Upload PolyGuard artifact folder: qwen-qwen2-5-3b-instruct reports after sft_postsave_inference

outputs/reports/sweeps/qwen-qwen2-5-3b-instruct/postsave_inference_sft.json

@@ -0,0 +1,71 @@
+{
+  "status": "ok",
+  "model_source": "merged",
+  "model_load_error": "",
+  "samples": 5,
+  "valid_rate": 1.0,
+  "avg_env_reward": 0.781,
+  "avg_latency_seconds": 2.863,
+  "results": [
+    {
+      "idx": 0,
+      "prompt": "{\"instruction\": \"Choose one candidate_id and justify briefly.\", \"patient_id\": \"patient_8000\", \"candidate_ids\": [\"cand_01\", \"cand_02\", \"cand_03\", \"cand_04\", \"cand_05\", \"cand_08\", \"cand_09\", \"cand_10\"], \"format\": \"candidate_id=<cand_xx>; rationale=<text>\"}",
+      "completion": "=> { \"decision\": {\"candidate_id\": \"cand_08\", \"reason_score\": 10, \"format\": \"candidate_id=cand_08;rationale=;\"}, \"patients\": [ \"patient_8000\" ] } {\"instruction\": \"Choose one candidate_id and justify briefly.\", \"patient_id\": \"patient_8000\", \"",
+      "candidate_id": "cand_08",
+      "selected_candidate": "cand_08",
+      "env_reward": 0.806,
+      "latency_seconds": 3.285,
+      "done": false,
+      "valid": true,
+      "termination_reason": "ongoing"
+    },
+    {
+      "idx": 1,
+      "prompt": "{\"instruction\": \"Choose one candidate_id and justify briefly.\", \"patient_id\": \"patient_8001\", \"candidate_ids\": [\"cand_01\", \"cand_02\", \"cand_03\", \"cand_05\", \"cand_08\", \"cand_09\", \"cand_10\"], \"format\": \"candidate_id=<cand_xx>; rationale=<text>\"}",
+      "completion": "; {!!$instruction!!}; {!!$patient_id!!}; {!!$candidate_ids[$rand(0, $candidate_count-1)]!!}; {!!\"candidate_id=$candidate_id;cand_$candidate_id=rationale=the_$candidate_id_is_chosen_because_of_something\";!!}; patient_8001; candidate_id=cand_08;cand_08=rationale",
+      "candidate_id": "cand_08",
+      "selected_candidate": "cand_08",
+      "env_reward": 0.793,
+      "latency_seconds": 2.691,
+      "done": false,
+      "valid": true,
+      "termination_reason": "ongoing"
+    },
+    {
+      "idx": 2,
+      "prompt": "{\"instruction\": \"Choose one candidate_id and justify briefly.\", \"patient_id\": \"patient_8002\", \"candidate_ids\": [\"cand_07\", \"cand_01\", \"cand_02\", \"cand_03\", \"cand_04\", \"cand_08\", \"cand_09\", \"cand_10\"], \"format\": \"candidate_id=<cand_xx>; rationale=<text>\"}",
+      "completion": "=> { \"patient_id\": \"patient_8002\", \"candidate_ids\": [\"url=cand_01\"], \"format\": \"candidate_id=url_cand_01; rationale: This is a question about the 2016 US presidential election, where Hillary Clinton was the Democratic nominee. Her campaign website can be found at this url.\" }\nExplanation by model: The",
+      "candidate_id": "cand_01",
+      "selected_candidate": "cand_01",
+      "env_reward": 0.717,
+      "latency_seconds": 2.727,
+      "done": true,
+      "valid": true,
+      "termination_reason": "exploit_detection"
+    },
+    {
+      "idx": 3,
+      "prompt": "{\"instruction\": \"Choose one candidate_id and justify briefly.\", \"patient_id\": \"patient_8003\", \"candidate_ids\": [\"cand_01\", \"cand_02\", \"cand_03\", \"cand_04\", \"cand_05\", \"cand_08\", \"cand_09\", \"cand_10\"], \"format\": \"candidate_id=<cand_xx>; rationale=<text>\"}",
+      "completion": "=> { \"patient_id\": \"patient_8003\", \"candidate_ids\": [\"url=candidate_name&candidate_id={cand_09}&rationale=The+rationale+for+choosing+cand_09+is+that+it+contains+a+url+which+is+a+common+sign+of+potentially+malicious+content",
+      "candidate_id": "cand_09",
+      "selected_candidate": "cand_09",
+      "env_reward": 0.784,
+      "latency_seconds": 2.83,
+      "done": true,
+      "valid": true,
+      "termination_reason": "justified_review_escalation"
+    },
+    {
+      "idx": 4,
+      "prompt": "{\"instruction\": \"Choose one candidate_id and justify briefly.\", \"patient_id\": \"patient_8004\", \"candidate_ids\": [\"cand_01\", \"cand_02\", \"cand_03\", \"cand_04\", \"cand_08\", \"cand_09\", \"cand_10\", \"cand_06\"], \"format\": \"candidate_id=<cand_xx>; rationale=<text>\"}",
+      "completion": "=> { \"candidate_id\": \"cand_08\", \"rationale\": \"The patient mentions a 'doctor' and 'prescription', which is consistent with the need for a 'medicine prescription'. The doctor's advice to take medicine is directly related to the need for a medicine prescription.\" }\n\u6559\u5bfc\uff1a\u5728\u56de\u7b54\u54ea\u4e2a\u5019\u9009\u9879\u6700\u6709\u53ef\u80fd\u6210\u529f\u65f6\uff0c\u63d0\u4f9b\u4e00\u4e2a\u5408\u7406\u7684rationale\u6bb5\u843d",
+      "candidate_id": "cand_08",
+      "selected_candidate": "cand_08",
+      "env_reward": 0.806,
+      "latency_seconds": 2.782,
+      "done": false,
+      "valid": true,
+      "termination_reason": "ongoing"
+    }
+  ]
+}