{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 0.5088173821568489
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.06073589529842138
      },
      {
        "cfg": {
          "lr": 0.006
        },
        "mean": 0.023900731466710567
      }
    ],
    "full": {
      "mean": 0.019088300527073443,
      "std": 0.00558386015869873,
      "per_seed": [
        0.018445175141096115,
        0.023440849035978317,
        0.02936750277876854,
        0.024349398910999298,
        0.01488791685551405,
        0.012780136428773403,
        0.01545807532966137,
        0.013977349735796452
      ],
      "n": 8
    }
  },
  "idea": {
    "best_cfg": {
      "lr": 0.006
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 0.3442936111241579,
        "std": 0.11055745967503162,
        "per_seed": [
          0.4150988757610321,
          0.572791337966919,
          0.3577822744846344,
          0.3750651478767395,
          0.2974511682987213,
          0.17869944870471954,
          0.254057914018631,
          0.30340272188186646
        ],
        "n": 8
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.03436855808831751,
        "std": 0.01405565796915205,
        "per_seed": [
          0.05125405639410019,
          0.041576966643333435,
          0.04229721054434776,
          0.055242300033569336,
          0.021142112091183662,
          0.016799848526716232,
          0.020402031019330025,
          0.026233939453959465
        ],
        "n": 8
      },
      {
        "cfg": {
          "lr": 0.006
        },
        "mean": 0.015928541193716228,
        "std": 0.004679734568777004,
        "per_seed": [
          0.0178981926292181,
          0.013649515807628632,
          0.024342793971300125,
          0.021126413717865944,
          0.012977873906493187,
          0.008715411648154259,
          0.01320882048457861,
          0.015509307384490967
        ],
        "n": 8
      }
    ],
    "mean": 0.015928541193716228,
    "std": 0.004679734568777004,
    "per_seed": [
      0.0178981926292181,
      0.013649515807628632,
      0.024342793971300125,
      0.021126413717865944,
      0.012977873906493187,
      0.008715411648154259,
      0.01320882048457861,
      0.015509307384490967
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.003159759333357215,
    "idea_wins": 7,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.0005469825118780136,
      -0.009791333228349686,
      -0.005024708807468414,
      -0.003222985193133354,
      -0.0019100429490208626,
      -0.0040647247806191444,
      -0.00224925484508276,
      0.0015319576486945152
    ],
    "p_value": 0.02315,
    "mde": 0.002818017003243901,
    "mde_rel_pct": 14.763058655992097,
    "verdict": "idea better (significant)",
    "system_worked": true
  },
  "mechanism_signature": {
    "collective_detectability": {
      "window": 2,
      "predicted_lambda_min_positive": true,
      "observed_mean_fused_information": 1.459951639175415,
      "observed_positive_fraction": 1.0,
      "tolerance": "positivity and >0.99 positive fraction",
      "confirmed": true
    }
  },
  "protocol_notes": {
    "structural_match": "dynamics: recurrent pendulum rollout and latent-state stability",
    "paired_seeds": 8,
    "epochs": 12,
    "batch": 128,
    "shared_lr_union": [
      0.001,
      0.003,
      0.006
    ],
    "baseline_method_knob": "equal arithmetic mean (fixed by definition)",
    "system_parity": "separately trained identical two-agent GRU encoders and predictor; only fusion differs"
  }
}