{
  "bench_version": 1,
  "track": "tabular",
  "model": "candidate_mlp_shared",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 119.02437210083008
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 30.62220811843872
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 21.151901245117188
      }
    ],
    "full": {
      "mean": 21.90489387512207,
      "std": 1.9880925147812747,
      "per_seed": [
        23.918453216552734,
        18.010822296142578,
        22.42278480529785,
        20.255544662475586,
        22.012264251708984,
        22.661293029785156,
        21.173133850097656,
        24.784854888916016
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 21.988378763198853,
    "std": 1.863640983270208,
    "per_seed": [
      23.71569061279297,
      18.270889282226562,
      22.492475509643555,
      20.57891082763672,
      22.177988052368164,
      22.6539249420166,
      21.250635147094727,
      24.766515731811523
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.08348488807678223,
    "idea_wins": 3,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.20276260375976562,
      0.2600669860839844,
      0.06969070434570312,
      0.3233661651611328,
      0.1657238006591797,
      -0.0073680877685546875,
      0.07750129699707031,
      -0.018339157104492188
    ],
    "p_value": 0.18505,
    "mde": 0.14001291226171048,
    "mde_rel_pct": 0.639185531141635,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "mechanism_signature": {
    "predicted_projected_gain_mean": 26.90758864581585,
    "observed_heldout_selected_response_mse_mean": 65.35227632522583,
    "predicted_per_seed": [
      21.2170729637146,
      7.129383563995361,
      23.922744750976562,
      49.054978370666504,
      8.374441027641296,
      25.982917308807373,
      46.1535587310791,
      33.425612449645996
    ],
    "observed_per_seed": [
      64.23245239257812,
      69.09458923339844,
      59.128604888916016,
      57.94867706298828,
      67.77635955810547,
      68.72184753417969,
      67.34278106689453,
      68.5728988647461
    ],
    "correlation": -0.415484140116091,
    "confirmed": false
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001
      },
      "result": {
        "mean": 117.0982551574707,
        "std": 6.973504377883443,
        "per_seed": [
          121.33436584472656,
          115.73588562011719,
          113.34111785888672,
          109.39837646484375,
          118.32662963867188,
          115.21646118164062,
          110.5782699584961,
          132.8549346923828
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.003
      },
      "result": {
        "mean": 31.2304470539093,
        "std": 3.2026968213759157,
        "per_seed": [
          33.56842803955078,
          26.45305633544922,
          31.497737884521484,
          29.350759506225586,
          30.496389389038086,
          29.599700927734375,
          30.8070011138916,
          38.07050323486328
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.01
      },
      "result": {
        "mean": 21.988378763198853,
        "std": 1.863640983270208,
        "per_seed": [
          23.71569061279297,
          18.270889282226562,
          22.492475509643555,
          20.57891082763672,
          22.177988052368164,
          22.6539249420166,
          21.250635147094727,
          24.766515731811523
        ],
        "n": 8
      }
    }
  ],
  "selection": {
    "candidates": 8,
    "active_blocks": 3,
    "epochs": 18,
    "n_train": 400,
    "n_test": 200,
    "track_rationale": "tabular regression is the built-in optimizer/regularizer/architecture-compatible small MLP track; no PDE/control structure is required by this adapter-bank idea"
  }
}