{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 0.0021690295106964186
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.001074885185516905
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 0.00054517226817552
      }
    ],
    "full": {
      "mean": 0.00054517226817552,
      "std": 0.00020361842727135583,
      "per_seed": [
        0.0004930912400595844,
        0.0004765945195686072,
        0.0004673648509196937,
        0.0010376222198829055,
        0.000511663849465549,
        0.00027233464061282575,
        0.0005571767687797546,
        0.0005455300561152399
      ],
      "n": 8
    },
    "protocol_note": "8 paired seeds; sweep grid shared with idea; baseline is canonical rnn_small."
  },
  "idea": {
    "mean": 0.0016794730763649568,
    "std": 0.0005899470001703297,
    "per_seed": [
      0.00201487448066473,
      0.0009998989989981055,
      0.0017773952567949891,
      0.0030421584378927946,
      0.001546183368191123,
      0.001343107083812356,
      0.0013974033063277602,
      0.0013147636782377958
    ],
    "n": 8,
    "cfg": {
      "lr": 0.003
    }
  },
  "comparison": {
    "delta_mean": 0.0011343008081894368,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.0015217832406051457,
      0.0005233044794294983,
      0.0013100304058752954,
      0.002004536218009889,
      0.001034519518725574,
      0.0010707724431995302,
      0.0008402265375480056,
      0.000769233622122556
    ],
    "p_value": 0.0081,
    "mde": 0.0003928773272119451,
    "mde_rel_pct": 72.06480412636415,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "method_note": "Braid transport replaces the GRU recurrence; same dynamics data, epochs, batch, and lr grid.",
  "mechanism_signature": {
    "predicted": "skew residual near zero and transport-only norm drift near zero",
    "observed_mean_skew_residual": 0.0,
    "observed_mean_transport_norm_drift": 7.996568260337789e-08,
    "confirmed": true
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001
      },
      "mean": 0.002948012319393456,
      "per_seed": [
        0.003085256787016988,
        0.002189615275710821,
        0.0050039333291351795,
        0.0036421462427824736,
        0.0022776462137699127,
        0.001872690161690116,
        0.0024267486296594143,
        0.003086061915382743
      ]
    },
    {
      "cfg": {
        "lr": 0.003
      },
      "mean": 0.0016794730763649568,
      "per_seed": [
        0.00201487448066473,
        0.0009998989989981055,
        0.0017773952567949891,
        0.0030421584378927946,
        0.001546183368191123,
        0.001343107083812356,
        0.0013974033063277602,
        0.0013147636782377958
      ]
    },
    {
      "cfg": {
        "lr": 0.01
      },
      "mean": 0.0018830011467798613,
      "per_seed": [
        0.0020360236521810293,
        0.0009326969156973064,
        0.00206544017419219,
        0.0026314484421163797,
        0.0011524695437401533,
        0.0020754176657646894,
        0.001974960323423147,
        0.002195552457123995
      ]
    }
  ]
}