{
  "bench_version": 1,
  "track": "tabular",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 17.09206533432007
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 11.12595009803772
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 6.983492374420166
      }
    ],
    "full": {
      "mean": 7.1135499477386475,
      "std": 0.44626176436687165,
      "per_seed": [
        6.864531517028809,
        7.168848514556885,
        7.027814865112305,
        6.872774600982666,
        6.30631685256958,
        7.218708515167236,
        7.5617499351501465,
        7.887654781341553
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 7.093362271785736,
    "std": 0.45241222062673386,
    "per_seed": [
      6.777974605560303,
      7.191386699676514,
      7.057932376861572,
      6.801158428192139,
      6.306606292724609,
      7.222118854522705,
      7.4917168617248535,
      7.898004055023193
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.020187675952911377,
    "idea_wins": 3,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.08655691146850586,
      0.022538185119628906,
      0.030117511749267578,
      -0.07161617279052734,
      0.0002894401550292969,
      0.00341033935546875,
      -0.07003307342529297,
      0.010349273681640625
    ],
    "p_value": 0.258,
    "mde": 0.03972406104311891,
    "mde_rel_pct": 0.5584280891391918,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "mechanism_signature": {
    "predicted_mass_error": 0.0,
    "observed_mass_error": 4.440892098500626e-16,
    "predicted_conservative": true,
    "observed_mean_activity_variance": 1346.4940232817462,
    "density_mean": 1.0,
    "density_std": 0.6109128074216446,
    "confirmed": true
  },
  "idea_sweep": {
    "0.001": {
      "mean": 17.594895243644714,
      "std": 1.455041940026395,
      "per_seed": [
        15.97562313079834,
        17.117998123168945,
        18.321380615234375,
        16.85783576965332,
        15.77338981628418,
        17.865934371948242,
        18.211532592773438,
        20.635467529296875
      ],
      "n": 8
    },
    "0.003": {
      "mean": 10.802422881126404,
      "std": 0.9665300052415943,
      "per_seed": [
        9.816262245178223,
        11.428475379943848,
        10.502289772033691,
        10.368227005004883,
        9.419925689697266,
        10.578962326049805,
        11.818642616271973,
        12.486598014831543
      ],
      "n": 8
    },
    "0.01": {
      "mean": 7.093362271785736,
      "std": 0.45241222062673386,
      "per_seed": [
        6.777974605560303,
        7.191386699676514,
        7.057932376861572,
        6.801158428192139,
        6.306606292724609,
        7.222118854522705,
        7.4917168617248535,
        7.898004055023193
      ],
      "n": 8
    }
  },
  "protocol": {
    "paired_seeds": [
      0,
      1,
      2,
      3,
      4,
      5,
      6,
      7
    ],
    "shared_lr_union": [
      0.001,
      0.003,
      0.01
    ],
    "track_rationale": "tabular is the built-in structural match for optimizer interventions"
  }
}