{
  "protocol": {
    "post_training_dataset": "OpenAI GSM8K official train/test JSONL",
    "post_training_train_validation": [
      128,
      32
    ],
    "post_training_steps": 120,
    "agent_episodes": 120,
    "seed": 42,
    "fidelity": "mechanism reproduction; not original full-size LLM training"
  },
  "results": {
    "gkd": {
      "baseline_accuracy": 0.25,
      "accuracy": 0.9375,
      "mean_reward": 0.9288500410880138,
      "kl_from_reference": 0.5122175188556003,
      "student_rollouts": 480,
      "online_teacher_calls": 480
    },
    "minillm": {
      "baseline_accuracy": 0.25,
      "accuracy": 0.875,
      "mean_reward": 0.8817471475152978,
      "kl_from_reference": 0.5476656946243136,
      "teacher_mix": 0.2,
      "online_teacher_calls": 480
    },
    "search-r1": {
      "benchmark": "scalemcp-mini",
      "joint_success": 1.0,
      "average_cost": 2.75,
      "search_queries": 240,
      "retrieved_tokens_masked": 1800,
      "outcome_rewards": 120
    },
    "ragen": {
      "benchmark": "planbench-mini",
      "joint_success": 1.0,
      "average_cost": 1.6,
      "trajectory_rollouts": 480,
      "trajectory_filters": 120,
      "critic_baseline_updates": 120,
      "echo_trap_events": 19,
      "gradient_clips": 19
    }
  },
  "schema_version": 2,
  "manifest_ref": "experiments:classic-agentic-rl-opd-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/experiments/classic-agentic-rl-opd-seed42.json",
    "historical_migration": "historical-metrics-v2-2026-08-09",
    "original_code_commit": "not recorded",
    "dataset_fingerprint": "not recorded in historical artifact"
  }
}