{
  "bench_version": 1,
  "track": "sequence",
  "model": "transformer_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.001,
      "temperature": 1.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "temperature": 1.0
        },
        "mean": 0.34783465415239334
      },
      {
        "cfg": {
          "lr": 0.003,
          "temperature": 1.0
        },
        "mean": 0.3491433709859848
      },
      {
        "cfg": {
          "lr": 0.005,
          "temperature": 1.0
        },
        "mean": 0.40824612975120544
      }
    ],
    "full": {
      "mean": 0.36294760182499886,
      "std": 0.054345281417173805,
      "per_seed": [
        0.39175504446029663,
        0.3038984537124634,
        0.30141177773475647,
        0.3942733407020569,
        0.35095956921577454,
        0.3010490834712982,
        0.4579026699066162,
        0.4023308753967285
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.3599541373550892,
    "std": 0.05518226062931197,
    "per_seed": [
      0.3880409896373749,
      0.2888491451740265,
      0.32456186413764954,
      0.3811890184879303,
      0.3370722830295563,
      0.29420310258865356,
      0.4586992561817169,
      0.40701743960380554
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.002993464469909668,
    "idea_wins": 5,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.003714054822921753,
      -0.01504930853843689,
      0.023150086402893066,
      -0.013084322214126587,
      -0.013887286186218262,
      -0.006845980882644653,
      0.000796586275100708,
      0.004686564207077026
    ],
    "p_value": 0.53235,
    "mde": 0.010661430606466927,
    "mde_rel_pct": 2.9374572397939445,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "mechanism_signature": {
    "idea_config": {
      "lr": 0.001,
      "rank": 4
    },
    "candidate_results": [
      {
        "cfg": {
          "lr": 0.001,
          "rank": 2
        },
        "result": {
          "mean": 0.3600270599126816,
          "std": 0.05461125040141159,
          "per_seed": [
            0.3894391655921936,
            0.2890293002128601,
            0.3245948255062103,
            0.3806515336036682,
            0.33679887652397156,
            0.2958018183708191,
            0.4570717513561249,
            0.40682920813560486
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "lr": 0.001,
          "rank": 4
        },
        "result": {
          "mean": 0.3599541373550892,
          "std": 0.05518226062931197,
          "per_seed": [
            0.3880409896373749,
            0.2888491451740265,
            0.32456186413764954,
            0.3811890184879303,
            0.3370722830295563,
            0.29420310258865356,
            0.4586992561817169,
            0.40701743960380554
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "lr": 0.001,
          "rank": 6
        },
        "result": {
          "mean": 0.3600568361580372,
          "std": 0.05501619060413237,
          "per_seed": [
            0.38814911246299744,
            0.2887286841869354,
            0.32432663440704346,
            0.38098469376564026,
            0.3378050625324249,
            0.29527634382247925,
            0.45912855863571167,
            0.40605559945106506
          ],
          "n": 8
        }
      }
    ],
    "signature": {
      "epsilon": 0.001,
      "predicted_rank_bound": 5,
      "observed_rank": 4,
      "trained_model_cross_output_relative_error": 3.261449819547124e-05,
      "predicted_error_bound": 0.001,
      "confirmed": true
    }
  },
  "protocol_notes": {
    "paired_seeds": [
      0,
      1,
      2,
      3,
      4,
      5,
      6,
      7
    ],
    "epochs": 20,
    "architecture_difference_only": "dense self-attention vs within-side dense plus cached dyadic Hankel cross branch"
  }
}