768,
+      "reward": 2.4018173813819885,
+      "reward_std": 0.8271952420473099,
+      "rewards/accuracy_reward": 0.6406250223517418,
+      "rewards/reasoning_steps_reward": 0.9513889104127884,
+      "rewards/repetition_penalty_reward": -0.1159778069704771,
+      "rewards/tag_count_reward": 0.92578125,
+      "step": 517
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 195.9947967529297,
+      "epoch": 0.7769028871391076,
+      "grad_norm": 0.7917880481551695,
+      "kl": 0.3603515625,
+      "learning_rate": 2.8643307233918192e-06,
+      "loss": 0.3036,
+      "reward": 2.690393328666687,
+      "reward_std": 0.3900110796093941,
+      "rewards/accuracy_reward": 0.786458358168602,
+      "rewards/reasoning_steps_reward": 0.982638880610466,
+      "rewards/repetition_penalty_reward": -0.05136016756296158,
+      "rewards/tag_count_reward": 0.9726562649011612,
+      "step": 518
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.57812881469727,
+      "epoch": 0.7784026996625422,
+      "grad_norm": 0.8157613306440419,
+      "kl": 0.3388671875,
+      "learning_rate": 2.827685139778511e-06,
+      "loss": 0.6223,
+      "reward": 2.477262020111084,
+      "reward_std": 0.7161538153886795,
+      "rewards/accuracy_reward": 0.6718750149011612,
+      "rewards/reasoning_steps_reward": 0.9618055671453476,
+      "rewards/repetition_penalty_reward": -0.09782484546303749,
+      "rewards/tag_count_reward": 0.9414062649011612,
+      "step": 519
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.87500381469727,
+      "epoch": 0.7799025121859767,
+      "grad_norm": 0.9727421187621869,
+      "kl": 0.35498046875,
+      "learning_rate": 2.7912368458830295e-06,
+      "loss": 0.6298,
+      "reward": 2.616613507270813,
+      "reward_std": 0.6122787222266197,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.08390727639198303,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 520
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 208.30208587646484,
+      "epoch": 0.7814023247094113,
+      "grad_norm": 0.8429352541214471,
+      "kl": 0.327392578125,
+      "learning_rate": 2.7549868442929286e-06,
+      "loss": 0.4077,
+      "reward": 2.774833917617798,
+      "reward_std": 0.5252001956105232,
+      "rewards/accuracy_reward": 0.911458358168602,
+      "rewards/reasoning_steps_reward": 0.9722222238779068,
+      "rewards/repetition_penalty_reward": -0.06848222017288208,
+      "rewards/tag_count_reward": 0.9596354514360428,
+      "step": 521
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 211.09375381469727,
+      "epoch": 0.7829021372328459,
+      "grad_norm": 1.0326110093541125,
+      "kl": 0.31884765625,
+      "learning_rate": 2.7189361321413144e-06,
+      "loss": 0.421,
+      "reward": 2.417696237564087,
+      "reward_std": 0.3683694452047348,
+      "rewards/accuracy_reward": 0.5260416865348816,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.06190443877130747,
+      "rewards/tag_count_reward": 0.9674479216337204,
+      "step": 522
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.7604217529297,
+      "epoch": 0.7844019497562804,
+      "grad_norm": 2.2355283965994865,
+      "kl": 0.39013671875,
+      "learning_rate": 2.683085701079412e-06,
+      "loss": 0.5656,
+      "reward": 2.6360539197921753,
+      "reward_std": 0.5976300239562988,
+      "rewards/accuracy_reward": 0.7864583507180214,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.06924132350832224,
+      "rewards/tag_count_reward": 0.9518229365348816,
+      "step": 523
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 202.9322967529297,
+      "epoch": 0.7859017622797151,
+      "grad_norm": 0.9756200081911167,
+      "kl": 0.3701171875,
+      "learning_rate": 2.647436537249294e-06,
+      "loss": 0.545,
+      "reward": 2.322978913784027,
+      "reward_std": 0.5455097928643227,
+      "rewards/accuracy_reward": 0.4531250223517418,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.0633058724924922,
+      "rewards/tag_count_reward": 0.9557292014360428,
+      "step": 524
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 203.1145896911621,
+      "epoch": 0.7874015748031497,
+      "grad_norm": 0.7876826353251758,
+      "kl": 0.40576171875,
+      "learning_rate": 2.611989621256745e-06,
+      "loss": 0.5895,
+      "reward": 2.726710319519043,
+      "reward_std": 0.5035083070397377,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9809027910232544,
+      "rewards/repetition_penalty_reward": -0.06669262330979109,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 525
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.5364646911621,
+      "epoch": 0.7889013873265842,
+      "grad_norm": 1.2271941862677709,
+      "kl": 0.29833984375,
+      "learning_rate": 2.5767459281443064e-06,
+      "loss": 0.6082,
+      "reward": 2.505469024181366,
+      "reward_std": 0.6449078023433685,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.07352408953011036,
+      "rewards/tag_count_reward": 0.9557291716337204,
+      "step": 526
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 239.5677146911621,
+      "epoch": 0.7904011998500188,
+      "grad_norm": 0.9183905328986752,
+      "kl": 0.4296875,
+      "learning_rate": 2.541706427364431e-06,
+      "loss": 0.8342,
+      "reward": 2.675832152366638,
+      "reward_std": 0.8275108933448792,
+      "rewards/accuracy_reward": 0.864583358168602,
+      "rewards/reasoning_steps_reward": 0.954861119389534,
+      "rewards/repetition_penalty_reward": -0.09283116087317467,
+      "rewards/tag_count_reward": 0.9492187649011612,
+      "step": 527
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.96875762939453,
+      "epoch": 0.7919010123734533,
+      "grad_norm": 0.7482995574885967,
+      "kl": 0.32275390625,
+      "learning_rate": 2.506872082752834e-06,
+      "loss": 0.6547,
+      "reward": 2.569875657558441,
+      "reward_std": 0.5808727741241455,
+      "rewards/accuracy_reward": 0.6927083432674408,
+      "rewards/reasoning_steps_reward": 0.9861111342906952,
+      "rewards/repetition_penalty_reward": -0.0724856061860919,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 528
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 208.06250381469727,
+      "epoch": 0.7934008248968879,
+      "grad_norm": 1.2950704438743703,
+      "kl": 0.35400390625,
+      "learning_rate": 2.4722438525019764e-06,
+      "loss": 0.3873,
+      "reward": 2.7978073358535767,
+      "reward_std": 0.486385278403759,
+      "rewards/accuracy_reward": 0.9010416865348816,
+      "rewards/reasoning_steps_reward": 0.9809027910232544,
+      "rewards/repetition_penalty_reward": -0.052887264639139175,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 529
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.96875381469727,
+      "epoch": 0.7949006374203225,
+      "grad_norm": 0.9168395385162036,
+      "kl": 0.33203125,
+      "learning_rate": 2.4378226891347056e-06,
+      "loss": 0.5429,
+      "reward": 2.631447494029999,
+      "reward_std": 0.530893087387085,
+      "rewards/accuracy_reward": 0.7500000149011612,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.07167758233845234,
+      "rewards/tag_count_reward": 0.9687500149011612,
+      "step": 530
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 237.88021087646484,
+      "epoch": 0.796400449943757,
+      "grad_norm": 2.5084733671661934,
+      "kl": 0.4423828125,
+      "learning_rate": 2.403609539478056e-06,
+      "loss": 0.5644,
+      "reward": 2.736763119697571,
+      "reward_std": 0.614450603723526,
+      "rewards/accuracy_reward": 0.890625,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.08224734663963318,
+      "rewards/tag_count_reward": 0.9596354514360428,
+      "step": 531
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 201.6822967529297,
+      "epoch": 0.7979002624671916,
+      "grad_norm": 1.2103141736384226,
+      "kl": 0.32958984375,
+      "learning_rate": 2.3696053446372026e-06,
+      "loss": 0.2212,
+      "reward": 2.7792577147483826,
+      "reward_std": 0.2872903672978282,
+      "rewards/accuracy_reward": 0.8541666865348816,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.047565262764692307,
+      "rewards/tag_count_reward": 0.9778645932674408,
+      "step": 532
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 205.01562881469727,
+      "epoch": 0.7994000749906262,
+      "grad_norm": 0.9409501748390533,
+      "kl": 0.33740234375,
+      "learning_rate": 2.3358110399695788e-06,
+      "loss": 0.5166,
+      "reward": 2.526670515537262,
+      "reward_std": 0.5588072910904884,
+      "rewards/accuracy_reward": 0.6510416865348816,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.057530895806849,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 533
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.4322967529297,
+      "epoch": 0.8008998875140607,
+      "grad_norm": 0.9487060971483162,
+      "kl": 0.3642578125,
+      "learning_rate": 2.302227555059141e-06,
+      "loss": 0.5154,
+      "reward": 2.5382038950920105,
+      "reward_std": 0.591885045170784,
+      "rewards/accuracy_reward": 0.6614583432674408,
+      "rewards/reasoning_steps_reward": 0.970486119389534,
+      "rewards/repetition_penalty_reward": -0.06769899744540453,
+      "rewards/tag_count_reward": 0.973958358168602,
+      "step": 534
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 211.07292556762695,
+      "epoch": 0.8023997000374953,
+      "grad_norm": 0.929609225864085,
+      "kl": 0.396484375,
+      "learning_rate": 2.2688558136908025e-06,
+      "loss": 0.5148,
+      "reward": 2.7448110580444336,
+      "reward_std": 0.46144695580005646,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.05336613114923239,
+      "rewards/tag_count_reward": 0.9700520932674408,
+      "step": 535
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.8802146911621,
+      "epoch": 0.8038995125609298,
+      "grad_norm": 1.584357940413058,
+      "kl": 0.30126953125,
+      "learning_rate": 2.2356967338250223e-06,
+      "loss": 0.431,
+      "reward": 2.501081109046936,
+      "reward_std": 0.4514058753848076,
+      "rewards/accuracy_reward": 0.6250000074505806,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.07140162866562605,
+      "rewards/tag_count_reward": 0.9648437798023224,
+      "step": 536
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 233.20834350585938,
+      "epoch": 0.8053993250843644,
+      "grad_norm": 64.06543251128925,
+      "kl": 1.220703125,
+      "learning_rate": 2.202751227572556e-06,
+      "loss": 0.7764,
+      "reward": 2.6655821800231934,
+      "reward_std": 0.7584892809391022,
+      "rewards/accuracy_reward": 0.848958358168602,
+      "rewards/reasoning_steps_reward": 0.958333358168602,
+      "rewards/repetition_penalty_reward": -0.08441784046590328,
+      "rewards/tag_count_reward": 0.9427083432674408,
+      "step": 537
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 227.40625762939453,
+      "epoch": 0.8068991376077991,
+      "grad_norm": 0.7719266837120854,
+      "kl": 0.37646484375,
+      "learning_rate": 2.1700202011693573e-06,
+      "loss": 0.5498,
+      "reward": 2.6957992911338806,
+      "reward_std": 0.6702239066362381,
+      "rewards/accuracy_reward": 0.8854166865348816,
+      "rewards/reasoning_steps_reward": 0.9496528059244156,
+      "rewards/repetition_penalty_reward": -0.0819785725325346,
+      "rewards/tag_count_reward": 0.942708358168602,
+      "step": 538
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 234.98958587646484,
+      "epoch": 0.8083989501312336,
+      "grad_norm": 2.038897171924456,
+      "kl": 0.32861328125,
+      "learning_rate": 2.1375045549516636e-06,
+      "loss": 0.5421,
+      "reward": 2.608824074268341,
+      "reward_std": 0.5336829051375389,
+      "rewards/accuracy_reward": 0.7239583730697632,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.07867603516206145,
+      "rewards/tag_count_reward": 0.973958358168602,
+      "step": 539
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 213.40105056762695,
+      "epoch": 0.8098987626546682,
+      "grad_norm": 1.2689341920662447,
+      "kl": 0.39599609375,
+      "learning_rate": 2.105205183331224e-06,
+      "loss": 0.6115,
+      "reward": 2.6301316022872925,
+      "reward_std": 0.6800608858466148,
+      "rewards/accuracy_reward": 0.8229166865348816,
+      "rewards/reasoning_steps_reward": 0.9461806118488312,
+      "rewards/repetition_penalty_reward": -0.08427836652845144,
+      "rewards/tag_count_reward": 0.9453125298023224,
+      "step": 540
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 229.3697967529297,
+      "epoch": 0.8113985751781028,
+      "grad_norm": 0.6531166991811557,
+      "kl": 0.2939453125,
+      "learning_rate": 2.0731229747706926e-06,
+      "loss": 0.5031,
+      "reward": 2.5187647342681885,
+      "reward_std": 0.5751441568136215,
+      "rewards/accuracy_reward": 0.6562500149011612,
+      "rewards/reasoning_steps_reward": 0.965277761220932,
+      "rewards/repetition_penalty_reward": -0.06630483735352755,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 541
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 230.78125381469727,
+      "epoch": 0.8128983877015373,
+      "grad_norm": 0.6612265161899963,
+      "kl": 0.31640625,
+      "learning_rate": 2.041258811759195e-06,
+      "loss": 0.5724,
+      "reward": 2.5856213569641113,
+      "reward_std": 0.6729920580983162,
+      "rewards/accuracy_reward": 0.7395833507180214,
+      "rewards/reasoning_steps_reward": 0.9722222536802292,
+      "rewards/repetition_penalty_reward": -0.08842381555587053,
+      "rewards/tag_count_reward": 0.9622395932674408,
+      "step": 542
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 187.7239646911621,
+      "epoch": 0.8143982002249719,
+      "grad_norm": 1.1508949391019638,
+      "kl": 0.290283203125,
+      "learning_rate": 2.009613570788057e-06,
+      "loss": 0.3922,
+      "reward": 2.7094662189483643,
+      "reward_std": 0.4391992464661598,
+      "rewards/accuracy_reward": 0.7968750149011612,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.042704006657004356,
+      "rewards/tag_count_reward": 0.9778645932674408,
+      "step": 543
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 209.62500762939453,
+      "epoch": 0.8158980127484065,
+      "grad_norm": 0.7726809000433558,
+      "kl": 0.29736328125,
+      "learning_rate": 1.978188122326683e-06,
+      "loss": 0.4558,
+      "reward": 2.690703511238098,
+      "reward_std": 0.49602875113487244,
+      "rewards/accuracy_reward": 0.8125000149011612,
+      "rewards/reasoning_steps_reward": 0.9722222238779068,
+      "rewards/repetition_penalty_reward": -0.06537291780114174,
+      "rewards/tag_count_reward": 0.9713541716337204,
+      "step": 544
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.43750762939453,
+      "epoch": 0.817397825271841,
+      "grad_norm": 1.4290665690326807,
+      "kl": 0.3125,
+      "learning_rate": 1.946983330798621e-06,
+      "loss": 0.5241,
+      "reward": 2.562969446182251,
+      "reward_std": 0.5197947286069393,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.06203062180429697,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 545
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 208.3645896911621,
+      "epoch": 0.8188976377952756,
+      "grad_norm": 1.0265157040309887,
+      "kl": 0.2939453125,
+      "learning_rate": 1.916000054557783e-06,
+      "loss": 0.6364,
+      "reward": 2.629048526287079,
+      "reward_std": 0.6721891462802887,
+      "rewards/accuracy_reward": 0.770833358168602,
+      "rewards/reasoning_steps_reward": 0.9704861491918564,
+      "rewards/repetition_penalty_reward": -0.0693023600615561,
+      "rewards/tag_count_reward": 0.9570312798023224,
+      "step": 546
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.62500762939453,
+      "epoch": 0.8203974503187101,
+      "grad_norm": 1.2230109106160663,
+      "kl": 0.31201171875,
+      "learning_rate": 1.8852391458648323e-06,
+      "loss": 0.4479,
+      "reward": 2.6805137395858765,
+      "reward_std": 0.4807371646165848,
+      "rewards/accuracy_reward": 0.7916666865348816,
+      "rewards/reasoning_steps_reward": 0.9843750149011612,
+      "rewards/repetition_penalty_reward": -0.06297590211033821,
+      "rewards/tag_count_reward": 0.9674479365348816,
+      "step": 547
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 200.45312881469727,
+      "epoch": 0.8218972628421447,
+      "grad_norm": 0.8191568879278863,
+      "kl": 0.3046875,
+      "learning_rate": 1.854701450863744e-06,
+      "loss": 0.4753,
+      "reward": 2.549809992313385,
+      "reward_std": 0.5221479944884777,
+      "rewards/accuracy_reward": 0.6562500298023224,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.053922670893371105,
+      "rewards/tag_count_reward": 0.9648437649011612,
+      "step": 548
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 221.0989646911621,
+      "epoch": 0.8233970753655793,
+      "grad_norm": 0.9675577366532915,
+      "kl": 0.36083984375,
+      "learning_rate": 1.8243878095585244e-06,
+      "loss": 0.6371,
+      "reward": 2.734562575817108,
+      "reward_std": 0.630949005484581,
+      "rewards/accuracy_reward": 0.8906250298023224,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.08314584195613861,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 549
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 207.06250762939453,
+      "epoch": 0.8248968878890138,
+      "grad_norm": 0.9898045914966018,
+      "kl": 0.30859375,
+      "learning_rate": 1.7942990557901119e-06,
+      "loss": 0.3167,
+      "reward": 2.5159996151924133,
+      "reward_std": 0.4262428246438503,
+      "rewards/accuracy_reward": 0.5937500149011612,
+      "rewards/reasoning_steps_reward": 0.9895833432674408,
+      "rewards/repetition_penalty_reward": -0.04780261078849435,
+      "rewards/tag_count_reward": 0.9804687649011612,
+      "step": 550
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.83855056762695,
+      "epoch": 0.8263967004124484,
+      "grad_norm": 1.3960120493498462,
+      "kl": 0.44775390625,
+      "learning_rate": 1.7644360172134323e-06,
+      "loss": 0.7773,
+      "reward": 2.514442801475525,
+      "reward_std": 0.7914524525403976,
+      "rewards/accuracy_reward": 0.6927083432674408,
+      "rewards/reasoning_steps_reward": 0.9496528059244156,
+      "rewards/repetition_penalty_reward": -0.07974134013056755,
+      "rewards/tag_count_reward": 0.9518229365348816,
+      "step": 551
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.77605056762695,
+      "epoch": 0.8278965129358831,
+      "grad_norm": 0.7179910870567698,
+      "kl": 0.3623046875,
+      "learning_rate": 1.734799515274641e-06,
+      "loss": 0.507,
+      "reward": 2.6635890007019043,
+      "reward_std": 0.5695291832089424,
+      "rewards/accuracy_reward": 0.8072916716337204,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.07078609801828861,
+      "rewards/tag_count_reward": 0.9635416716337204,
+      "step": 552
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 184.89583587646484,
+      "epoch": 0.8293963254593176,
+      "grad_norm": 1.660541115002357,
+      "kl": 0.3603515625,
+      "learning_rate": 1.7053903651885217e-06,
+      "loss": 0.3674,
+      "reward": 2.609849989414215,
+      "reward_std": 0.3290810212492943,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9791667014360428,
+      "rewards/repetition_penalty_reward": -0.0516084156697616,
+      "rewards/tag_count_reward": 0.9739583432674408,
+      "step": 553
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.2968864440918,
+      "epoch": 0.8308961379827522,
+      "grad_norm": 1.1421644219209315,
+      "kl": 0.3388671875,
+      "learning_rate": 1.6762093759160614e-06,
+      "loss": 0.6045,
+      "reward": 2.6770655512809753,
+      "reward_std": 0.5518500655889511,
+      "rewards/accuracy_reward": 0.8177083432674408,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.08161513973027468,
+      "rewards/tag_count_reward": 0.958333358168602,
+      "step": 554
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 271.15625762939453,
+      "epoch": 0.8323959505061868,
+      "grad_norm": 2.2156251851166755,
+      "kl": 0.462890625,
+      "learning_rate": 1.647257350142204e-06,
+      "loss": 0.8727,
+      "reward": 2.427699863910675,
+      "reward_std": 0.8616139888763428,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/reasoning_steps_reward": 0.954861119389534,
+      "rewards/repetition_penalty_reward": -0.12221339344978333,
+      "rewards/tag_count_reward": 0.9283854514360428,
+      "step": 555
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 203.6354217529297,
+      "epoch": 0.8338957630296213,
+      "grad_norm": 4.90298950509675,
+      "kl": 0.46337890625,
+      "learning_rate": 1.618535084253765e-06,
+      "loss": 0.6818,
+      "reward": 2.632472336292267,
+      "reward_std": 0.5403143912553787,
+      "rewards/accuracy_reward": 0.7656250298023224,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.06457635015249252,
+      "rewards/tag_count_reward": 0.9609375149011612,
+      "step": 556
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 279.3125114440918,
+      "epoch": 0.8353955755530559,
+      "grad_norm": 2.3574073221661074,
+      "kl": 0.6494140625,
+      "learning_rate": 1.5900433683175277e-06,
+      "loss": 0.7594,
+      "reward": 2.3048887848854065,
+      "reward_std": 0.8010806366801262,
+      "rewards/accuracy_reward": 0.572916679084301,
+      "rewards/reasoning_steps_reward": 0.9531250149011612,
+      "rewards/repetition_penalty_reward": -0.1365174800157547,
+      "rewards/tag_count_reward": 0.9153645932674408,
+      "step": 557
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.7135543823242,
+      "epoch": 0.8368953880764904,
+      "grad_norm": 1.9632445237865204,
+      "kl": 0.466796875,
+      "learning_rate": 1.5617829860585087e-06,
+      "loss": 0.6438,
+      "reward": 2.505094826221466,
+      "reward_std": 0.7284489870071411,
+      "rewards/accuracy_reward": 0.7187500149011612,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.10992250964045525,
+      "rewards/tag_count_reward": 0.930989608168602,
+      "step": 558
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.82292556762695,
+      "epoch": 0.838395200599925,
+      "grad_norm": 1.3529276790590763,
+      "kl": 0.50244140625,
+      "learning_rate": 1.533754714838408e-06,
+      "loss": 0.7146,
+      "reward": 2.5260643362998962,
+      "reward_std": 0.6827712506055832,
+      "rewards/accuracy_reward": 0.7083333432674408,
+      "rewards/reasoning_steps_reward": 0.953125,
+      "rewards/repetition_penalty_reward": -0.07810238003730774,
+      "rewards/tag_count_reward": 0.942708358168602,
+      "step": 559
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 295.43750762939453,
+      "epoch": 0.8398950131233596,
+      "grad_norm": 1.3595681175219,
+      "kl": 0.5419921875,
+      "learning_rate": 1.5059593256342142e-06,
+      "loss": 0.5809,
+      "reward": 2.471269369125366,
+      "reward_std": 0.7434787154197693,
+      "rewards/accuracy_reward": 0.7239583507180214,
+      "rewards/reasoning_steps_reward": 0.9739583432674408,
+      "rewards/repetition_penalty_reward": -0.14852236676961184,
+      "rewards/tag_count_reward": 0.9218750149011612,
+      "step": 560
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.3073043823242,
+      "epoch": 0.8413948256467941,
+      "grad_norm": 3.39049441324729,
+      "kl": 0.57568359375,
+      "learning_rate": 1.4783975830170028e-06,
+      "loss": 0.7564,
+      "reward": 2.414111316204071,
+      "reward_std": 0.9404040277004242,
+      "rewards/accuracy_reward": 0.6875000149011612,
+      "rewards/reasoning_steps_reward": 0.9392361491918564,
+      "rewards/repetition_penalty_reward": -0.1253854325041175,
+      "rewards/tag_count_reward": 0.9127604365348816,
+      "step": 561
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 256.3125114440918,
+      "epoch": 0.8428946381702287,
+      "grad_norm": 2.8979024715584325,
+      "kl": 0.689453125,
+      "learning_rate": 1.4510702451309055e-06,
+      "loss": 0.8145,
+      "reward": 2.4051551818847656,
+      "reward_std": 0.7071145176887512,
+      "rewards/accuracy_reward": 0.6250000298023224,
+      "rewards/reasoning_steps_reward": 0.972222238779068,
+      "rewards/repetition_penalty_reward": -0.12175463326275349,
+      "rewards/tag_count_reward": 0.9296875149011612,
+      "step": 562
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.7708435058594,
+      "epoch": 0.8443944506936633,
+      "grad_norm": 1.408501605320147,
+      "kl": 0.505859375,
+      "learning_rate": 1.4239780636722555e-06,
+      "loss": 0.7652,
+      "reward": 2.3463268280029297,
+      "reward_std": 0.7379028648138046,
+      "rewards/accuracy_reward": 0.5625000149011612,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.11374275013804436,
+      "rewards/tag_count_reward": 0.9322916865348816,
+      "step": 563
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.82813262939453,
+      "epoch": 0.8458942632170978,
+      "grad_norm": 2.6676821866666742,
+      "kl": 0.42431640625,
+      "learning_rate": 1.39712178386891e-06,
+      "loss": 0.8024,
+      "reward": 2.43689888715744,
+      "reward_std": 0.8974379524588585,
+      "rewards/accuracy_reward": 0.739583358168602,
+      "rewards/reasoning_steps_reward": 0.928819477558136,
+      "rewards/repetition_penalty_reward": -0.13905610889196396,
+      "rewards/tag_count_reward": 0.907552108168602,
+      "step": 564
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 264.0989685058594,
+      "epoch": 0.8473940757405324,
+      "grad_norm": 0.8627370825958761,
+      "kl": 0.43115234375,
+      "learning_rate": 1.3705021444597521e-06,
+      "loss": 0.7752,
+      "reward": 2.607667088508606,
+      "reward_std": 0.782805323600769,
+      "rewards/accuracy_reward": 0.817708358168602,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.11672540940344334,
+      "rewards/tag_count_reward": 0.9309895932674408,
+      "step": 565
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.8802185058594,
+      "epoch": 0.8488938882639671,
+      "grad_norm": 0.9159232949782135,
+      "kl": 0.4345703125,
+      "learning_rate": 1.344119877674368e-06,
+      "loss": 0.9667,
+      "reward": 2.3450130224227905,
+      "reward_std": 0.9461807906627655,
+      "rewards/accuracy_reward": 0.6250000149011612,
+      "rewards/reasoning_steps_reward": 0.9375000149011612,
+      "rewards/repetition_penalty_reward": -0.1341537069529295,
+      "rewards/tag_count_reward": 0.9166667014360428,
+      "step": 566
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 273.5364685058594,
+      "epoch": 0.8503937007874016,
+      "grad_norm": 0.9183785293807163,
+      "kl": 0.412109375,
+      "learning_rate": 1.3179757092129087e-06,
+      "loss": 0.7443,
+      "reward": 2.350478619337082,
+      "reward_std": 0.8575991243124008,
+      "rewards/accuracy_reward": 0.6250000223517418,
+      "rewards/reasoning_steps_reward": 0.9322917014360428,
+      "rewards/repetition_penalty_reward": -0.13129226304590702,
+      "rewards/tag_count_reward": 0.9244791865348816,
+      "step": 567
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 260.76563262939453,
+      "epoch": 0.8518935133108362,
+      "grad_norm": 0.7672423282948194,
+      "kl": 0.32177734375,
+      "learning_rate": 1.292070358226124e-06,
+      "loss": 0.6237,
+      "reward": 2.4249655306339264,
+      "reward_std": 0.6657977253198624,
+      "rewards/accuracy_reward": 0.6302083507180214,
+      "rewards/reasoning_steps_reward": 0.9652778208255768,
+      "rewards/repetition_penalty_reward": -0.11583315394818783,
+      "rewards/tag_count_reward": 0.9453125149011612,
+      "step": 568
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 242.81250762939453,
+      "epoch": 0.8533933258342707,
+      "grad_norm": 1.510802559149807,
+      "kl": 0.3408203125,
+      "learning_rate": 1.2664045372955858e-06,
+      "loss": 0.8092,
+      "reward": 2.3671931624412537,
+      "reward_std": 0.747007891535759,
+      "rewards/accuracy_reward": 0.5677083507180214,
+      "rewards/reasoning_steps_reward": 0.9583333730697632,
+      "rewards/repetition_penalty_reward": -0.10155703499913216,
+      "rewards/tag_count_reward": 0.942708358168602,
+      "step": 569
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 245.81250762939453,
+      "epoch": 0.8548931383577053,
+      "grad_norm": 1.2507925070262236,
+      "kl": 0.45947265625,
+      "learning_rate": 1.2409789524140813e-06,
+      "loss": 0.7475,
+      "reward": 2.5338348150253296,
+      "reward_std": 0.7423350065946579,
+      "rewards/accuracy_reward": 0.7500000298023224,
+      "rewards/reasoning_steps_reward": 0.9461805820465088,
+      "rewards/repetition_penalty_reward": -0.10375208128243685,
+      "rewards/tag_count_reward": 0.9414062798023224,
+      "step": 570
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 252.5000114440918,
+      "epoch": 0.8563929508811399,
+      "grad_norm": 0.898004858587946,
+      "kl": 0.33984375,
+      "learning_rate": 1.2157943029661977e-06,
+      "loss": 0.7782,
+      "reward": 2.514720618724823,
+      "reward_std": 0.7610578685998917,
+      "rewards/accuracy_reward": 0.7343750298023224,
+      "rewards/reasoning_steps_reward": 0.9548611491918564,
+      "rewards/repetition_penalty_reward": -0.1081093717366457,
+      "rewards/tag_count_reward": 0.9335937649011612,
+      "step": 571
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 208.18750762939453,
+      "epoch": 0.8578927634045744,
+      "grad_norm": 1.539413571101912,
+      "kl": 0.33935546875,
+      "learning_rate": 1.1908512817090833e-06,
+      "loss": 0.8174,
+      "reward": 2.4172632098197937,
+      "reward_std": 0.6794392615556717,
+      "rewards/accuracy_reward": 0.5520833432674408,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.06580978166311979,
+      "rewards/tag_count_reward": 0.962239608168602,
+      "step": 572
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 240.3072967529297,
+      "epoch": 0.859392575928009,
+      "grad_norm": 0.8770707592548431,
+      "kl": 0.335205078125,
+      "learning_rate": 1.1661505747533897e-06,
+      "loss": 0.5468,
+      "reward": 2.615067780017853,
+      "reward_std": 0.6282743141055107,
+      "rewards/accuracy_reward": 0.78125,
+      "rewards/reasoning_steps_reward": 0.9739583432674408,
+      "rewards/repetition_penalty_reward": -0.09066134784370661,
+      "rewards/tag_count_reward": 0.9505208432674408,
+      "step": 573
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 207.15625762939453,
+      "epoch": 0.8608923884514436,
+      "grad_norm": 0.952206754417322,
+      "kl": 0.29248046875,
+      "learning_rate": 1.1416928615444013e-06,
+      "loss": 0.4348,
+      "reward": 2.639313220977783,
+      "reward_std": 0.514240987598896,
+      "rewards/accuracy_reward": 0.7447916865348816,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.054697235114872456,
+      "rewards/tag_count_reward": 0.9700520932674408,
+      "step": 574
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 221.3072967529297,
+      "epoch": 0.8623922009748781,
+      "grad_norm": 0.9603023631517865,
+      "kl": 0.35107421875,
+      "learning_rate": 1.1174788148433423e-06,
+      "loss": 0.4322,
+      "reward": 2.698141098022461,
+      "reward_std": 0.5075501780956984,
+      "rewards/accuracy_reward": 0.8541667014360428,
+      "rewards/reasoning_steps_reward": 0.9704861342906952,
+      "rewards/repetition_penalty_reward": -0.08093887567520142,
+      "rewards/tag_count_reward": 0.9544270932674408,
+      "step": 575
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.0677146911621,
+      "epoch": 0.8638920134983127,
+      "grad_norm": 1.2989707455803579,
+      "kl": 0.3115234375,
+      "learning_rate": 1.0935091007088761e-06,
+      "loss": 0.7183,
+      "reward": 2.5909855365753174,
+      "reward_std": 0.5980718731880188,
+      "rewards/accuracy_reward": 0.7447916716337204,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.08566383551806211,
+      "rewards/tag_count_reward": 0.954427108168602,
+      "step": 576
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 203.7135467529297,
+      "epoch": 0.8653918260217472,
+      "grad_norm": 1.2239167370700748,
+      "kl": 0.29150390625,
+      "learning_rate": 1.069784378478781e-06,
+      "loss": 0.6738,
+      "reward": 2.5854490995407104,
+      "reward_std": 0.7286294102668762,
+      "rewards/accuracy_reward": 0.7291666865348816,
+      "rewards/reasoning_steps_reward": 0.9704861491918564,
+      "rewards/repetition_penalty_reward": -0.07123497780412436,
+      "rewards/tag_count_reward": 0.9570312649011612,
+      "step": 577
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 201.0989646911621,
+      "epoch": 0.8668916385451818,
+      "grad_norm": 2.135809055954815,
+      "kl": 0.3212890625,
+      "learning_rate": 1.046305300751811e-06,
+      "loss": 0.679,
+      "reward": 2.635705530643463,
+      "reward_std": 0.49747517332434654,
+      "rewards/accuracy_reward": 0.7395833507180214,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.05787099711596966,
+      "rewards/tag_count_reward": 0.9713542014360428,
+      "step": 578
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.46875381469727,
+      "epoch": 0.8683914510686164,
+      "grad_norm": 1.1578027094404484,
+      "kl": 0.3466796875,
+      "learning_rate": 1.0230725133697495e-06,
+      "loss": 0.7601,
+      "reward": 2.575222373008728,
+      "reward_std": 0.6852890402078629,
+      "rewards/accuracy_reward": 0.7500000149011612,
+      "rewards/reasoning_steps_reward": 0.9652777761220932,
+      "rewards/repetition_penalty_reward": -0.0905762929469347,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 579
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 216.43750762939453,
+      "epoch": 0.869891263592051,
+      "grad_norm": 0.7213394754759388,
+      "kl": 0.404296875,
+      "learning_rate": 1.0000866553996436e-06,
+      "loss": 0.5595,
+      "reward": 2.5885783433914185,
+      "reward_std": 0.6372941508889198,
+      "rewards/accuracy_reward": 0.7500000149011612,
+      "rewards/reasoning_steps_reward": 0.9531250149011612,
+      "rewards/repetition_penalty_reward": -0.06767181493341923,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 580
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 193.3697967529297,
+      "epoch": 0.8713910761154856,
+      "grad_norm": 352.68134893034215,
+      "kl": 4.25341796875,
+      "learning_rate": 9.773483591162203e-07,
+      "loss": 1.1519,
+      "reward": 2.7451828718185425,
+      "reward_std": 0.4861754924058914,
+      "rewards/accuracy_reward": 0.8593750149011612,
+      "rewards/reasoning_steps_reward": 0.9722222238779068,
+      "rewards/repetition_penalty_reward": -0.05646645650267601,
+      "rewards/tag_count_reward": 0.9700520932674408,
+      "step": 581
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.60938262939453,
+      "epoch": 0.8728908886389202,
+      "grad_norm": 1.0240611001356141,
+      "kl": 0.33935546875,
+      "learning_rate": 9.548582499845015e-07,
+      "loss": 0.4393,
+      "reward": 2.83210825920105,
+      "reward_std": 0.46040425822138786,
+      "rewards/accuracy_reward": 0.9479167014360428,
+      "rewards/reasoning_steps_reward": 0.9774305522441864,
+      "rewards/repetition_penalty_reward": -0.06329114036634564,
+      "rewards/tag_count_reward": 0.9700520932674408,
+      "step": 582
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 211.89062881469727,
+      "epoch": 0.8743907011623547,
+      "grad_norm": 0.898626007525376,
+      "kl": 0.32861328125,
+      "learning_rate": 9.326169466425916e-07,
+      "loss": 0.7951,
+      "reward": 2.764460861682892,
+      "reward_std": 0.6820637285709381,
+      "rewards/accuracy_reward": 0.9062500149011612,
+      "rewards/reasoning_steps_reward": 0.9687500298023224,
+      "rewards/repetition_penalty_reward": -0.07408086117357016,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 583
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 203.23437881469727,
+      "epoch": 0.8758905136857893,
+      "grad_norm": 1.0744117383136889,
+      "kl": 0.33251953125,
+      "learning_rate": 9.106250608846679e-07,
+      "loss": 0.4911,
+      "reward": 2.625205934047699,
+      "reward_std": 0.5258033722639084,
+      "rewards/accuracy_reward": 0.739583358168602,
+      "rewards/reasoning_steps_reward": 0.9774305671453476,
+      "rewards/repetition_penalty_reward": -0.05795381683856249,
+      "rewards/tag_count_reward": 0.966145858168602,
+      "step": 584
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 251.5260467529297,
+      "epoch": 0.8773903262092239,
+      "grad_norm": 1.325096473638462,
+      "kl": 0.3818359375,
+      "learning_rate": 8.888831976441481e-07,
+      "loss": 0.5365,
+      "reward": 2.3645836114883423,
+      "reward_std": 0.6718242466449738,
+      "rewards/accuracy_reward": 0.5833333432674408,
+      "rewards/reasoning_steps_reward": 0.9531250149011612,
+      "rewards/repetition_penalty_reward": -0.10026027075946331,
+      "rewards/tag_count_reward": 0.9283854365348816,
+      "step": 585
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.04166793823242,
+      "epoch": 0.8788901387326584,
+      "grad_norm": 0.7704092563337279,
+      "kl": 0.35693359375,
+      "learning_rate": 8.673919549770483e-07,
+      "loss": 0.6,
+      "reward": 2.743525743484497,
+      "reward_std": 0.6834591180086136,
+      "rewards/accuracy_reward": 0.9010416865348816,
+      "rewards/reasoning_steps_reward": 0.96180559694767,
+      "rewards/repetition_penalty_reward": -0.07505077961832285,
+      "rewards/tag_count_reward": 0.9557292014360428,
+      "step": 586
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 230.9010467529297,
+      "epoch": 0.880389951256093,
+      "grad_norm": 5.725122628830429,
+      "kl": 0.403564453125,
+      "learning_rate": 8.461519240455362e-07,
+      "loss": 0.6512,
+      "reward": 2.6582990884780884,
+      "reward_std": 0.676504597067833,
+      "rewards/accuracy_reward": 0.8281250149011612,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.08388850279152393,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 587
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 219.5572967529297,
+      "epoch": 0.8818897637795275,
+      "grad_norm": 0.6438408724607655,
+      "kl": 0.33056640625,
+      "learning_rate": 8.251636891016702e-07,
+      "loss": 0.6033,
+      "reward": 2.701736092567444,
+      "reward_std": 0.5977243855595589,
+      "rewards/accuracy_reward": 0.8489583432674408,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.074305709451437,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 588
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 205.6458396911621,
+      "epoch": 0.8833895763029621,
+      "grad_norm": 0.6460605343575523,
+      "kl": 0.360595703125,
+      "learning_rate": 8.044278274713246e-07,
+      "loss": 0.6139,
+      "reward": 2.7414156198501587,
+      "reward_std": 0.5733753256499767,
+      "rewards/accuracy_reward": 0.8697916865348816,
+      "rewards/reasoning_steps_reward": 0.9739583730697632,
+      "rewards/repetition_penalty_reward": -0.06717826426029205,
+      "rewards/tag_count_reward": 0.9648437649011612,
+      "step": 589
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.70313262939453,
+      "epoch": 0.8848893888263967,
+      "grad_norm": 0.7977081741142966,
+      "kl": 0.2939453125,
+      "learning_rate": 7.839449095383111e-07,
+      "loss": 0.6594,
+      "reward": 2.5514962673187256,
+      "reward_std": 0.6692121252417564,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9635417014360428,
+      "rewards/repetition_penalty_reward": -0.07741007022559643,
+      "rewards/tag_count_reward": 0.95703125,
+      "step": 590
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.4739646911621,
+      "epoch": 0.8863892013498312,
+      "grad_norm": 0.6568761050630232,
+      "kl": 0.31201171875,
+      "learning_rate": 7.637154987286888e-07,
+      "loss": 0.4408,
+      "reward": 2.6473821997642517,
+      "reward_std": 0.49707260727882385,
+      "rewards/accuracy_reward": 0.7656250149011612,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.0652914484962821,
+      "rewards/tag_count_reward": 0.9713541865348816,
+      "step": 591
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.76562881469727,
+      "epoch": 0.8878890138732658,
+      "grad_norm": 1.190854319926877,
+      "kl": 0.37451171875,
+      "learning_rate": 7.437401514952646e-07,
+      "loss": 0.8083,
+      "reward": 2.5822657346725464,
+      "reward_std": 0.691101536154747,
+      "rewards/accuracy_reward": 0.7552083432674408,
+      "rewards/reasoning_steps_reward": 0.973958358168602,
+      "rewards/repetition_penalty_reward": -0.0974218100309372,
+      "rewards/tag_count_reward": 0.9505208432674408,
+      "step": 592
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 220.5312614440918,
+      "epoch": 0.8893888263967004,
+      "grad_norm": 0.9514705424643457,
+      "kl": 0.33349609375,
+      "learning_rate": 7.240194173022941e-07,
+      "loss": 0.5884,
+      "reward": 2.755159914493561,
+      "reward_std": 0.5876848474144936,
+      "rewards/accuracy_reward": 0.9218750149011612,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.08945825602859259,
+      "rewards/tag_count_reward": 0.9557292014360428,
+      "step": 593
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 208.0208396911621,
+      "epoch": 0.890888638920135,
+      "grad_norm": 3.0869314018149523,
+      "kl": 0.38623046875,
+      "learning_rate": 7.045538386103579e-07,
+      "loss": 0.4621,
+      "reward": 2.473863959312439,
+      "reward_std": 0.5747430324554443,
+      "rewards/accuracy_reward": 0.6093750149011612,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.06650061067193747,
+      "rewards/tag_count_reward": 0.962239608168602,
+      "step": 594
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.0052146911621,
+      "epoch": 0.8923884514435696,
+      "grad_norm": 0.7131069872089879,
+      "kl": 0.281494140625,
+      "learning_rate": 6.853439508614412e-07,
+      "loss": 0.4492,
+      "reward": 2.5537226796150208,
+      "reward_std": 0.550928995013237,
+      "rewards/accuracy_reward": 0.6927083432674408,
+      "rewards/reasoning_steps_reward": 0.9704861491918564,
+      "rewards/repetition_penalty_reward": -0.07040940225124359,
+      "rewards/tag_count_reward": 0.9609375149011612,
+      "step": 595
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.3333396911621,
+      "epoch": 0.8938882639670042,
+      "grad_norm": 1.0361329256509342,
+      "kl": 0.32666015625,
+      "learning_rate": 6.663902824642132e-07,
+      "loss": 0.7393,
+      "reward": 2.6442973017692566,
+      "reward_std": 0.7471431717276573,
+      "rewards/accuracy_reward": 0.817708358168602,
+      "rewards/reasoning_steps_reward": 0.9635416716337204,
+      "rewards/repetition_penalty_reward": -0.08226529462262988,
+      "rewards/tag_count_reward": 0.9453125298023224,
+      "step": 596
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 197.64583587646484,
+      "epoch": 0.8953880764904387,
+      "grad_norm": 0.8179894329191135,
+      "kl": 0.2880859375,
+      "learning_rate": 6.47693354779484e-07,
+      "loss": 0.3874,
+      "reward": 2.6881872415542603,
+      "reward_std": 0.4334303140640259,
+      "rewards/accuracy_reward": 0.7864583432674408,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.053132264874875546,
+      "rewards/tag_count_reward": 0.9791667014360428,
+      "step": 597
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 226.76562881469727,
+      "epoch": 0.8968878890138733,
+      "grad_norm": 1.1693577356428269,
+      "kl": 0.33154296875,
+      "learning_rate": 6.29253682105866e-07,
+      "loss": 0.4538,
+      "reward": 2.6256433725357056,
+      "reward_std": 0.4847990833222866,
+      "rewards/accuracy_reward": 0.7812500298023224,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.08789830794557929,
+      "rewards/tag_count_reward": 0.9531250149011612,
+      "step": 598
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 200.71354293823242,
+      "epoch": 0.8983877015373078,
+      "grad_norm": 53.704269130642075,
+      "kl": 0.4990234375,
+      "learning_rate": 6.110717716656289e-07,
+      "loss": 0.4649,
+      "reward": 2.6057077050209045,
+      "reward_std": 0.5076233521103859,
+      "rewards/accuracy_reward": 0.7187500223517418,
+      "rewards/reasoning_steps_reward": 0.9739583432674408,
+      "rewards/repetition_penalty_reward": -0.058354973793029785,
+      "rewards/tag_count_reward": 0.9713541716337204,
+      "step": 599
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.81250381469727,
+      "epoch": 0.8998875140607424,
+      "grad_norm": 1.8065228297910518,
+      "kl": 0.36083984375,
+      "learning_rate": 5.931481235907466e-07,
+      "loss": 0.5947,
+      "reward": 2.6084994673728943,
+      "reward_std": 0.6163917481899261,
+      "rewards/accuracy_reward": 0.7395833432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0698860315605998,
+      "rewards/tag_count_reward": 0.9596354216337204,
+      "step": 600
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 242.0989646911621,
+      "epoch": 0.901387326584177,
+      "grad_norm": 0.7642591170420632,
+      "kl": 0.32080078125,
+      "learning_rate": 5.754832309091362e-07,
+      "loss": 0.7586,
+      "reward": 2.654149353504181,
+      "reward_std": 0.6473118215799332,
+      "rewards/accuracy_reward": 0.8281250149011612,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.10366316325962543,
+      "rewards/tag_count_reward": 0.9505208730697632,
+      "step": 601
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 197.6666717529297,
+      "epoch": 0.9028871391076115,
+      "grad_norm": 0.782148037165739,
+      "kl": 0.28662109375,
+      "learning_rate": 5.580775795311033e-07,
+      "loss": 0.541,
+      "reward": 2.641181170940399,
+      "reward_std": 0.4843425452709198,
+      "rewards/accuracy_reward": 0.7656250298023224,
+      "rewards/reasoning_steps_reward": 0.9687500149011612,
+      "rewards/repetition_penalty_reward": -0.05543356016278267,
+      "rewards/tag_count_reward": 0.9622395932674408,
+      "step": 602
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 247.57812881469727,
+      "epoch": 0.9043869516310461,
+      "grad_norm": 0.6409249324968989,
+      "kl": 0.29345703125,
+      "learning_rate": 5.409316482359694e-07,
+      "loss": 0.7317,
+      "reward": 2.541344404220581,
+      "reward_std": 0.7429018467664719,
+      "rewards/accuracy_reward": 0.723958358168602,
+      "rewards/reasoning_steps_reward": 0.9600694924592972,
+      "rewards/repetition_penalty_reward": -0.09320436045527458,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 603
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 207.79688262939453,
+      "epoch": 0.9058867641544807,
+      "grad_norm": 0.8747881283239337,
+      "kl": 0.3603515625,
+      "learning_rate": 5.240459086589056e-07,
+      "loss": 0.7424,
+      "reward": 2.6282320618629456,
+      "reward_std": 0.7034382522106171,
+      "rewards/accuracy_reward": 0.7968750298023224,
+      "rewards/reasoning_steps_reward": 0.9548611342906952,
+      "rewards/repetition_penalty_reward": -0.07662920840084553,
+      "rewards/tag_count_reward": 0.9531250149011612,
+      "step": 604
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 236.0885467529297,
+      "epoch": 0.9073865766779152,
+      "grad_norm": 1.25088846306247,
+      "kl": 0.345703125,
+      "learning_rate": 5.074208252779589e-07,
+      "loss": 0.6663,
+      "reward": 2.4092161655426025,
+      "reward_std": 0.7467872053384781,
+      "rewards/accuracy_reward": 0.583333358168602,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.08427347056567669,
+      "rewards/tag_count_reward": 0.946614608168602,
+      "step": 605
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 212.01562881469727,
+      "epoch": 0.9088863892013498,
+      "grad_norm": 0.9993410900248464,
+      "kl": 0.33935546875,
+      "learning_rate": 4.910568554012751e-07,
+      "loss": 0.6047,
+      "reward": 2.4471018314361572,
+      "reward_std": 0.4864268973469734,
+      "rewards/accuracy_reward": 0.5572917014360428,
+      "rewards/reasoning_steps_reward": 0.987847238779068,
+      "rewards/repetition_penalty_reward": -0.06808935943990946,
+      "rewards/tag_count_reward": 0.970052108168602,
+      "step": 606
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.16146850585938,
+      "epoch": 0.9103862017247843,
+      "grad_norm": 1.6025896501489152,
+      "kl": 0.474609375,
+      "learning_rate": 4.749544491545199e-07,
+      "loss": 0.651,
+      "reward": 2.516863226890564,
+      "reward_std": 0.6264981552958488,
+      "rewards/accuracy_reward": 0.6770833507180214,
+      "rewards/reasoning_steps_reward": 0.9670139253139496,
+      "rewards/repetition_penalty_reward": -0.08166120201349258,
+      "rewards/tag_count_reward": 0.954427108168602,
+      "step": 607
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 213.46875762939453,
+      "epoch": 0.911886014248219,
+      "grad_norm": 0.8390719889566893,
+      "kl": 0.30908203125,
+      "learning_rate": 4.591140494684965e-07,
+      "loss": 0.3055,
+      "reward": 2.4858875274658203,
+      "reward_std": 0.4557424336671829,
+      "rewards/accuracy_reward": 0.5989583432674408,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.060987500473856926,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 608
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.48958587646484,
+      "epoch": 0.9133858267716536,
+      "grad_norm": 0.603093011230838,
+      "kl": 0.2841796875,
+      "learning_rate": 4.435360920669618e-07,
+      "loss": 0.5458,
+      "reward": 2.7762425541877747,
+      "reward_std": 0.5177741958759725,
+      "rewards/accuracy_reward": 0.9010416865348816,
+      "rewards/reasoning_steps_reward": 0.9774305522441864,
+      "rewards/repetition_penalty_reward": -0.06577137997373939,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 609
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 224.37500381469727,
+      "epoch": 0.9148856392950881,
+      "grad_norm": 1.0751560656614039,
+      "kl": 0.35693359375,
+      "learning_rate": 4.282210054546454e-07,
+      "loss": 0.6732,
+      "reward": 2.5695890188217163,
+      "reward_std": 0.6582420766353607,
+      "rewards/accuracy_reward": 0.7187500149011612,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.08535904064774513,
+      "rewards/tag_count_reward": 0.9570312798023224,
+      "step": 610
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 231.01563262939453,
+      "epoch": 0.9163854518185227,
+      "grad_norm": 0.7771230124140907,
+      "kl": 0.32470703125,
+      "learning_rate": 4.1316921090545305e-07,
+      "loss": 0.5413,
+      "reward": 2.4763482809066772,
+      "reward_std": 0.6009985208511353,
+      "rewards/accuracy_reward": 0.6354166865348816,
+      "rewards/reasoning_steps_reward": 0.9739583432674408,
+      "rewards/repetition_penalty_reward": -0.09136011637747288,
+      "rewards/tag_count_reward": 0.9583333432674408,
+      "step": 611
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.92708587646484,
+      "epoch": 0.9178852643419573,
+      "grad_norm": 1.1584104652860914,
+      "kl": 0.359375,
+      "learning_rate": 3.9838112245088934e-07,
+      "loss": 0.5215,
+      "reward": 2.5999078154563904,
+      "reward_std": 0.6818754225969315,
+      "rewards/accuracy_reward": 0.7500000149011612,
+      "rewards/reasoning_steps_reward": 0.9670139253139496,
+      "rewards/repetition_penalty_reward": -0.07283534575253725,
+      "rewards/tag_count_reward": 0.9557291865348816,
+      "step": 612
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.7604217529297,
+      "epoch": 0.9193850768653918,
+      "grad_norm": 0.7260908756916117,
+      "kl": 0.314453125,
+      "learning_rate": 3.8385714686866137e-07,
+      "loss": 0.5458,
+      "reward": 2.6635963916778564,
+      "reward_std": 0.5627726316452026,
+      "rewards/accuracy_reward": 0.802083358168602,
+      "rewards/reasoning_steps_reward": 0.9756944477558136,
+      "rewards/repetition_penalty_reward": -0.07381692994385958,
+      "rewards/tag_count_reward": 0.9596354365348816,
+      "step": 613
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 237.6979217529297,
+      "epoch": 0.9208848893888264,
+      "grad_norm": 0.9565075434583165,
+      "kl": 0.298828125,
+      "learning_rate": 3.695976836714932e-07,
+      "loss": 0.503,
+      "reward": 2.370998799800873,
+      "reward_std": 0.6577398786321282,
+      "rewards/accuracy_reward": 0.5572916939854622,
+      "rewards/reasoning_steps_reward": 0.9600694626569748,
+      "rewards/repetition_penalty_reward": -0.09297691145911813,
+      "rewards/tag_count_reward": 0.946614608168602,
+      "step": 614
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 229.15625381469727,
+      "epoch": 0.922384701912261,
+      "grad_norm": 1.135581810486635,
+      "kl": 0.34228515625,
+      "learning_rate": 3.556031250961356e-07,
+      "loss": 0.5926,
+      "reward": 2.516046404838562,
+      "reward_std": 0.5974976867437363,
+      "rewards/accuracy_reward": 0.6770833432674408,
+      "rewards/reasoning_steps_reward": 0.9670139253139496,
+      "rewards/repetition_penalty_reward": -0.08377996645867825,
+      "rewards/tag_count_reward": 0.9557291865348816,
+      "step": 615
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 254.8489646911621,
+      "epoch": 0.9238845144356955,
+      "grad_norm": 0.8735579754836578,
+      "kl": 0.34765625,
+      "learning_rate": 3.4187385609257275e-07,
+      "loss": 0.7876,
+      "reward": 2.537445366382599,
+      "reward_std": 0.8377581238746643,
+      "rewards/accuracy_reward": 0.7604166865348816,
+      "rewards/reasoning_steps_reward": 0.960069477558136,
+      "rewards/repetition_penalty_reward": -0.11533251963555813,
+      "rewards/tag_count_reward": 0.9322917014360428,
+      "step": 616
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 220.47396087646484,
+      "epoch": 0.9253843269591301,
+      "grad_norm": 0.7011606608162222,
+      "kl": 0.40087890625,
+      "learning_rate": 3.284102543134426e-07,
+      "loss": 0.4159,
+      "reward": 2.5848045349121094,
+      "reward_std": 0.5568958222866058,
+      "rewards/accuracy_reward": 0.6979166865348816,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.06189696677029133,
+      "rewards/tag_count_reward": 0.9713541865348816,
+      "step": 617
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 194.4947967529297,
+      "epoch": 0.9268841394825647,
+      "grad_norm": 0.779792570829681,
+      "kl": 0.267578125,
+      "learning_rate": 3.152126901036401e-07,
+      "loss": 0.3907,
+      "reward": 2.7681113481521606,
+      "reward_std": 0.4018867686390877,
+      "rewards/accuracy_reward": 0.8385416865348816,
+      "rewards/reasoning_steps_reward": 0.9947916716337204,
+      "rewards/repetition_penalty_reward": -0.04438872542232275,
+      "rewards/tag_count_reward": 0.9791666716337204,
+      "step": 618
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.7916717529297,
+      "epoch": 0.9283839520059992,
+      "grad_norm": 0.9727936242036457,
+      "kl": 0.268310546875,
+      "learning_rate": 3.0228152649013133e-07,
+      "loss": 0.5183,
+      "reward": 2.6839698553085327,
+      "reward_std": 0.5690836161375046,
+      "rewards/accuracy_reward": 0.8229166865348816,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.0716724512167275,
+      "rewards/tag_count_reward": 0.9570312798023224,
+      "step": 619
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.92708587646484,
+      "epoch": 0.9298837645294338,
+      "grad_norm": 950.8548852246184,
+      "kl": 19.261474609375,
+      "learning_rate": 2.896171191719743e-07,
+      "loss": 2.7638,
+      "reward": 2.5930131673812866,
+      "reward_std": 0.6507796198129654,
+      "rewards/accuracy_reward": 0.7343750149011612,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.06931330915540457,
+      "rewards/tag_count_reward": 0.9609375149011612,
+      "step": 620
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 172.92708587646484,
+      "epoch": 0.9313835770528683,
+      "grad_norm": 0.977711243590189,
+      "kl": 0.29345703125,
+      "learning_rate": 2.772198165105267e-07,
+      "loss": 0.2329,
+      "reward": 2.6393333673477173,
+      "reward_std": 0.36761191859841347,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.9913194626569748,
+      "rewards/repetition_penalty_reward": -0.0316736598033458,
+      "rewards/tag_count_reward": 0.9869791716337204,
+      "step": 621
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.26562881469727,
+      "epoch": 0.932883389576303,
+      "grad_norm": 1.0412352430260128,
+      "kl": 0.42578125,
+      "learning_rate": 2.6508995951986526e-07,
+      "loss": 0.6586,
+      "reward": 2.772718071937561,
+      "reward_std": 0.5336792543530464,
+      "rewards/accuracy_reward": 0.8906250149011612,
+      "rewards/reasoning_steps_reward": 0.9809027761220932,
+      "rewards/repetition_penalty_reward": -0.06495565082877874,
+      "rewards/tag_count_reward": 0.966145858168602,
+      "step": 622
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 232.16667556762695,
+      "epoch": 0.9343832020997376,
+      "grad_norm": 0.7362485021348766,
+      "kl": 0.3427734375,
+      "learning_rate": 2.532278818574108e-07,
+      "loss": 0.7523,
+      "reward": 2.6175056099891663,
+      "reward_std": 0.7539031505584717,
+      "rewards/accuracy_reward": 0.817708358168602,
+      "rewards/reasoning_steps_reward": 0.9548611491918564,
+      "rewards/repetition_penalty_reward": -0.09647011943161488,
+      "rewards/tag_count_reward": 0.9414062649011612,
+      "step": 623
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 201.53646087646484,
+      "epoch": 0.9358830146231721,
+      "grad_norm": 1.5000993664801836,
+      "kl": 0.28955078125,
+      "learning_rate": 2.4163390981474354e-07,
+      "loss": 0.4566,
+      "reward": 2.5779688954353333,
+      "reward_std": 0.4579782895743847,
+      "rewards/accuracy_reward": 0.6718750149011612,
+      "rewards/reasoning_steps_reward": 0.975694477558136,
+      "rewards/repetition_penalty_reward": -0.048767429776489735,
+      "rewards/tag_count_reward": 0.9791666865348816,
+      "step": 624
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 212.2552146911621,
+      "epoch": 0.9373828271466067,
+      "grad_norm": 1.0373325356696304,
+      "kl": 0.3564453125,
+      "learning_rate": 2.3030836230863108e-07,
+      "loss": 0.5518,
+      "reward": 2.4835113286972046,
+      "reward_std": 0.6010262817144394,
+      "rewards/accuracy_reward": 0.6354166939854622,
+      "rewards/reasoning_steps_reward": 0.9618056118488312,
+      "rewards/repetition_penalty_reward": -0.06813805643469095,
+      "rewards/tag_count_reward": 0.954427108168602,
+      "step": 625
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 225.14062881469727,
+      "epoch": 0.9388826396700413,
+      "grad_norm": 0.8742390265642193,
+      "kl": 0.3603515625,
+      "learning_rate": 2.192515508722559e-07,
+      "loss": 0.5906,
+      "reward": 2.5894395112991333,
+      "reward_std": 0.679242342710495,
+      "rewards/accuracy_reward": 0.7760416865348816,
+      "rewards/reasoning_steps_reward": 0.9461805820465088,
+      "rewards/repetition_penalty_reward": -0.07809533644467592,
+      "rewards/tag_count_reward": 0.9453125149011612,
+      "step": 626
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.52084350585938,
+      "epoch": 0.9403824521934758,
+      "grad_norm": 1.1468331051712704,
+      "kl": 0.361328125,
+      "learning_rate": 2.08463779646646e-07,
+      "loss": 0.4132,
+      "reward": 2.633046269416809,
+      "reward_std": 0.46498178830370307,
+      "rewards/accuracy_reward": 0.7708333432674408,
+      "rewards/reasoning_steps_reward": 0.960069477558136,
+      "rewards/repetition_penalty_reward": -0.06400250736624002,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 627
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 211.15625381469727,
+      "epoch": 0.9418822647169104,
+      "grad_norm": 1.4986712256290156,
+      "kl": 0.365234375,
+      "learning_rate": 1.979453453723057e-07,
+      "loss": 0.6512,
+      "reward": 2.6833993792533875,
+      "reward_std": 0.7203980311751366,
+      "rewards/accuracy_reward": 0.833333358168602,
+      "rewards/reasoning_steps_reward": 0.958333358168602,
+      "rewards/repetition_penalty_reward": -0.06529862061142921,
+      "rewards/tag_count_reward": 0.9570312649011612,
+      "step": 628
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 229.2447967529297,
+      "epoch": 0.943382077240345,
+      "grad_norm": 0.7753884291535859,
+      "kl": 0.2939453125,
+      "learning_rate": 1.8769653738105797e-07,
+      "loss": 0.499,
+      "reward": 2.6547706723213196,
+      "reward_std": 0.4733446016907692,
+      "rewards/accuracy_reward": 0.786458358168602,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.077868377789855,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 629
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.28125381469727,
+      "epoch": 0.9448818897637795,
+      "grad_norm": 1.3985799794968885,
+      "kl": 0.2900390625,
+      "learning_rate": 1.7771763758808403e-07,
+      "loss": 0.6038,
+      "reward": 2.550901174545288,
+      "reward_std": 0.5799007415771484,
+      "rewards/accuracy_reward": 0.677083358168602,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.07236280757933855,
+      "rewards/tag_count_reward": 0.9635416865348816,
+      "step": 630
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 209.70313262939453,
+      "epoch": 0.9463817022872141,
+      "grad_norm": 1.4263168938028885,
+      "kl": 0.3466796875,
+      "learning_rate": 1.6800892048416618e-07,
+      "loss": 0.7189,
+      "reward": 2.665283203125,
+      "reward_std": 0.6292329207062721,
+      "rewards/accuracy_reward": 0.8072917014360428,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.0756023209542036,
+      "rewards/tag_count_reward": 0.9544271230697632,
+      "step": 631
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.7864646911621,
+      "epoch": 0.9478815148106486,
+      "grad_norm": 0.7415274923254211,
+      "kl": 0.3037109375,
+      "learning_rate": 1.5857065312814058e-07,
+      "loss": 0.5161,
+      "reward": 2.5291183590888977,
+      "reward_std": 0.5103181153535843,
+      "rewards/accuracy_reward": 0.651041679084301,
+      "rewards/reasoning_steps_reward": 0.973958358168602,
+      "rewards/repetition_penalty_reward": -0.06332961097359657,
+      "rewards/tag_count_reward": 0.9674479365348816,
+      "step": 632
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 211.75000381469727,
+      "epoch": 0.9493813273340832,
+      "grad_norm": 1.3628095814309313,
+      "kl": 0.322265625,
+      "learning_rate": 1.4940309513955088e-07,
+      "loss": 0.6802,
+      "reward": 2.667730212211609,
+      "reward_std": 0.6961818635463715,
+      "rewards/accuracy_reward": 0.8281250149011612,
+      "rewards/reasoning_steps_reward": 0.958333358168602,
+      "rewards/repetition_penalty_reward": -0.07315511163324118,
+      "rewards/tag_count_reward": 0.954427108168602,
+      "step": 633
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 220.2291717529297,
+      "epoch": 0.9508811398575178,
+      "grad_norm": 0.9795351526659545,
+      "kl": 0.369140625,
+      "learning_rate": 1.405064986915028e-07,
+      "loss": 0.6991,
+      "reward": 2.5635343194007874,
+      "reward_std": 0.6203908771276474,
+      "rewards/accuracy_reward": 0.708333358168602,
+      "rewards/reasoning_steps_reward": 0.9826389104127884,
+      "rewards/repetition_penalty_reward": -0.08056299947202206,
+      "rewards/tag_count_reward": 0.9531250298023224,
+      "step": 634
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 194.6927146911621,
+      "epoch": 0.9523809523809523,
+      "grad_norm": 0.8730687949726732,
+      "kl": 0.32666015625,
+      "learning_rate": 1.3188110850373527e-07,
+      "loss": 0.4296,
+      "reward": 2.716920852661133,
+      "reward_std": 0.47606247290968895,
+      "rewards/accuracy_reward": 0.8281250149011612,
+      "rewards/reasoning_steps_reward": 0.967013880610466,
+      "rewards/repetition_penalty_reward": -0.050874427892267704,
+      "rewards/tag_count_reward": 0.9726562649011612,
+      "step": 635
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 221.78125381469727,
+      "epoch": 0.953880764904387,
+      "grad_norm": 1.1050385339520583,
+      "kl": 0.4296875,
+      "learning_rate": 1.2352716183588022e-07,
+      "loss": 0.7716,
+      "reward": 2.4985339045524597,
+      "reward_std": 0.7328899428248405,
+      "rewards/accuracy_reward": 0.6822916865348816,
+      "rewards/reasoning_steps_reward": 0.954861119389534,
+      "rewards/repetition_penalty_reward": -0.08393146842718124,
+      "rewards/tag_count_reward": 0.9453125298023224,
+      "step": 636
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 220.67188262939453,
+      "epoch": 0.9553805774278216,
+      "grad_norm": 1.7375333895919391,
+      "kl": 0.349609375,
+      "learning_rate": 1.1544488848094338e-07,
+      "loss": 0.5867,
+      "reward": 2.678426444530487,
+      "reward_std": 0.5324868559837341,
+      "rewards/accuracy_reward": 0.7968750298023224,
+      "rewards/reasoning_steps_reward": 0.9843750298023224,
+      "rewards/repetition_penalty_reward": -0.07287566550076008,
+      "rewards/tag_count_reward": 0.9700520932674408,
+      "step": 637
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 210.47916793823242,
+      "epoch": 0.9568803899512561,
+      "grad_norm": 1.4004750786860491,
+      "kl": 0.3212890625,
+      "learning_rate": 1.0763451075897713e-07,
+      "loss": 0.8042,
+      "reward": 2.6482399106025696,
+      "reward_std": 0.6535500586032867,
+      "rewards/accuracy_reward": 0.7812500149011612,
+      "rewards/reasoning_steps_reward": 0.9756944626569748,
+      "rewards/repetition_penalty_reward": -0.07485033478587866,
+      "rewards/tag_count_reward": 0.966145858168602,
+      "step": 638
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 190.1354217529297,
+      "epoch": 0.9583802024746907,
+      "grad_norm": 1.0307017049735754,
+      "kl": 0.316650390625,
+      "learning_rate": 1.0009624351097313e-07,
+      "loss": 0.4842,
+      "reward": 2.681637763977051,
+      "reward_std": 0.5066114142537117,
+      "rewards/accuracy_reward": 0.7812500298023224,
+      "rewards/reasoning_steps_reward": 0.9739583432674408,
+      "rewards/repetition_penalty_reward": -0.0462268297560513,
+      "rewards/tag_count_reward": 0.9726562798023224,
+      "step": 639
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 196.38021087646484,
+      "epoch": 0.9598800149981253,
+      "grad_norm": 0.7851546590808128,
+      "kl": 0.29345703125,
+      "learning_rate": 9.283029409294263e-08,
+      "loss": 0.5919,
+      "reward": 2.8382840156555176,
+      "reward_std": 0.507376492023468,
+      "rewards/accuracy_reward": 0.942708358168602,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.05624724645167589,
+      "rewards/tag_count_reward": 0.9726562649011612,
+      "step": 640
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 206.6770896911621,
+      "epoch": 0.9613798275215598,
+      "grad_norm": 0.8675967780931345,
+      "kl": 0.32421875,
+      "learning_rate": 8.583686237022105e-08,
+      "loss": 0.5485,
+      "reward": 2.5225971341133118,
+      "reward_std": 0.504936508834362,
+      "rewards/accuracy_reward": 0.6354166716337204,
+      "rewards/reasoning_steps_reward": 0.982638880610466,
+      "rewards/repetition_penalty_reward": -0.06551066134124994,
+      "rewards/tag_count_reward": 0.9700520932674408,
+      "step": 641
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 242.56250762939453,
+      "epoch": 0.9628796400449944,
+      "grad_norm": 1.4191184069887932,
+      "kl": 0.40283203125,
+      "learning_rate": 7.911614071196671e-08,
+      "loss": 0.6117,
+      "reward": 2.552360415458679,
+      "reward_std": 0.6927186399698257,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/reasoning_steps_reward": 0.953125,
+      "rewards/repetition_penalty_reward": -0.09477510303258896,
+      "rewards/tag_count_reward": 0.9440104514360428,
+      "step": 642
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.17188262939453,
+      "epoch": 0.9643794525684289,
+      "grad_norm": 0.9682440226139685,
+      "kl": 0.35498046875,
+      "learning_rate": 7.266831398587082e-08,
+      "loss": 0.5619,
+      "reward": 2.6066702604293823,
+      "reward_std": 0.5623406581580639,
+      "rewards/accuracy_reward": 0.7552083432674408,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.07171526318416,
+      "rewards/tag_count_reward": 0.9596354365348816,
+      "step": 643
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 265.9270896911621,
+      "epoch": 0.9658792650918635,
+      "grad_norm": 108.30769155391553,
+      "kl": 1.31689453125,
+      "learning_rate": 6.649355955306802e-08,
+      "loss": 0.7916,
+      "reward": 2.512334644794464,
+      "reward_std": 0.77546027302742,
+      "rewards/accuracy_reward": 0.7135417014360428,
+      "rewards/reasoning_steps_reward": 0.9670139402151108,
+      "rewards/repetition_penalty_reward": -0.10962733440101147,
+      "rewards/tag_count_reward": 0.9414062798023224,
+      "step": 644
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 212.09375762939453,
+      "epoch": 0.9673790776152981,
+      "grad_norm": 0.9073762922270306,
+      "kl": 0.27783203125,
+      "learning_rate": 6.059204726326373e-08,
+      "loss": 0.5087,
+      "reward": 2.4138490557670593,
+      "reward_std": 0.6195433586835861,
+      "rewards/accuracy_reward": 0.5364583507180214,
+      "rewards/reasoning_steps_reward": 0.9895833283662796,
+      "rewards/repetition_penalty_reward": -0.0731301549822092,
+      "rewards/tag_count_reward": 0.9609375149011612,
+      "step": 645
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 205.7291717529297,
+      "epoch": 0.9688788901387326,
+      "grad_norm": 0.6693528675529495,
+      "kl": 0.2998046875,
+      "learning_rate": 5.4963939450057846e-08,
+      "loss": 0.4844,
+      "reward": 2.7782857418060303,
+      "reward_std": 0.5295575931668282,
+      "rewards/accuracy_reward": 0.9010416865348816,
+      "rewards/reasoning_steps_reward": 0.973958358168602,
+      "rewards/repetition_penalty_reward": -0.06025600666180253,
+      "rewards/tag_count_reward": 0.9635416716337204,
+      "step": 646
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 200.9947967529297,
+      "epoch": 0.9703787026621672,
+      "grad_norm": 60.998327736404455,
+      "kl": 48.24267578125,
+      "learning_rate": 4.960939092648165e-08,
+      "loss": 0.6613,
+      "reward": 2.77788108587265,
+      "reward_std": 0.5485429763793945,
+      "rewards/accuracy_reward": 0.9062500298023224,
+      "rewards/reasoning_steps_reward": 0.9670139253139496,
+      "rewards/repetition_penalty_reward": -0.06022655125707388,
+      "rewards/tag_count_reward": 0.9648437649011612,
+      "step": 647
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 226.3489646911621,
+      "epoch": 0.9718785151856018,
+      "grad_norm": 1.0839111689107002,
+      "kl": 0.3359375,
+      "learning_rate": 4.452854898073788e-08,
+      "loss": 0.6235,
+      "reward": 2.5908501744270325,
+      "reward_std": 0.5783855766057968,
+      "rewards/accuracy_reward": 0.7395833432674408,
+      "rewards/reasoning_steps_reward": 0.9826388955116272,
+      "rewards/repetition_penalty_reward": -0.08189292438328266,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 648
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.47917938232422,
+      "epoch": 0.9733783277090363,
+      "grad_norm": 1.1002454367127348,
+      "kl": 0.41015625,
+      "learning_rate": 3.9721553372150665e-08,
+      "loss": 0.6111,
+      "reward": 2.5614100098609924,
+      "reward_std": 0.7080376967787743,
+      "rewards/accuracy_reward": 0.7343750223517418,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.08702756371349096,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 649
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 241.73438262939453,
+      "epoch": 0.974878140232471,
+      "grad_norm": 4.156144128333803,
+      "kl": 2.935546875,
+      "learning_rate": 3.5188536327318554e-08,
+      "loss": 0.533,
+      "reward": 2.4512782096862793,
+      "reward_std": 0.528814010322094,
+      "rewards/accuracy_reward": 0.6354166716337204,
+      "rewards/reasoning_steps_reward": 0.9635416865348816,
+      "rewards/repetition_penalty_reward": -0.09169065579771996,
+      "rewards/tag_count_reward": 0.9440104365348816,
+      "step": 650
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 176.15625381469727,
+      "epoch": 0.9763779527559056,
+      "grad_norm": 0.9986867176141776,
+      "kl": 0.37939453125,
+      "learning_rate": 3.092962253648302e-08,
+      "loss": 0.352,
+      "reward": 2.7320080399513245,
+      "reward_std": 0.40247857104986906,
+      "rewards/accuracy_reward": 0.8072916865348816,
+      "rewards/reasoning_steps_reward": 0.9791666716337204,
+      "rewards/repetition_penalty_reward": -0.03752335952594876,
+      "rewards/tag_count_reward": 0.9830729365348816,
+      "step": 651
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 194.25000381469727,
+      "epoch": 0.9778777652793401,
+      "grad_norm": 0.7209747458961953,
+      "kl": 0.27294921875,
+      "learning_rate": 2.694492915009006e-08,
+      "loss": 0.3393,
+      "reward": 2.59340101480484,
+      "reward_std": 0.36479785293340683,
+      "rewards/accuracy_reward": 0.6979166865348816,
+      "rewards/reasoning_steps_reward": 0.9791666865348816,
+      "rewards/repetition_penalty_reward": -0.05764078535139561,
+      "rewards/tag_count_reward": 0.973958358168602,
+      "step": 652
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 216.52083587646484,
+      "epoch": 0.9793775778027747,
+      "grad_norm": 0.8162473177671608,
+      "kl": 0.30859375,
+      "learning_rate": 2.3234565775575034e-08,
+      "loss": 0.5575,
+      "reward": 2.596774399280548,
+      "reward_std": 0.6674105823040009,
+      "rewards/accuracy_reward": 0.7447917014360428,
+      "rewards/reasoning_steps_reward": 0.9670139253139496,
+      "rewards/repetition_penalty_reward": -0.07466658856719732,
+      "rewards/tag_count_reward": 0.9596354365348816,
+      "step": 653
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 218.06771087646484,
+      "epoch": 0.9808773903262092,
+      "grad_norm": 1.3998005104320381,
+      "kl": 0.35595703125,
+      "learning_rate": 1.9798634474345048e-08,
+      "loss": 0.6469,
+      "reward": 2.451699197292328,
+      "reward_std": 0.6794094815850258,
+      "rewards/accuracy_reward": 0.6093750149011612,
+      "rewards/reasoning_steps_reward": 0.9670139104127884,
+      "rewards/repetition_penalty_reward": -0.07651268597692251,
+      "rewards/tag_count_reward": 0.9518229216337204,
+      "step": 654
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 197.5833396911621,
+      "epoch": 0.9823772028496438,
+      "grad_norm": 0.8512758770399678,
+      "kl": 0.31396484375,
+      "learning_rate": 1.6637229758970087e-08,
+      "loss": 0.4539,
+      "reward": 2.7271097898483276,
+      "reward_std": 0.4929804429411888,
+      "rewards/accuracy_reward": 0.817708358168602,
+      "rewards/reasoning_steps_reward": 0.9843750298023224,
+      "rewards/repetition_penalty_reward": -0.05283822864294052,
+      "rewards/tag_count_reward": 0.9778645932674408,
+      "step": 655
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 221.85417556762695,
+      "epoch": 0.9838770153730784,
+      "grad_norm": 1.2498346639772517,
+      "kl": 0.3740234375,
+      "learning_rate": 1.3750438590586223e-08,
+      "loss": 0.5368,
+      "reward": 2.681597590446472,
+      "reward_std": 0.568534217774868,
+      "rewards/accuracy_reward": 0.833333358168602,
+      "rewards/reasoning_steps_reward": 0.960069477558136,
+      "rewards/repetition_penalty_reward": -0.07404482085257769,
+      "rewards/tag_count_reward": 0.962239608168602,
+      "step": 656
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 215.24480056762695,
+      "epoch": 0.9853768278965129,
+      "grad_norm": 0.8636210438416793,
+      "kl": 0.30517578125,
+      "learning_rate": 1.1138340376501966e-08,
+      "loss": 0.5454,
+      "reward": 2.747868835926056,
+      "reward_std": 0.6248507276177406,
+      "rewards/accuracy_reward": 0.8697916865348816,
+      "rewards/reasoning_steps_reward": 0.9722222536802292,
+      "rewards/repetition_penalty_reward": -0.0602909866720438,
+      "rewards/tag_count_reward": 0.9661458432674408,
+      "step": 657
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 199.9270896911621,
+      "epoch": 0.9868766404199475,
+      "grad_norm": 1.3618095357818232,
+      "kl": 0.3759765625,
+      "learning_rate": 8.801006968012227e-09,
+      "loss": 0.6015,
+      "reward": 2.544634699821472,
+      "reward_std": 0.524773295968771,
+      "rewards/accuracy_reward": 0.6614583544433117,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.06300420686602592,
+      "rewards/tag_count_reward": 0.9687500298023224,
+      "step": 658
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 238.6614646911621,
+      "epoch": 0.9883764529433821,
+      "grad_norm": 0.8680279006911963,
+      "kl": 0.3662109375,
+      "learning_rate": 6.738502658426571e-09,
+      "loss": 0.6921,
+      "reward": 2.585418939590454,
+      "reward_std": 0.7059072330594063,
+      "rewards/accuracy_reward": 0.7812500149011612,
+      "rewards/reasoning_steps_reward": 0.9652777910232544,
+      "rewards/repetition_penalty_reward": -0.09860894456505775,
+      "rewards/tag_count_reward": 0.9375,
+      "step": 659
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 227.03125381469727,
+      "epoch": 0.9898762654668166,
+      "grad_norm": 1.0096929149450597,
+      "kl": 0.33642578125,
+      "learning_rate": 4.950884181295079e-09,
+      "loss": 0.6266,
+      "reward": 2.6765416860580444,
+      "reward_std": 0.5609493404626846,
+      "rewards/accuracy_reward": 0.8281250149011612,
+      "rewards/reasoning_steps_reward": 0.9774305820465088,
+      "rewards/repetition_penalty_reward": -0.0886493306607008,
+      "rewards/tag_count_reward": 0.9596354365348816,
+      "step": 660
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 217.0104217529297,
+      "epoch": 0.9913760779902512,
+      "grad_norm": 0.858588179176331,
+      "kl": 0.3212890625,
+      "learning_rate": 3.4382007088518134e-09,
+      "loss": 0.7794,
+      "reward": 2.381414532661438,
+      "reward_std": 0.6443270593881607,
+      "rewards/accuracy_reward": 0.5468750149011612,
+      "rewards/reasoning_steps_reward": 0.9618055671453476,
+      "rewards/repetition_penalty_reward": -0.0777870174497366,
+      "rewards/tag_count_reward": 0.950520858168602,
+      "step": 661
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 235.0833396911621,
+      "epoch": 0.9928758905136857,
+      "grad_norm": 2.317140366000632,
+      "kl": 0.34619140625,
+      "learning_rate": 2.200493850662566e-09,
+      "loss": 0.7031,
+      "reward": 2.485570192337036,
+      "reward_std": 0.7142708599567413,
+      "rewards/accuracy_reward": 0.692708358168602,
+      "rewards/reasoning_steps_reward": 0.9513888955116272,
+      "rewards/repetition_penalty_reward": -0.09863131493330002,
+      "rewards/tag_count_reward": 0.9401041865348816,
+      "step": 662
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 219.2239646911621,
+      "epoch": 0.9943757030371203,
+      "grad_norm": 0.8136792892151121,
+      "kl": 0.34765625,
+      "learning_rate": 1.2377976524746705e-09,
+      "loss": 0.6723,
+      "reward": 2.587967872619629,
+      "reward_std": 0.66152124106884,
+      "rewards/accuracy_reward": 0.7343750149011612,
+      "rewards/reasoning_steps_reward": 0.9739583432674408,
+      "rewards/repetition_penalty_reward": -0.07609467767179012,
+      "rewards/tag_count_reward": 0.9557291865348816,
+      "step": 663
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 214.48958587646484,
+      "epoch": 0.995875515560555,
+      "grad_norm": 1.3047927636843426,
+      "kl": 0.3359375,
+      "learning_rate": 5.501385952888516e-10,
+      "loss": 0.751,
+      "reward": 2.50458562374115,
+      "reward_std": 0.6785130500793457,
+      "rewards/accuracy_reward": 0.6718750149011612,
+      "rewards/reasoning_steps_reward": 0.960069477558136,
+      "rewards/repetition_penalty_reward": -0.07397346664220095,
+      "rewards/tag_count_reward": 0.9466146230697632,
+      "step": 664
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 224.18750762939453,
+      "epoch": 0.9973753280839895,
+      "grad_norm": 0.6420501828977786,
+      "kl": 0.3134765625,
+      "learning_rate": 1.375355946242607e-10,
+      "loss": 0.6873,
+      "reward": 2.3991820216178894,
+      "reward_std": 0.6460757553577423,
+      "rewards/accuracy_reward": 0.5677083432674408,
+      "rewards/reasoning_steps_reward": 0.9600694626569748,
+      "rewards/repetition_penalty_reward": -0.08041872084140778,
+      "rewards/tag_count_reward": 0.9518229365348816,
+      "step": 665
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 242.2552146911621,
+      "epoch": 0.9988751406074241,
+      "grad_norm": 1.2779320053093417,
+      "kl": 0.5048828125,
+      "learning_rate": 0.0,
+      "loss": 0.8081,
+      "reward": 2.5781980752944946,
+      "reward_std": 0.8304520845413208,
+      "rewards/accuracy_reward": 0.7968750149011612,
+      "rewards/reasoning_steps_reward": 0.9496528059244156,
+      "rewards/repetition_penalty_reward": -0.10322570707648993,
+      "rewards/tag_count_reward": 0.934895858168602,
+      "step": 666
+    },
+    {
+      "epoch": 0.9988751406074241,
+      "step": 666,
       "total_flos": 0.0,
-      "train_loss": 0.6876773679259796,
-      "train_runtime": 4819.3131,
-      "train_samples_per_second": 0.83,
-      "train_steps_per_second": 0.069
+      "train_loss": 566.2065416834479,
+      "train_runtime": 15651.159,
+      "train_samples_per_second": 0.511,
+      "train_steps_per_second": 0.043
     }
   ],
   "logging_steps": 1,
-  "max_steps": 333,
+  "max_steps": 666,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 200,