{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.005
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001
        },
        "mean": 0.004298779065720737
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.0034615940821822733
      },
      {
        "cfg": {
          "lr": 0.005
        },
        "mean": 0.0015488484350498766
      }
    ],
    "full": {
      "mean": 0.0015987288497854024,
      "std": 0.0004723450985826372,
      "per_seed": [
        0.0010614258935675025,
        0.0021564248017966747,
        0.001086847041733563,
        0.0018906960031017661,
        0.0020614659879356623,
        0.0012517774011939764,
        0.0011406542034819722,
        0.002140539465472102
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.006309549498837441,
    "std": 0.0013831181007421675,
    "per_seed": [
      0.004766013007611036,
      0.0075876968912780285,
      0.00614228006452322,
      0.005290781147778034,
      0.005994749721139669,
      0.0061586033552885056,
      0.009300675243139267,
      0.005235596559941769
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.004710820649052039,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.003704587114043534,
      0.005431272089481354,
      0.005055433022789657,
      0.003400085144676268,
      0.003933283733204007,
      0.004906825954094529,
      0.008160021039657295,
      0.0030950570944696665
    ],
    "p_value": 0.0081,
    "mde": 0.0013600218273413377,
    "mde_rel_pct": 85.06894884169343,
    "verdict": "idea worse (significant)",
    "system_worked": false
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.005,
        "step": 0.04
      },
      "result": {
        "mean": 0.006823674135375768,
        "std": 0.0018729123134178251,
        "per_seed": [
          0.005168338771909475,
          0.010784340091049671,
          0.0066774641163647175,
          0.005866168532520533,
          0.0061936150304973125,
          0.005813745316118002,
          0.008938142098486423,
          0.005147579126060009
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.005,
        "step": 0.08
      },
      "result": {
        "mean": 0.006543989293277264,
        "std": 0.00158231784982001,
        "per_seed": [
          0.005307216662913561,
          0.009000664576888084,
          0.0062515996396541595,
          0.0055503034964203835,
          0.005550468806177378,
          0.006075830198824406,
          0.00943143293261528,
          0.005184398032724857
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.005,
        "step": 0.12
      },
      "result": {
        "mean": 0.006309549498837441,
        "std": 0.0013831181007421675,
        "per_seed": [
          0.004766013007611036,
          0.0075876968912780285,
          0.00614228006452322,
          0.005290781147778034,
          0.005994749721139669,
          0.0061586033552885056,
          0.009300675243139267,
          0.005235596559941769
        ],
        "n": 8
      }
    }
  ],
  "audit": {
    "n_train": 400,
    "n_test": 200,
    "epochs": 12,
    "matched_architecture_role": "GRU replaced by symplectic hidden transition",
    "baseline_lr_union": [
      0.001,
      0.003,
      0.005
    ]
  },
  "mechanism_signature": {
    "prediction": "symplectic transition det should remain near 1 (log|det| near 0)",
    "observed_idea_log_abs_det": 0.00387675897218287,
    "observed_baseline_log_abs_det": -42.42223358154297,
    "absolute_logdet_error_idea": 0.00387675897218287,
    "absolute_logdet_error_baseline": 42.42223358154297,
    "confirmed": true
  }
}