{
  "bench_version": 1,
  "track": "custom_relativistic_conservative_regression",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006,
      "epochs": 12
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 12
        },
        "mean": 0.95502670109272
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 12
        },
        "mean": 0.2731384299695492
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 12
        },
        "mean": 0.2562924362719059
      }
    ],
    "full": {
      "mean": 0.25546916387975216,
      "std": 0.020297905471986302,
      "per_seed": [
        0.26548606157302856,
        0.26569655537605286,
        0.222167506814003,
        0.2718196213245392,
        0.24948517978191376,
        0.23331035673618317,
        0.2888605296611786,
        0.24692749977111816
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.2924688458442688,
    "std": 0.11434509573619478,
    "per_seed": [
      0.24210484325885773,
      0.26006338000297546,
      0.22749106585979462,
      0.2767752707004547,
      0.5917121171951294,
      0.23255784809589386,
      0.2707672119140625,
      0.23827902972698212
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.03699968196451664,
    "idea_wins": 5,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.023381218314170837,
      -0.005633175373077393,
      0.005323559045791626,
      0.004955649375915527,
      0.34222693741321564,
      -0.0007525086402893066,
      -0.01809331774711609,
      -0.008648470044136047
    ],
    "p_value": 0.92165,
    "mde": 0.10347551087691399,
    "mde_rel_pct": 40.504109891563786,
    "verdict": "no significant win",
    "system_worked": false
  },
  "custom_track": {
    "name": "relativistic_conservative_regression",
    "file": "gql_track.py",
    "domain": "conservative_state_admissibility"
  },
  "sweep_parity": {
    "union_grid": [
      {
        "lr": 0.001,
        "epochs": 12
      },
      {
        "lr": 0.003,
        "epochs": 12
      },
      {
        "lr": 0.006,
        "epochs": 12
      }
    ],
    "idea_sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 12
        },
        "mean": 0.7376375496387482
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 12
        },
        "mean": 0.2556793764233589
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 12
        },
        "mean": 0.25160863995552063
      }
    ]
  },
  "mechanism_signature": {
    "claim": "limiting a valid baseline candidate along the complete residual enforces admissibility",
    "predicted_invalid_rate_after_limit": 0.0,
    "observed_invalid_rate_after_limit": 0.019999999552965164,
    "observed_raw_invalid_rate": 0.1299999998882413,
    "trained_model_measurements": [
      {
        "seed": 0,
        "invalid_raw": 0.029999999329447746,
        "invalid_limited": 0.0,
        "mean_theta": 0.9987504482269287
      },
      {
        "seed": 1,
        "invalid_raw": 0.009999999776482582,
        "invalid_limited": 0.0,
        "mean_theta": 0.9990912079811096
      },
      {
        "seed": 2,
        "invalid_raw": 0.0,
        "invalid_limited": 0.0,
        "mean_theta": 1.0
      },
      {
        "seed": 3,
        "invalid_raw": 0.0,
        "invalid_limited": 0.0,
        "mean_theta": 1.0
      },
      {
        "seed": 4,
        "invalid_raw": 1.0,
        "invalid_limited": 0.1599999964237213,
        "mean_theta": 0.5550016164779663
      },
      {
        "seed": 5,
        "invalid_raw": 0.0,
        "invalid_limited": 0.0,
        "mean_theta": 1.0
      },
      {
        "seed": 6,
        "invalid_raw": 0.0,
        "invalid_limited": 0.0,
        "mean_theta": 1.0
      },
      {
        "seed": 7,
        "invalid_raw": 0.0,
        "invalid_limited": 0.0,
        "mean_theta": 1.0
      }
    ],
    "confirmed": false
  }
}