{
  "bench_version": 1,
  "track": "tabular",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01,
      "wd": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "wd": 0.0
        },
        "mean": 14.000884532928467
      },
      {
        "cfg": {
          "lr": 0.003,
          "wd": 0.0
        },
        "mean": 7.579635977745056
      },
      {
        "cfg": {
          "lr": 0.01,
          "wd": 0.0
        },
        "mean": 6.920138120651245
      }
    ],
    "full": {
      "mean": 6.9205562472343445,
      "std": 0.5716103719461259,
      "per_seed": [
        7.096155643463135,
        6.974558353424072,
        6.666646480560303,
        6.943192005157471,
        5.7007622718811035,
        7.199983596801758,
        6.8888444900512695,
        7.8943071365356445
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 6.810217976570129,
    "std": 0.34114415830274575,
    "per_seed": [
      6.926016807556152,
      6.900168418884277,
      6.7102837562561035,
      6.759965896606445,
      6.147101879119873,
      6.569408893585205,
      7.09395170211792,
      7.374846458435059
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.11033827066421509,
    "idea_wins": 5,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.17013883590698242,
      -0.07438993453979492,
      0.04363727569580078,
      -0.1832261085510254,
      0.44633960723876953,
      -0.6305747032165527,
      0.2051072120666504,
      -0.5194606781005859
    ],
    "p_value": 0.39255,
    "mde": 0.29668947868390905,
    "mde_rel_pct": 4.2870756061331745,
    "verdict": "no significant win",
    "system_worked": false
  },
  "mechanism_signature": {
    "signature_type": "trained_nn_optimizer_dynamics",
    "alpha": 1.5,
    "lambda": 1.0,
    "lr": 0.01,
    "q": 0.01,
    "predicted_drift_multiplier": 0.99,
    "observed_median_drift_multiplier": 0.9900000095367432,
    "observed_nn_jump_cf_power": 1.6607667261616632,
    "jump_mad": 0.0009690423030406237,
    "confirmed": true,
    "n_observations": 5914800,
    "training_metric_for_signature": {
      "mean": 6.810217976570129,
      "std": 0.34114415830274575,
      "per_seed": [
        6.926016807556152,
        6.900168418884277,
        6.7102837562561035,
        6.759965896606445,
        6.147101879119873,
        6.569408893585205,
        7.09395170211792,
        7.374846458435059
      ],
      "n": 8
    }
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.01,
        "wd": 0.0,
        "lambda": 1.0,
        "beta": 0.95,
        "noise": 0.001
      },
      "mean": 6.824108719825745
    },
    {
      "cfg": {
        "lr": 0.001,
        "wd": 0.0,
        "lambda": 1.0,
        "beta": 0.95,
        "noise": 0.001
      },
      "mean": 13.783732652664185
    },
    {
      "cfg": {
        "lr": 0.003,
        "wd": 0.0,
        "lambda": 1.0,
        "beta": 0.95,
        "noise": 0.001
      },
      "mean": 7.443580269813538
    }
  ],
  "protocol": {
    "paired_seeds": [
      0,
      1,
      2,
      3,
      4,
      5,
      6,
      7
    ],
    "epochs": 15,
    "n_train": 1200,
    "n_test": 400,
    "structural_match": "optimizer intervention on tabular regression"
  }
}