{
  "bench_version": 1,
  "track": "tabular",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006,
      "weight_decay": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "weight_decay": 0.0
        },
        "mean": 15.53538990020752
      },
      {
        "cfg": {
          "lr": 0.003,
          "weight_decay": 0.0
        },
        "mean": 9.415347337722778
      },
      {
        "cfg": {
          "lr": 0.006,
          "weight_decay": 0.0
        },
        "mean": 7.360140562057495
      }
    ],
    "full": {
      "mean": 7.4583141803741455,
      "std": 0.6102747432817298,
      "per_seed": [
        7.289438247680664,
        8.00219440460205,
        7.081491947174072,
        7.067437648773193,
        6.286782264709473,
        7.646380424499512,
        8.093415260314941,
        8.199373245239258
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 7.388119101524353,
    "std": 0.6041003222700619,
    "per_seed": [
      6.846275329589844,
      7.8614349365234375,
      7.108870506286621,
      7.15531587600708,
      6.309638023376465,
      7.587562084197998,
      8.090027809143066,
      8.145828247070312
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.07019507884979248,
    "idea_wins": 5,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.4431629180908203,
      -0.14075946807861328,
      0.027378559112548828,
      0.08787822723388672,
      0.022855758666992188,
      -0.05881834030151367,
      -0.003387451171875,
      -0.05354499816894531
    ],
    "p_value": 0.2587,
    "mde": 0.13852366888165354,
    "mde_rel_pct": 1.8573053579073617,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "idea_trials": [
    {
      "cfg": {
        "lr": 0.001,
        "weight_decay": 0.0
      },
      "result": {
        "mean": 16.191248655319214,
        "std": 1.334610265709453,
        "per_seed": [
          15.566378593444824,
          16.270675659179688,
          17.070934295654297,
          14.182695388793945,
          14.616189002990723,
          15.975017547607422,
          17.418180465698242,
          18.42991828918457
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.003,
        "weight_decay": 0.0
      },
      "result": {
        "mean": 9.826780676841736,
        "std": 1.0954169590780658,
        "per_seed": [
          8.976614952087402,
          11.023777961730957,
          9.17143726348877,
          9.166444778442383,
          8.010053634643555,
          10.084756851196289,
          10.998445510864258,
          11.182714462280273
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.006,
        "weight_decay": 0.0
      },
      "result": {
        "mean": 7.388119101524353,
        "std": 0.6041003222700619,
        "per_seed": [
          6.846275329589844,
          7.8614349365234375,
          7.108870506286621,
          7.15531587600708,
          6.309638023376465,
          7.587562084197998,
          8.090027809143066,
          8.145828247070312
        ],
        "n": 8
      }
    }
  ],
  "task_match": "optimizer intervention on Friedman regression",
  "mechanism_signature": {
    "raw_grad_lag1_corr": 0.3716290132687427,
    "corrected_grad_lag1_corr": 0.12645231312280522,
    "baseline_raw_grad_lag1_corr": 0.5400456543155014,
    "confirmed": true
  }
}