{
  "bench_version": 1,
  "track": "sequence",
  "model": "transformer_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.001,
      "epochs": 8,
      "lam": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 8,
          "lam": 0.0
        },
        "mean": 0.19084082171320915
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 8,
          "lam": 0.0
        },
        "mean": 0.2167150229215622
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 8,
          "lam": 0.0
        },
        "mean": 0.26577289402484894
      }
    ],
    "full": {
      "mean": 0.19580462761223316,
      "std": 0.019276107610397688,
      "per_seed": [
        0.21272796392440796,
        0.16291894018650055,
        0.18857084214687347,
        0.19914554059505463,
        0.2006099969148636,
        0.1974794715642929,
        0.22919276356697083,
        0.17579150199890137
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.19599766097962856,
    "std": 0.018529453120078188,
    "per_seed": [
      0.21339234709739685,
      0.1631648987531662,
      0.19162587821483612,
      0.19664713740348816,
      0.1998937577009201,
      0.20089472830295563,
      0.22621849179267883,
      0.1761440485715866
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.000193033367395401,
    "idea_wins": 3,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.0006643831729888916,
      0.0002459585666656494,
      0.0030550360679626465,
      -0.0024984031915664673,
      -0.0007162392139434814,
      0.0034152567386627197,
      -0.002974271774291992,
      0.0003525465726852417
    ],
    "p_value": 0.7309,
    "mde": 0.0019204674874472793,
    "mde_rel_pct": 0.9808080181079926,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "mechanism_signature": {
    "prediction": "commutator regularization lowers trained attention commutator; multi-step compatibility follows from resolvent identity",
    "baseline_comm_rms": {
      "mean": 0.03365310910157859,
      "std": 0.005540028408003687,
      "per_seed": [
        0.028813593089580536,
        0.04024209827184677,
        0.037993812933564186,
        0.02756293211132288
      ],
      "n": 4
    },
    "idea_comm_rms": {
      "mean": 0.02128551318310201,
      "std": 0.0034537152021512133,
      "per_seed": [
        0.018550011795014143,
        0.024316378869116306,
        0.02507482608780265,
        0.01720083598047495
      ],
      "n": 4
    },
    "baseline_function_reversal_mse": {
      "mean": 1.6439854502677917,
      "std": 0.2084291753940242,
      "per_seed": [
        1.8240505456924438,
        1.5404354333877563,
        1.3530447483062744,
        1.8584110736846924
      ],
      "n": 4
    },
    "idea_function_reversal_mse": {
      "mean": 1.6318478286266327,
      "std": 0.20953659306825898,
      "per_seed": [
        1.841636061668396,
        1.5250027179718018,
        1.340692400932312,
        1.820060133934021
      ],
      "n": 4
    },
    "observed_reduction_factor": 1.5810334856429433,
    "confirmed": true,
    "math_check": {
      "resolvent_identity_rel_error": 4.1222850040650605e-15,
      "gamma_1_commutator": 0.0,
      "identity_confirmed": true
    },
    "idea_pilot_sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 8,
          "lam": 0.3
        },
        "result": {
          "mean": 0.19120756536722183,
          "std": 0.018085350569029532,
          "per_seed": [
            0.21339234709739685,
            0.1631648987531662,
            0.19162587821483612,
            0.19664713740348816
          ],
          "n": 4
        }
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 8,
          "lam": 0.3
        },
        "result": {
          "mean": 0.21730473637580872,
          "std": 0.029589204134821655,
          "per_seed": [
            0.2179270088672638,
            0.1916452795267105,
            0.1942574828863144,
            0.26538917422294617
          ],
          "n": 4
        }
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 8,
          "lam": 0.3
        },
        "result": {
          "mean": 0.26101431995630264,
          "std": 0.014287884388682429,
          "per_seed": [
            0.2409307062625885,
            0.26482611894607544,
            0.2576110064983368,
            0.28068944811820984
          ],
          "n": 4
        }
      }
    ]
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001,
        "epochs": 8,
        "lam": 0.3
      },
      "mean": 0.19120756536722183
    },
    {
      "cfg": {
        "lr": 0.003,
        "epochs": 8,
        "lam": 0.3
      },
      "mean": 0.21730473637580872
    },
    {
      "cfg": {
        "lr": 0.006,
        "epochs": 8,
        "lam": 0.3
      },
      "mean": 0.26101431995630264
    }
  ]
}