{
  "bench_version": 1,
  "track": "tabular",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01,
      "weight_decay": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "weight_decay": 0.0
        },
        "mean": 52.02932357788086
      },
      {
        "cfg": {
          "lr": 0.003,
          "weight_decay": 0.0
        },
        "mean": 14.625996589660645
      },
      {
        "cfg": {
          "lr": 0.01,
          "weight_decay": 0.0
        },
        "mean": 9.001914024353027
      }
    ],
    "full": {
      "mean": 9.259196519851685,
      "std": 1.0373002651087238,
      "per_seed": [
        8.302460670471191,
        10.292799949645996,
        8.400059700012207,
        9.012335777282715,
        7.509613037109375,
        9.842769622802734,
        10.44126033782959,
        10.272273063659668
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 15.81218934059143,
    "std": 1.949355068505579,
    "per_seed": [
      16.183080673217773,
      15.5823335647583,
      17.167373657226562,
      13.497793197631836,
      12.077912330627441,
      16.422466278076172,
      17.101774215698242,
      18.464780807495117
    ],
    "n": 8,
    "best_cfg": {
      "lr": 0.003,
      "weight_decay": 0.0
    }
  },
  "comparison": {
    "delta_mean": 6.552992820739746,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      7.880620002746582,
      5.289533615112305,
      8.767313957214355,
      4.485457420349121,
      4.568299293518066,
      6.5796966552734375,
      6.660513877868652,
      8.19250774383545
    ],
    "p_value": 0.0081,
    "mde": 1.3839413130321554,
    "mde_rel_pct": 14.94666745721392,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "track_choice": "tabular is the built-in optimizer/training-dynamics match; Friedman regression keeps architecture and task fixed.",
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001,
        "weight_decay": 0.0
      },
      "result": {
        "mean": 76.82785987854004,
        "std": 25.70975037882407,
        "per_seed": [
          70.13491821289062,
          102.93740844726562,
          84.74547576904297,
          91.09562683105469,
          33.59514236450195,
          98.26583099365234,
          37.17234420776367,
          96.67613220214844
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.003,
        "weight_decay": 0.0
      },
      "result": {
        "mean": 15.81218934059143,
        "std": 1.949355068505579,
        "per_seed": [
          16.183080673217773,
          15.5823335647583,
          17.167373657226562,
          13.497793197631836,
          12.077912330627441,
          16.422466278076172,
          17.101774215698242,
          18.464780807495117
        ],
        "n": 8,
        "best_cfg": {
          "lr": 0.003,
          "weight_decay": 0.0
        }
      }
    },
    {
      "cfg": {
        "lr": 0.01,
        "weight_decay": 0.0
      },
      "result": {
        "mean": 34.88474154472351,
        "std": 15.230981751784949,
        "per_seed": [
          34.66309356689453,
          59.120567321777344,
          43.088172912597656,
          53.1762580871582,
          21.203292846679688,
          34.56721496582031,
          14.101446151733398,
          19.157886505126953
        ],
        "n": 8
      }
    }
  ],
  "budget": {
    "baseline_epochs": 16,
    "idea_generations": 16,
    "population": 8
  },
  "mechanism_signature": {
    "observed_cov_increment": 0.000783836052484821,
    "predicted_2D_increment": 0.0005999999999999998,
    "relative_error": 0.3063934202973796,
    "curvature_median_final": -8.0,
    "confirmed": true,
    "trained_model_metric": 16.183080673217773
  }
}