{"status": "failed", "required": true, "checks": [{"path": "grpo_judge_advantage_via_custom_path", "check": "baseline_loss", "status": "passed", "tolerance": {"atol": 0.0001, "rtol": 0.001}, "measured": {"max_abs_error": 0.0, "reference_max_abs": 0.00714289303869009, "relative_l2_error": 0.0, "cosine_similarity": 1.0}}, {"path": "grpo_judge_advantage_via_custom_path", "check": "baseline_gradient", "status": "passed", "tolerance": {"atol": 0.0001, "rtol": 0.01, "min_cosine": 0.999}, "measured": {"max_abs_error": 0.0, "reference_max_abs": 0.0452161505818367, "relative_l2_error": 0.0, "cosine_similarity": 0.9999999999999998}}, {"path": "grpo_judge_advantage_via_custom_path", "check": "reference_loss", "status": "passed", "tolerance": {"atol": 0.0001, "rtol": 0.001}, "measured": {"max_abs_error": 0.0, "reference_max_abs": 0.00714289303869009, "relative_l2_error": 0.0, "cosine_similarity": 1.0}}, {"path": "grpo_judge_advantage_via_custom_path", "check": "reference_gradient", "status": "passed", "tolerance": {"atol": 0.0001, "rtol": 0.01, "min_cosine": 0.999}, "measured": {"max_abs_error": 0.0, "reference_max_abs": 0.0452161505818367, "relative_l2_error": 0.0, "cosine_similarity": 0.9999999999999998}}, {"path": "sdpo_self_teacher_advantage", "check": "baseline_loss", "status": "passed", "tolerance": {"atol": 0.0001, "rtol": 0.001}, "measured": {"max_abs_error": 0.0, "reference_max_abs": 0.0, "relative_l2_error": null, "cosine_similarity": null}}, {"path": "sdpo_self_teacher_advantage", "check": "baseline_gradient", "status": "unverified", "tolerance": {"atol": 0.0001, "rtol": 0.01, "min_cosine": 0.999}, "measured": {"max_abs_error": 0.0, "reference_max_abs": 0.0, "relative_l2_error": null, "cosine_similarity": null}}, {"path": "sdpo_self_teacher_advantage", "check": "reference_loss", "status": "failed", "tolerance": {"atol": 0.001, "rtol": 0.01}, "measured": {"max_abs_error": 0.13540398702025414, "reference_max_abs": 0.09341718256473541, "relative_l2_error": 1.4494548358533836, "cosine_similarity": -1.0}}, {"path": "sdpo_self_teacher_advantage", "check": "reference_gradient", "status": "passed", "tolerance": {"atol": 0.001, "rtol": 0.02, "min_cosine": 0.995}, "measured": {"max_abs_error": 0.0, "reference_max_abs": 0.04461999982595444, "relative_l2_error": 0.0, "cosine_similarity": 1.0}}], "reason": "sdpo_self_teacher_advantage: reference_loss; sdpo_self_teacher_advantage: baseline_gradient cosine/tolerance", "evidence_source": "executor report; remote execution is not attested", "applicability": "required"}