{
  "bench_version": 1,
  "track": "sequence",
  "model": "transformer_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.001,
      "epochs": 12
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 12
        },
        "mean": 0.16501786187291145
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 12
        },
        "mean": 0.1727909930050373
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 12
        },
        "mean": 0.17529388144612312
      }
    ],
    "full": {
      "mean": 0.16652636602520943,
      "std": 0.019344256210385904,
      "per_seed": [
        0.18950511515140533,
        0.15041682124137878,
        0.14130990207195282,
        0.17883960902690887,
        0.15188206732273102,
        0.1598561406135559,
        0.20002053678035736,
        0.16038073599338531
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.23773143999278545,
    "std": 0.022669798148183287,
    "per_seed": [
      0.24208609759807587,
      0.21345070004463196,
      0.24233895540237427,
      0.20291335880756378,
      0.27286380529403687,
      0.23195308446884155,
      0.26794296503067017,
      0.22830255329608917
    ],
    "n": 8,
    "best_cfg": {
      "lr": 0.001,
      "epochs": 12
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 12
        },
        "mean": 0.23773143999278545,
        "per_seed": [
          0.24208609759807587,
          0.21345070004463196,
          0.24233895540237427,
          0.20291335880756378,
          0.27286380529403687,
          0.23195308446884155,
          0.26794296503067017,
          0.22830255329608917
        ]
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 12
        },
        "mean": 0.25968209840357304,
        "per_seed": [
          0.28427907824516296,
          0.2242421656847,
          0.24691039323806763,
          0.1911633461713791,
          0.319369912147522,
          0.28412121534347534,
          0.29372480511665344,
          0.23364587128162384
        ]
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 12
        },
        "mean": 0.3195491824299097,
        "per_seed": [
          0.33699294924736023,
          0.29003939032554626,
          0.30690643191337585,
          0.20158611238002777,
          0.3861159682273865,
          0.33679136633872986,
          0.40094634890556335,
          0.29701489210128784
        ]
      }
    ]
  },
  "comparison": {
    "delta_mean": 0.07120507396757603,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.05258098244667053,
      0.06303387880325317,
      0.10102905333042145,
      0.024073749780654907,
      0.12098173797130585,
      0.07209694385528564,
      0.0679224282503128,
      0.06792181730270386
    ],
    "p_value": 0.0081,
    "mde": 0.02453005366966053,
    "mde_rel_pct": 14.730432336430782,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "selection": {
    "baseline_best_cfg": {
      "lr": 0.001,
      "epochs": 12
    },
    "idea_best_cfg": {
      "lr": 0.001,
      "epochs": 12
    },
    "shared_lr_union": [
      0.001,
      0.003,
      0.006
    ]
  },
  "structural_match": "multi-token correlated sequence window",
  "mechanism_signature": {
    "quantity": "trained-model pair inclusion mass",
    "predicted_direction": "compatibility boost increases joint inclusion",
    "boost": 2.0,
    "observed_before": 0.039688840508461,
    "observed_after": 0.22817423939704895,
    "observed_change": 0.18848538398742676,
    "confirmed": true
  },
  "protocol_note": "Eight paired seeds; three shared learning rates; reduced 1200/400 samples and 12 epochs for runtime."
}