{
  "bench_version": 1,
  "track": "tabular",
  "model": "mlp_bottleneck",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.008,
      "eta": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.002,
          "eta": 0.0
        },
        "mean": 12.123852729797363
      },
      {
        "cfg": {
          "lr": 0.004,
          "eta": 0.0
        },
        "mean": 8.107524752616882
      },
      {
        "cfg": {
          "lr": 0.008,
          "eta": 0.0
        },
        "mean": 7.050076246261597
      }
    ],
    "full": {
      "mean": 7.273632645606995,
      "std": 0.5744234790485494,
      "per_seed": [
        6.92377233505249,
        7.258182048797607,
        6.877320289611816,
        7.141030311584473,
        6.306039810180664,
        7.486922264099121,
        8.12838363647461,
        8.067410469055176
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 7.273632645606995,
    "std": 0.5744234790485494,
    "per_seed": [
      6.92377233505249,
      7.258182048797607,
      6.877320289611816,
      7.141030311584473,
      6.306039810180664,
      7.486922264099121,
      8.12838363647461,
      8.067410469055176
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.0,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.0,
      0.0,
      0.0,
      0.0,
      0.0,
      0.0,
      0.0,
      0.0
    ],
    "p_value": 1.0,
    "mde": 0.0,
    "mde_rel_pct": 0.0,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "mechanism_signature": {
    "prediction": "larger eta produces larger trained GMM signature separation",
    "eta_rows": [
      {
        "cfg": {
          "lr": 0.008,
          "eta": 0.02
        },
        "observed_min_dist_mean": 2.4002824127674103,
        "observed_min_dist_per_seed": [
          2.228745698928833,
          2.3035049438476562,
          2.495332717895508,
          2.4624271392822266,
          2.388399839401245,
          2.2766027450561523,
          2.3105733394622803,
          2.736672878265381
        ]
      },
      {
        "cfg": {
          "lr": 0.008,
          "eta": 0.08
        },
        "observed_min_dist_mean": 2.4003629982471466,
        "observed_min_dist_per_seed": [
          2.2287771701812744,
          2.3035190105438232,
          2.4954094886779785,
          2.462498426437378,
          2.388439178466797,
          2.276618480682373,
          2.310865640640259,
          2.73677659034729
        ]
      },
      {
        "cfg": {
          "lr": 0.008,
          "eta": 0.2
        },
        "observed_min_dist_mean": 2.4003797471523285,
        "observed_min_dist_per_seed": [
          2.2287838459014893,
          2.3035221099853516,
          2.495425224304199,
          2.462512731552124,
          2.3884472846984863,
          2.2766220569610596,
          2.310927629470825,
          2.7367970943450928
        ]
      }
    ],
    "predicted_order": "eta .02 < .08 < .20",
    "confirmed": true
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.008,
        "eta": 0.02
      },
      "mean": 7.273632645606995,
      "std": 0.5744234790485494,
      "per_seed": [
        6.92377233505249,
        7.258182048797607,
        6.877320289611816,
        7.141030311584473,
        6.306039810180664,
        7.486922264099121,
        8.12838363647461,
        8.067410469055176
      ]
    },
    {
      "cfg": {
        "lr": 0.008,
        "eta": 0.08
      },
      "mean": 7.273632645606995,
      "std": 0.5744234790485494,
      "per_seed": [
        6.92377233505249,
        7.258182048797607,
        6.877320289611816,
        7.141030311584473,
        6.306039810180664,
        7.486922264099121,
        8.12838363647461,
        8.067410469055176
      ]
    },
    {
      "cfg": {
        "lr": 0.008,
        "eta": 0.2
      },
      "mean": 7.273632645606995,
      "std": 0.5744234790485494,
      "per_seed": [
        6.92377233505249,
        7.258182048797607,
        6.877320289611816,
        7.141030311584473,
        6.306039810180664,
        7.486922264099121,
        8.12838363647461,
        8.067410469055176
      ]
    }
  ],
  "protocol_note": "Tabular is the built-in regularization/optimizer track; both systems use the identical bottleneck MLP and differ only by the signature loss."
}