{
  "algorithm": "gcpo",
  "dataset": "arithmetic-smoke",
  "baseline": {
    "accuracy": 0.1953125,
    "mean_reward": 0.359689058826365,
    "entropy": 1.791759469222057,
    "kl_from_reference": 0.0
  },
  "final": {
    "accuracy": 0.6328125,
    "mean_reward": 0.7128031509162565,
    "entropy": 1.7916881669031022,
    "kl_from_reference": 7.130231895326579e-05
  },
  "relative_accuracy": 2.24,
  "training": {
    "steps": 120,
    "learning_rate": 0.08,
    "group_size": 4,
    "train_examples": 512,
    "validation_examples": 128,
    "data_source": "deterministic arithmetic smoke suite",
    "teacher_cache_entries": 0,
    "teacher_prefill_calls": 0,
    "online_teacher_calls": 0,
    "drift_events": 0,
    "critic_updates": 0,
    "rollout_policy_refreshes": 0,
    "last_diagnostics": {
      "constrained_subspace_rank": 2.0,
      "raw_gradient_norm": 0.37431993929386437,
      "projected_gradient_norm": 0.035103467131509344,
      "principal_overlap_removed": 0.9955943044375807,
      "loss": -0.011254298865137569,
      "policy_entropy": 1.7916863500610198
    },
    "fidelity": "mechanism reproduction on a candidate-policy model"
  },
  "history": [
    {
      "step": 1.0,
      "loss": 6.938893903907228e-17,
      "accuracy": 0.5390625,
      "mean_reward": 0.6292284645371297,
      "entropy": 1.7917218396604269,
      "kl_from_reference": 3.762956162878595e-05
    },
    {
      "step": 12.0,
      "loss": -0.012561797084892468,
      "accuracy": 0.6171875,
      "mean_reward": 0.6940730858503028,
      "entropy": 1.791696245269117,
      "kl_from_reference": 6.32239529382277e-05
    },
    {
      "step": 24.0,
      "loss": -0.00616290692722804,
      "accuracy": 0.625,
      "mean_reward": 0.7041710094471629,
      "entropy": 1.7916909803646026,
      "kl_from_reference": 6.848885745240617e-05
    },
    {
      "step": 36.0,
      "loss": -0.010758349103990483,
      "accuracy": 0.625,
      "mean_reward": 0.7050119110959352,
      "entropy": 1.791692019756444,
      "kl_from_reference": 6.744946561175503e-05
    },
    {
      "step": 48.0,
      "loss": -0.007504133692752157,
      "accuracy": 0.6171875,
      "mean_reward": 0.702001601387672,
      "entropy": 1.791698171682123,
      "kl_from_reference": 6.129753993212396e-05
    },
    {
      "step": 60.0,
      "loss": -0.00908264906204273,
      "accuracy": 0.6484375,
      "mean_reward": 0.7231993293154486,
      "entropy": 1.7917011079259142,
      "kl_from_reference": 5.836129614073517e-05
    },
    {
      "step": 72.0,
      "loss": -0.00865542323501603,
      "accuracy": 0.625,
      "mean_reward": 0.7091477087220518,
      "entropy": 1.7916855596534436,
      "kl_from_reference": 7.39095686117944e-05
    },
    {
      "step": 84.0,
      "loss": -0.011630576959737682,
      "accuracy": 0.6484375,
      "mean_reward": 0.7256784150789694,
      "entropy": 1.7916763904863091,
      "kl_from_reference": 8.307873574572784e-05
    },
    {
      "step": 96.0,
      "loss": -0.009459048247931146,
      "accuracy": 0.5390625,
      "mean_reward": 0.6489398993526901,
      "entropy": 1.7916648274519154,
      "kl_from_reference": 9.464177013967946e-05
    },
    {
      "step": 108.0,
      "loss": -0.005442774516017934,
      "accuracy": 0.6328125,
      "mean_reward": 0.7158202719987394,
      "entropy": 1.7916767121831514,
      "kl_from_reference": 8.275703890376279e-05
    },
    {
      "step": 120.0,
      "loss": -0.011254298865137569,
      "accuracy": 0.6328125,
      "mean_reward": 0.7128031509162565,
      "entropy": 1.7916881669031022,
      "kl_from_reference": 7.130231895326579e-05
    }
  ],
  "schema_version": 2,
  "manifest_ref": "post-training:arithmetic-smoke-seed42",
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42
    ],
    "formal_comparison": false,
    "claim_policy": "single/few-seed smoke result; do not claim a stable improvement"
  },
  "provenance": {
    "artifact_path": "docs/post-training/2608.11674-gcpo/metrics/arithmetic-smoke-seed42.json",
    "historical_migration": "historical-metrics-v2-2026-08-09",
    "original_code_commit": "not recorded",
    "dataset_fingerprint": "not recorded in historical artifact"
  }
}