{
  "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": 0.0032877353951334953
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.0012035724357701838
      },
      {
        "cfg": {
          "lr": 0.01
        },
        "mean": 0.0006944253206408272
      }
    ],
    "full": {
      "mean": 0.0006787698075640947,
      "std": 0.00018311611365834393,
      "per_seed": [
        0.0008624937618151307,
        0.0008604836766608059,
        0.000360298523446545,
        0.0006422948790714145,
        0.0006247925339266658,
        0.0009110610117204487,
        0.0004747614439111203,
        0.0006939726299606264
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.0002450490219416679,
    "std": 0.0001979632966612432,
    "per_seed": [
      0.00021305571135599166,
      0.0002364848623983562,
      0.0007525552064180374,
      0.0001690309727564454,
      0.00017366997781209648,
      8.526047167833894e-05,
      0.0002184970217058435,
      0.0001118379514082335
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.0004337207856224268,
    "idea_wins": 7,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.000649438050459139,
      -0.0006239988142624497,
      0.0003922566829714924,
      -0.00047326390631496906,
      -0.0004511225561145693,
      -0.0008258005400421098,
      -0.0002562644222052768,
      -0.0005821346785523929
    ],
    "p_value": 0.0231,
    "mde": 0.00031191512145065987,
    "mde_rel_pct": 45.953004682107405,
    "verdict": "idea better (significant)",
    "system_worked": true
  },
  "math_check": {
    "rls_batch_max_abs_error": 5.551115123125783e-17,
    "predicted_adaptation_timescale": 14.285714285714295
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001
      },
      "result": {
        "mean": 0.00044900484863319434,
        "std": 0.00010002042742588506,
        "per_seed": [
          0.0007017979514785111,
          0.0004040250787511468,
          0.00040901819011196494,
          0.00039653119165450335,
          0.00039476461824961007,
          0.0004773136170115322,
          0.00043600559001788497,
          0.00037258255179040134
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.003
      },
      "result": {
        "mean": 0.00025152570015052333,
        "std": 5.887341333514287e-05,
        "per_seed": [
          0.0002252471458632499,
          0.00016258451796602458,
          0.0003189823473803699,
          0.0002701112534850836,
          0.00025544039090164006,
          0.00024129776284098625,
          0.00035137461964040995,
          0.00018716756312642246
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.01
      },
      "result": {
        "mean": 0.0002450490219416679,
        "std": 0.0001979632966612432,
        "per_seed": [
          0.00021305571135599166,
          0.0002364848623983562,
          0.0007525552064180374,
          0.0001690309727564454,
          0.00017366997781209648,
          8.526047167833894e-05,
          0.0002184970217058435,
          0.0001118379514082335
        ],
        "n": 8
      }
    }
  ],
  "protocol_note": "Baseline sweep uses same lr union; final comparison is eight paired seeds.",
  "mechanism_signature": {
    "rho_trained_A": 0.7977778911590576,
    "predicted_asymptotic_norm_ratio": 0.7977778911590576,
    "observed_mean_last3_norm_ratio": 0.8155918888124241,
    "horizon_norms": [
      4.329426288604736,
      1.5744551420211792,
      0.8551374077796936,
      0.697711169719696,
      0.6099732518196106,
      0.518440306186676,
      0.42884668707847595,
      0.3488800525665283,
      0.2812177240848541
    ],
    "confirmed": true,
    "source": "trained benchmark Koopman model"
  }
}