{
  "bench_version": 1,
  "track": "long_1d_convolution",
  "model": "cnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01,
      "m": 8
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "m": 8
        },
        "mean": 0.4312499910593033
      },
      {
        "cfg": {
          "lr": 0.003,
          "m": 8
        },
        "mean": 0.4387499913573265
      },
      {
        "cfg": {
          "lr": 0.01,
          "m": 8
        },
        "mean": 0.35874998569488525
      }
    ],
    "full": {
      "mean": 0.35249999165534973,
      "std": 0.02772634371136278,
      "per_seed": [
        0.41499999165534973,
        0.35999998450279236,
        0.32999998331069946,
        0.32999998331069946,
        0.3449999988079071,
        0.3700000047683716,
        0.3449999988079071,
        0.32499998807907104
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.48499998450279236,
    "std": 0.03381936953316897,
    "per_seed": [
      0.429999977350235,
      0.4399999976158142,
      0.5299999713897705,
      0.5149999856948853,
      0.4899999797344208,
      0.4699999988079071,
      0.5149999856948853,
      0.4899999797344208
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.13249999284744263,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.014999985694885254,
      0.08000001311302185,
      0.19999998807907104,
      0.1850000023841858,
      0.14499998092651367,
      0.09999999403953552,
      0.16999998688697815,
      0.16499999165534973
    ],
    "p_value": 0.0081,
    "mde": 0.05245633272675812,
    "mde_rel_pct": 14.881229494622602,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "mechanism_signature": {
    "mechanism_signature": {
      "prediction": "implicit convolution equals explicit convolution; tile storage ratio M/(ceil(L/m)*m)",
      "predicted_relative_operator_error": 0.0,
      "observed_relative_operator_error": 1.0825973749160767,
      "predicted_storage_ratio": 2.0,
      "trained_system_output_gap": 1.0547434091567993,
      "confirmed": false
    },
    "custom_track": {
      "name": "long_1d_convolution",
      "file": "stage2_bench.py",
      "domain": "cnn"
    }
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001,
        "m": 8
      },
      "mean": 0.48499999195337296
    },
    {
      "cfg": {
        "lr": 0.003,
        "m": 8
      },
      "mean": 0.48624999076128006
    },
    {
      "cfg": {
        "lr": 0.01,
        "m": 8
      },
      "mean": 0.47874998301267624
    }
  ],
  "selection_note": "Baseline and idea share lr union [0.001, 0.003, 0.01]; idea m fixed a priori at 8."
}