{
  "algorithm": "grin",
  "dataset": "arithmetic-smoke",
  "baseline": {
    "accuracy": 0.125,
    "mean_reward": 0.3039061471113459,
    "entropy": 1.7917594692220566,
    "kl_from_reference": 0.0
  },
  "final": {
    "accuracy": 0.34375,
    "mean_reward": 0.5015669535289922,
    "entropy": 1.78177507692576,
    "kl_from_reference": 0.009984392296294699
  },
  "relative_accuracy": 1.75,
  "training": {
    "steps": 100,
    "learning_rate": 0.08,
    "group_size": 4,
    "train_examples": 64,
    "validation_examples": 64,
    "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": 6,
    "last_diagnostics": {
      "failed_rollout_fraction": 0.75,
      "golden_answer_injection_fraction": 0.75,
      "mixed_policy_importance_mean": 1.0011693423059347,
      "sft_loss_weight": 0.0,
      "loss": -0.05841395790548426,
      "policy_entropy": 1.7793068730374229
    },
    "fidelity": "mechanism reproduction on a candidate-policy model"
  },
  "history": [
    {
      "step": 1.0,
      "loss": -0.0,
      "accuracy": 0.125,
      "mean_reward": 0.3039061471113459,
      "entropy": 1.7917594692220566,
      "kl_from_reference": 0.0
    },
    {
      "step": 10.0,
      "loss": -0.0,
      "accuracy": 0.203125,
      "mean_reward": 0.38792858424974397,
      "entropy": 1.7914302101726771,
      "kl_from_reference": 0.0003292590493776422
    },
    {
      "step": 20.0,
      "loss": -0.016233461812480143,
      "accuracy": 0.203125,
      "mean_reward": 0.38792858424974397,
      "entropy": 1.7905569573613525,
      "kl_from_reference": 0.0012025118607025696
    },
    {
      "step": 30.0,
      "loss": -0.0006860813506200242,
      "accuracy": 0.265625,
      "mean_reward": 0.43943013355655564,
      "entropy": 1.7905217346026638,
      "kl_from_reference": 0.0012377346193916112
    },
    {
      "step": 40.0,
      "loss": -0.004161497483530466,
      "accuracy": 0.28125,
      "mean_reward": 0.45181326126453875,
      "entropy": 1.7897993624995125,
      "kl_from_reference": 0.0019601067225423242
    },
    {
      "step": 50.0,
      "loss": -0.0066235720112600704,
      "accuracy": 0.34375,
      "mean_reward": 0.5015669535289922,
      "entropy": 1.7891730177851524,
      "kl_from_reference": 0.0025864514369030862
    },
    {
      "step": 60.0,
      "loss": -0.0059264631412970364,
      "accuracy": 0.359375,
      "mean_reward": 0.5130744317531799,
      "entropy": 1.78871092531289,
      "kl_from_reference": 0.003048543909165282
    },
    {
      "step": 70.0,
      "loss": -0.03346042882470243,
      "accuracy": 0.390625,
      "mean_reward": 0.5356562676860482,
      "entropy": 1.7877490280542963,
      "kl_from_reference": 0.0040104411677585215
    },
    {
      "step": 80.0,
      "loss": -0.04072930687758103,
      "accuracy": 0.28125,
      "mean_reward": 0.452768114328201,
      "entropy": 1.7853659312151076,
      "kl_from_reference": 0.006393538006947649
    },
    {
      "step": 90.0,
      "loss": -0.03222791412686621,
      "accuracy": 0.390625,
      "mean_reward": 0.5361244882942591,
      "entropy": 1.7846492364129867,
      "kl_from_reference": 0.007110232809067988
    },
    {
      "step": 100.0,
      "loss": -0.05841395790548426,
      "accuracy": 0.34375,
      "mean_reward": 0.5015669535289922,
      "entropy": 1.78177507692576,
      "kl_from_reference": 0.009984392296294699
    }
  ],
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42
    ],
    "formal_comparison": false,
    "claim_policy": "single/few-seed mechanism result; do not claim a stable improvement"
  },
  "schema_version": 2,
  "manifest_ref": "post-training:grin-arithmetic-smoke-seed42",
  "provenance": {
    "artifact_path": "docs/post-training/2608.25243-grin/metrics/grin-arithmetic-smoke-seed42.json",
    "dataset_fingerprint": "not recorded in historical artifact",
    "historical_migration": "historical-metrics-v2-2026-08-09",
    "original_code_commit": "not recorded"
  }
}
