{
  "algorithm": "smopd",
  "dataset": "arithmetic-smoke",
  "baseline": {
    "accuracy": 0.1953125,
    "mean_reward": 0.359689058826365,
    "entropy": 1.791759469222057,
    "kl_from_reference": 0.0
  },
  "final": {
    "accuracy": 0.3125,
    "mean_reward": 0.4806971630154059,
    "entropy": 1.436163275620846,
    "kl_from_reference": 0.3555961936012087
  },
  "relative_accuracy": 0.6,
  "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": 512,
    "teacher_prefill_calls": 512,
    "online_teacher_calls": 0,
    "drift_events": 0,
    "critic_updates": 0,
    "rollout_policy_refreshes": 7,
    "last_diagnostics": {
      "mechanism_id": 11.0,
      "specialized_teachers": 4.0,
      "advantage_std": 1.1001581203655664,
      "loss": -0.739157586726989,
      "policy_entropy": 1.5188752013228874
    },
    "fidelity": "mechanism reproduction on a candidate-policy model"
  },
  "history": [
    {
      "step": 1.0,
      "loss": 2.220446049250313e-16,
      "accuracy": 0.6328125,
      "mean_reward": 0.7142799734654669,
      "entropy": 1.791704461786654,
      "kl_from_reference": 5.500743540123352e-05
    },
    {
      "step": 12.0,
      "loss": -0.04784054252429021,
      "accuracy": 0.53125,
      "mean_reward": 0.6447878429840455,
      "entropy": 1.7876528990874323,
      "kl_from_reference": 0.004106570134622215
    },
    {
      "step": 24.0,
      "loss": -0.2509664725184687,
      "accuracy": 0.359375,
      "mean_reward": 0.517814626907501,
      "entropy": 1.7699471790553598,
      "kl_from_reference": 0.02181229016669475
    },
    {
      "step": 36.0,
      "loss": -0.22293814884606933,
      "accuracy": 0.3671875,
      "mean_reward": 0.5240883563980406,
      "entropy": 1.7475282267507184,
      "kl_from_reference": 0.04423124247133692
    },
    {
      "step": 48.0,
      "loss": -0.024518706107619415,
      "accuracy": 0.296875,
      "mean_reward": 0.468677279507009,
      "entropy": 1.6960389764308512,
      "kl_from_reference": 0.09572049279120357
    },
    {
      "step": 60.0,
      "loss": -0.0942908237987018,
      "accuracy": 0.28125,
      "mean_reward": 0.45564911291653953,
      "entropy": 1.615877161014569,
      "kl_from_reference": 0.175882308207485
    },
    {
      "step": 72.0,
      "loss": -0.018256824629011348,
      "accuracy": 0.28125,
      "mean_reward": 0.45564911291653953,
      "entropy": 1.5795641054830223,
      "kl_from_reference": 0.21219536373903197
    },
    {
      "step": 84.0,
      "loss": -0.11847395188953136,
      "accuracy": 0.3046875,
      "mean_reward": 0.47475757647766553,
      "entropy": 1.5811746321263176,
      "kl_from_reference": 0.21058483709573633
    },
    {
      "step": 96.0,
      "loss": 0.033956603383559936,
      "accuracy": 0.3203125,
      "mean_reward": 0.4866164071619053,
      "entropy": 1.5531212029510608,
      "kl_from_reference": 0.23863826627099438
    },
    {
      "step": 108.0,
      "loss": 0.009089696176576262,
      "accuracy": 0.3046875,
      "mean_reward": 0.47475757647766553,
      "entropy": 1.481437567159204,
      "kl_from_reference": 0.3103219020628508
    },
    {
      "step": 120.0,
      "loss": -0.739157586726989,
      "accuracy": 0.3125,
      "mean_reward": 0.4806971630154059,
      "entropy": 1.436163275620846,
      "kl_from_reference": 0.3555961936012087
    }
  ],
  "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/2608.03092-smopd/metrics/arithmetic-smoke-seed42.json",
    "dataset_fingerprint": "not recorded in historical artifact",
    "historical_migration": "historical-metrics-v2-2026-08-09",
    "original_code_commit": "not recorded"
  }
}