{
  "bench_version": 1,
  "track": "dynamics",
  "model": "factor_rnn",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006,
      "epochs": 18,
      "beta": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.0015,
          "epochs": 18,
          "beta": 0.0
        },
        "mean": 0.11639470420777798
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 18,
          "beta": 0.0
        },
        "mean": 0.058107024524360895
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 18,
          "beta": 0.0
        },
        "mean": 0.026168195763602853
      }
    ],
    "full": {
      "mean": 0.02109241997823119,
      "std": 0.014265532680015599,
      "per_seed": [
        0.05811716988682747,
        0.019254598766565323,
        0.013698350638151169,
        0.01360266376286745,
        0.014953133650124073,
        0.012289082631468773,
        0.021011028438806534,
        0.015813332051038742
      ],
      "n": 8
    },
    "full_all_union": [
      {
        "cfg": {
          "lr": 0.0015,
          "epochs": 18,
          "beta": 0.0
        },
        "mean": 0.08842666260898113,
        "std": 0.06345898927987136,
        "per_seed": [
          0.2491457462310791,
          0.04528465121984482,
          0.08107621967792511,
          0.09007219970226288,
          0.05925949290394783,
          0.04771618917584419,
          0.09064487367868423,
          0.044213928282260895
        ],
        "n": 8
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 18,
          "beta": 0.0
        },
        "mean": 0.04415802680887282,
        "std": 0.03152132845115148,
        "per_seed": [
          0.1255890429019928,
          0.02912258543074131,
          0.03296443447470665,
          0.04475203529000282,
          0.029448043555021286,
          0.025108061730861664,
          0.041623830795288086,
          0.024656180292367935
        ],
        "n": 8
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 18,
          "beta": 0.0
        },
        "mean": 0.02109241997823119,
        "std": 0.014265532680015599,
        "per_seed": [
          0.05811716988682747,
          0.019254598766565323,
          0.013698350638151169,
          0.01360266376286745,
          0.014953133650124073,
          0.012289082631468773,
          0.021011028438806534,
          0.015813332051038742
        ],
        "n": 8
      }
    ]
  },
  "idea": {
    "mean": 0.022512523224577308,
    "std": 0.012331270086282655,
    "per_seed": [
      0.04500043764710426,
      0.017105672508478165,
      0.03219568356871605,
      0.034598458558321,
      0.01256189402192831,
      0.008819940499961376,
      0.019357675686478615,
      0.010460423305630684
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.001420103246346116,
    "idea_wins": 6,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.013116732239723206,
      -0.002148926258087158,
      0.01849733293056488,
      0.02099579479545355,
      -0.0023912396281957626,
      -0.0034691421315073967,
      -0.001653352752327919,
      -0.005352908745408058
    ],
    "p_value": 0.78325,
    "mde": 0.009957143075416026,
    "mde_rel_pct": 47.20721038976311,
    "verdict": "no significant win",
    "system_worked": false
  },
  "mechanism_signature": {
    "track_structure": "controlled pendulum multi-step dynamics",
    "metric": 0.04500043764710426,
    "max_component_violation": -0.35021981596946716,
    "max_observed_ratio": 0.7172038555145264,
    "predicted_bound": "nonpositive violation",
    "observed_envelope_row_sum": 2.4505488872528076,
    "observed_sum_abs_A": 45.71932601928711,
    "observed_sum_P": 65.28011322021484,
    "confirmed": true,
    "config": {
      "lr": 0.006,
      "epochs": 18,
      "beta": 0.02,
      "target": 0.92
    },
    "wall_seconds": 27.75654865699471
  },
  "idea_grid": [
    {
      "cfg": {
        "lr": 0.0015,
        "epochs": 18,
        "beta": 0.005,
        "target": 0.92
      },
      "mean": 0.09474506694823503,
      "std": 0.10094635313979229,
      "per_seed": [
        0.35983920097351074,
        0.059099163860082626,
        0.058137185871601105,
        0.07874950021505356,
        0.054427292197942734,
        0.044902503490448,
        0.0674169585108757,
        0.035388730466365814
      ],
      "n": 8
    },
    {
      "cfg": {
        "lr": 0.003,
        "epochs": 18,
        "beta": 0.01,
        "target": 0.92
      },
      "mean": 0.036793723003938794,
      "std": 0.02819267034786451,
      "per_seed": [
        0.1090393140912056,
        0.0328064002096653,
        0.026808833703398705,
        0.040870074182748795,
        0.023899797350168228,
        0.01856924220919609,
        0.024410061538219452,
        0.017946060746908188
      ],
      "n": 8
    },
    {
      "cfg": {
        "lr": 0.006,
        "epochs": 18,
        "beta": 0.02,
        "target": 0.92
      },
      "mean": 0.022512523224577308,
      "std": 0.012331270086282655,
      "per_seed": [
        0.04500043764710426,
        0.017105672508478165,
        0.03219568356871605,
        0.034598458558321,
        0.01256189402192831,
        0.008819940499961376,
        0.019357675686478615,
        0.010460423305630684
      ],
      "n": 8
    }
  ],
  "device": "cuda",
  "note": "Baseline and idea are end-to-end trained systems with identical FactorRNN architecture; only envelope loss differs. All shared learning rates were evaluated on the same eight paired seeds."
}