{
  "bench_version": 1,
  "track": "tabular",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006,
      "weight_decay": 0.0001,
      "epochs": 12
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.0015,
          "weight_decay": 0.0,
          "epochs": 12
        },
        "mean": 13.675705671310425
      },
      {
        "cfg": {
          "lr": 0.003,
          "weight_decay": 0.0,
          "epochs": 12
        },
        "mean": 9.352656126022339
      },
      {
        "cfg": {
          "lr": 0.006,
          "weight_decay": 0.0,
          "epochs": 12
        },
        "mean": 7.226524114608765
      },
      {
        "cfg": {
          "lr": 0.0015,
          "weight_decay": 0.0001,
          "epochs": 12
        },
        "mean": 13.609990000724792
      },
      {
        "cfg": {
          "lr": 0.003,
          "weight_decay": 0.0001,
          "epochs": 12
        },
        "mean": 9.076743841171265
      },
      {
        "cfg": {
          "lr": 0.006,
          "weight_decay": 0.0001,
          "epochs": 12
        },
        "mean": 7.0532397627830505
      }
    ],
    "full": {
      "mean": 7.0532397627830505,
      "std": 0.2264412705403704,
      "per_seed": [
        6.572167873382568,
        7.1756792068481445,
        7.042057037353516,
        7.032951831817627,
        7.018033504486084,
        7.022153854370117,
        7.112776756286621,
        7.450098037719727
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 7.014131605625153,
    "std": 0.24308566746895208,
    "per_seed": [
      6.6641693115234375,
      7.260889053344727,
      6.737545013427734,
      6.856451034545898,
      6.927297592163086,
      7.039077281951904,
      7.303320407867432,
      7.324303150177002
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.03910815715789795,
    "idea_wins": 4,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.09200143814086914,
      0.08520984649658203,
      -0.30451202392578125,
      -0.17650079727172852,
      -0.09073591232299805,
      0.01692342758178711,
      0.19054365158081055,
      -0.1257948875427246
    ],
    "p_value": 0.50145,
    "mde": 0.13703746835599878,
    "mde_rel_pct": 1.9429010350546605,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "selection_details": "Entropy and validation-risk masks selected from train split; final systems use identical mlp_tiny/training.",
  "mechanism_signature": {
    "prediction": "kappa>=0.90 routes away from entropy mask and avoids materially worse trained-model MSE",
    "observed": [
      {
        "seed": 0,
        "kappa": 1.0,
        "gamma": 13.140283130679085,
        "used_entropy": false,
        "predicted_fallback": true,
        "observed_subset_minus_full_mse": 0.09200143814086914
      },
      {
        "seed": 1,
        "kappa": 1.0,
        "gamma": 16.30871552694337,
        "used_entropy": false,
        "predicted_fallback": true,
        "observed_subset_minus_full_mse": 0.08520984649658203
      },
      {
        "seed": 2,
        "kappa": 1.0,
        "gamma": 12.264759413539208,
        "used_entropy": false,
        "predicted_fallback": true,
        "observed_subset_minus_full_mse": -0.30451202392578125
      },
      {
        "seed": 3,
        "kappa": 1.0,
        "gamma": 13.648493168331017,
        "used_entropy": false,
        "predicted_fallback": true,
        "observed_subset_minus_full_mse": -0.17650079727172852
      },
      {
        "seed": 4,
        "kappa": 1.0,
        "gamma": 15.0961523290884,
        "used_entropy": false,
        "predicted_fallback": true,
        "observed_subset_minus_full_mse": -0.09073591232299805
      },
      {
        "seed": 5,
        "kappa": 1.0,
        "gamma": 11.799654276914342,
        "used_entropy": false,
        "predicted_fallback": true,
        "observed_subset_minus_full_mse": 0.01692342758178711
      },
      {
        "seed": 6,
        "kappa": 1.0,
        "gamma": 15.35830490498792,
        "used_entropy": false,
        "predicted_fallback": true,
        "observed_subset_minus_full_mse": 0.19054365158081055
      },
      {
        "seed": 7,
        "kappa": 1.0,
        "gamma": 13.753987138531764,
        "used_entropy": false,
        "predicted_fallback": true,
        "observed_subset_minus_full_mse": -0.1257948875427246
      }
    ],
    "high_kappa_n": 8,
    "high_kappa_observed_mean_delta": -0.03910815715789795,
    "confirmed": true
  }
}