{
  "schema_version": 3,
  "config": {
    "objective": "orpo",
    "dataset": "ultrafeedback",
    "model_id": "HuggingFaceTB/SmolLM2-135M-Instruct",
    "model_revision": "12fd25f77366fa6b3b4b768ec3050bf629380bac",
    "dataset_revision": "292c16329d921287c4166934cac1a6ad1e13a6c5",
    "steps": 40,
    "maximum_examples": 64,
    "evaluation_examples": 24,
    "seeds": [42, 43, 44]
  },
  "seed_results": [
    {"seed": 42, "baseline": {"preference_accuracy": 0.3333333333, "preference_margin": -49.0416667}, "final": {"preference_accuracy": 0.3333333333, "preference_margin": -48.75}},
    {"seed": 43, "baseline": {"preference_accuracy": 0.3333333333, "preference_margin": -49.0416667}, "final": {"preference_accuracy": 0.3333333333, "preference_margin": -48.2916667}},
    {"seed": 44, "baseline": {"preference_accuracy": 0.3333333333, "preference_margin": -49.0416667}, "final": {"preference_accuracy": 0.3333333333, "preference_margin": -48.125}}
  ],
  "metrics": {
    "metric": "preference_accuracy",
    "baseline": {"mean": 0.3333333333, "std": 0.0, "ci95_low": 0.3333333333, "ci95_high": 0.3333333333},
    "final": {"mean": 0.3333333333, "std": 0.0, "ci95_low": 0.3333333333, "ci95_high": 0.3333333333},
    "baseline_preference_margin": {"mean": -49.0416667, "std": 0.0, "ci95_low": -49.0416667, "ci95_high": -49.0416667},
    "final_preference_margin": {"mean": -48.3888889, "std": 0.3236439, "ci95_low": -48.7551264, "ci95_high": -48.0226513}
  },
  "evaluation_protocol": {"three_independent_training_seeds": true, "fixed_public_test_split": true, "test_used_for_model_selection": false}
}