{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small_matched_spectral_head",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01,
      "epochs": 18
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 18
        },
        "mean": 0.009911961504258215
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 18
        },
        "mean": 0.004386652202811092
      },
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 18
        },
        "mean": 0.0008802848460618407
      }
    ],
    "full": {
      "mean": 0.0006743472804373596,
      "std": 0.00026491514577006763,
      "per_seed": [
        0.0005103786825202405,
        0.001095399260520935,
        0.0009623044752515852,
        0.0009530569659546018,
        0.0005656004068441689,
        0.0003435070684645325,
        0.0004539852379821241,
        0.0005105461459606886
      ],
      "n": 8
    }
  },
  "idea": {
    "cfg": {
      "lr": 0.01,
      "epochs": 18
    },
    "mean": 0.0902025755494833,
    "std": 0.014189437586170238,
    "per_seed": [
      0.07012719660997391,
      0.09647830575704575,
      0.08879011869430542,
      0.11882545053958893,
      0.09636852890253067,
      0.07268789410591125,
      0.08943840861320496,
      0.08890470117330551
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.08952822826904594,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.06961681792745367,
      0.09538290649652481,
      0.08782781421905383,
      0.11787239357363433,
      0.0958029284956865,
      0.07234438703744672,
      0.08898442337522283,
      0.08839415502734482
    ],
    "p_value": 0.0081,
    "mde": 0.012535387409262248,
    "mde_rel_pct": 1858.8919645575206,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.001,
        "epochs": 18
      },
      "mean": 0.35300457291305065,
      "std": 0.08097430981045273,
      "per_seed": [
        0.22397439181804657,
        0.44731879234313965,
        0.26465779542922974,
        0.43655097484588623,
        0.34434905648231506,
        0.45171505212783813,
        0.35042715072631836,
        0.30504336953163147
      ],
      "n": 8
    },
    {
      "cfg": {
        "lr": 0.003,
        "epochs": 18
      },
      "mean": 0.1979203224182129,
      "std": 0.038848589270492305,
      "per_seed": [
        0.1307593286037445,
        0.23984569311141968,
        0.1644514948129654,
        0.25781843066215515,
        0.19592295587062836,
        0.22193168103694916,
        0.19744279980659485,
        0.17519019544124603
      ],
      "n": 8
    },
    {
      "cfg": {
        "lr": 0.01,
        "epochs": 18
      },
      "mean": 0.0902025755494833,
      "std": 0.014189437586170238,
      "per_seed": [
        0.07012719660997391,
        0.09647830575704575,
        0.08879011869430542,
        0.11882545053958893,
        0.09636852890253067,
        0.07268789410591125,
        0.08943840861320496,
        0.08890470117330551
      ],
      "n": 8
    }
  ],
  "structural_match": "dynamics: controlled pendulum rollout",
  "mechanism_signature": {
    "prediction": "noncommuting Hamiltonian terms should create a measurable nonlinear readout",
    "commutator_frobenius": 4.0,
    "trained_idea_test_mse_mean": 0.0902025755494833,
    "trained_idea_nonlinear_residual_mean": 0.1242971978952532,
    "confirmed": true
  },
  "budget": {
    "epochs": 18,
    "n_train": 400,
    "n_test": 400,
    "seeds": 8
  }
}