{
  "bench_version": 1,
  "track": "tabular",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.003,
      "epochs": 8,
      "weight_decay": 0.0,
      "lam": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 8,
          "weight_decay": 0.0,
          "lam": 0.0
        },
        "mean": 188.35083389282227
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 8,
          "weight_decay": 0.0,
          "lam": 0.0
        },
        "mean": 19.947585344314575
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 8,
          "weight_decay": 0.0,
          "lam": 0.0
        },
        "mean": 22.523260593414307
      }
    ],
    "full": {
      "mean": 21.810271620750427,
      "std": 6.031322024802775,
      "per_seed": [
        17.455322265625,
        15.50521183013916,
        29.37735366821289,
        17.45245361328125,
        33.27134323120117,
        16.96589469909668,
        22.887645721435547,
        21.56694793701172
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 21.810271620750427,
    "std": 6.031322024802775,
    "per_seed": [
      17.455322265625,
      15.50521183013916,
      29.37735366821289,
      17.45245361328125,
      33.27134323120117,
      16.96589469909668,
      22.887645721435547,
      21.56694793701172
    ],
    "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
  },
  "selected_idea_cfg": {
    "lr": 0.003,
    "epochs": 8,
    "weight_decay": 0.0,
    "lam": 0.0001
  },
  "math_check": {
    "tau": 0.5,
    "target": [
      0.6065306663513184,
      0.5518191456794739
    ],
    "empirical": [
      0.600790664279383,
      0.5476969697311627
    ],
    "abs_error": [
      0.005740002071935368,
      0.004122175948311213
    ],
    "confirmed": true
  },
  "mechanism_signature": {
    "source": "trained mlp_tiny Jacobian on held-out tabular inputs",
    "predicted_moments": [
      0.9692332148551941,
      0.9687697291374207
    ],
    "observed_moments": [
      5.242537975311279,
      34.17074203491211
    ],
    "log_moment_rmse": 2.7879295349121094,
    "confirmed": false
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.0001
      },
      "result": {
        "mean": 187.1381435394287,
        "std": 11.829779321921778,
        "per_seed": [
          200.69190979003906,
          191.61203002929688,
          186.99557495117188,
          174.10382080078125,
          177.56117248535156,
          178.70860290527344,
          177.95843505859375,
          209.47360229492188
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.001,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.0003
      },
      "result": {
        "mean": 187.1381435394287,
        "std": 11.829779321921778,
        "per_seed": [
          200.69190979003906,
          191.61203002929688,
          186.99557495117188,
          174.10382080078125,
          177.56117248535156,
          178.70860290527344,
          177.95843505859375,
          209.47360229492188
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.001,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.001
      },
      "result": {
        "mean": 187.1381435394287,
        "std": 11.829779321921778,
        "per_seed": [
          200.69190979003906,
          191.61203002929688,
          186.99557495117188,
          174.10382080078125,
          177.56117248535156,
          178.70860290527344,
          177.95843505859375,
          209.47360229492188
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.003,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.0001
      },
      "result": {
        "mean": 21.810271620750427,
        "std": 6.031322024802775,
        "per_seed": [
          17.455322265625,
          15.50521183013916,
          29.37735366821289,
          17.45245361328125,
          33.27134323120117,
          16.96589469909668,
          22.887645721435547,
          21.56694793701172
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.003,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.0003
      },
      "result": {
        "mean": 21.810271620750427,
        "std": 6.031322024802775,
        "per_seed": [
          17.455322265625,
          15.50521183013916,
          29.37735366821289,
          17.45245361328125,
          33.27134323120117,
          16.96589469909668,
          22.887645721435547,
          21.56694793701172
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.003,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.001
      },
      "result": {
        "mean": 21.810271620750427,
        "std": 6.031322024802775,
        "per_seed": [
          17.455322265625,
          15.50521183013916,
          29.37735366821289,
          17.45245361328125,
          33.27134323120117,
          16.96589469909668,
          22.887645721435547,
          21.56694793701172
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.006,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.0001
      },
      "result": {
        "mean": 22.986173629760742,
        "std": 2.470365935456011,
        "per_seed": [
          25.04693603515625,
          23.222103118896484,
          21.88385772705078,
          19.94014549255371,
          19.829004287719727,
          21.487890243530273,
          25.67311668395996,
          26.80633544921875
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.006,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.0003
      },
      "result": {
        "mean": 22.986173629760742,
        "std": 2.470365935456011,
        "per_seed": [
          25.04693603515625,
          23.222103118896484,
          21.88385772705078,
          19.94014549255371,
          19.829004287719727,
          21.487890243530273,
          25.67311668395996,
          26.80633544921875
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.006,
        "epochs": 8,
        "weight_decay": 0.0,
        "lam": 0.001
      },
      "result": {
        "mean": 22.986173629760742,
        "std": 2.470365935456011,
        "per_seed": [
          25.04693603515625,
          23.222103118896484,
          21.88385772705078,
          19.94014549255371,
          19.829004287719727,
          21.487890243530273,
          25.67311668395996,
          26.80633544921875
        ],
        "n": 8
      }
    }
  ]
}