{
  "bench_version": 1,
  "track": "sequence",
  "model": "frozen_random_attention_prompt",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 1.0786777287721634
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 1.0726252794265747
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 0.921186164021492
      }
    ],
    "full": {
      "mean": 0.9356768578290939,
      "std": 0.10686743285939519,
      "per_seed": [
        1.0163524150848389,
        0.7307133078575134,
        1.0717743635177612,
        0.8659045696258545,
        0.892680823802948,
        1.0664112567901611,
        0.9147705435752869,
        0.9268075823783875
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 1.7406605532973132,
    "std": 0.15897193757942363,
    "per_seed": [
      2.0928197670190434,
      1.6468335788158937,
      1.6924736704763956,
      1.668518826794135,
      1.6360904797108695,
      1.7143411545296539,
      1.89549804751905,
      1.5787089015134637
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.8049836954682192,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      1.0764673519342045,
      0.9161202709583802,
      0.6206993069586344,
      0.8026142571682806,
      0.7434096559079215,
      0.6479298977394927,
      0.9807275039437631,
      0.6519013191350762
    ],
    "p_value": 0.0081,
    "mde": 0.14232114117001493,
    "mde_rel_pct": 15.21050135836646,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "mechanism_signature": {
    "track_choice": "sequence: multi-token temporal-window correlations require sequence attention",
    "best_sigma": 0.8,
    "idea_sweep": [
      {
        "cfg": {
          "sigma": 0.4,
          "lr": 0.001
        },
        "result": {
          "mean": 1.8291889989446934,
          "std": 0.16079926338695508,
          "per_seed": [
            2.187105139019409,
            1.7452456227531217,
            1.7904076713809234,
            1.7679971041431175,
            1.6905353869220097,
            1.8289776945690013,
            1.9642818351456248,
            1.6589615376243407
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "sigma": 0.6,
          "lr": 0.003
        },
        "result": {
          "mean": 1.7922968691831327,
          "std": 0.15999984262010897,
          "per_seed": [
            2.1471634715202894,
            1.7083331174848266,
            1.74349576800059,
            1.722126782043419,
            1.669563403023254,
            1.7793366851501542,
            1.940943109339341,
            1.6274126169031866
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "sigma": 0.8,
          "lr": 0.01
        },
        "result": {
          "mean": 1.7406605532973132,
          "std": 0.15897193757942363,
          "per_seed": [
            2.0928197670190434,
            1.6468335788158937,
            1.6924736704763956,
            1.668518826794135,
            1.6360904797108695,
            1.7143411545296539,
            1.89549804751905,
            1.5787089015134637
          ],
          "n": 8
        }
      }
    ],
    "trained_baseline_behavior": {
      "0": {
        "prompt_norm": 19.529014587402344,
        "attention_entropy": 4.115479946136475
      },
      "1": {
        "prompt_norm": 15.537064552307129,
        "attention_entropy": 4.137622833251953
      },
      "2": {
        "prompt_norm": 9.14969253540039,
        "attention_entropy": 4.145155906677246
      },
      "3": {
        "prompt_norm": 17.088436126708984,
        "attention_entropy": 4.135089874267578
      },
      "4": {
        "prompt_norm": 19.799318313598633,
        "attention_entropy": 4.124661445617676
      },
      "5": {
        "prompt_norm": 11.53364086151123,
        "attention_entropy": 4.150853157043457
      },
      "6": {
        "prompt_norm": 18.757871627807617,
        "attention_entropy": 4.126466751098633
      },
      "7": {
        "prompt_norm": 12.221356391906738,
        "attention_entropy": 4.149312973022461
      }
    },
    "mechanism_prediction": {
      "prediction": "prompt norm scales as sigma^-2; affine logit residual is near zero",
      "observed_prompt_norms": {
        "0.4": 2036.919040636862,
        "0.6": 905.2893090361015,
        "0.8": 509.21965520351125
      },
      "predicted_norm_ratio_sigma_0.4_to_0.8": 4.0,
      "observed_norm_ratio_sigma_0.4_to_0.8": 4.000079376006806,
      "observed_best_logit_residual": 8.765288996216841e-14,
      "confirmed": true
    }
  },
  "baseline_sweep_shared_lr_grid": [
    0.001,
    0.003,
    0.01
  ],
  "cut": "Only sequence track tested; no CIFAR/MNIST transfer and no latency/FLOP study."
}