{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01,
      "penalty": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "penalty": 0.0
        },
        "mean": 0.11797566118184477
      },
      {
        "cfg": {
          "lr": 0.003,
          "penalty": 0.0
        },
        "mean": 0.007828571717254817
      },
      {
        "cfg": {
          "lr": 0.01,
          "penalty": 0.0
        },
        "mean": 0.0025760137650649995
      }
    ],
    "full": {
      "mean": 0.002240322035504505,
      "std": 0.0007111673246843795,
      "per_seed": [
        0.0027267802506685257,
        0.003143619978800416,
        0.0015068432549014688,
        0.0029268115758895874,
        0.0025003517512232065,
        0.001474660588428378,
        0.002521744929254055,
        0.0011217639548704028
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.0023943046835483983,
    "std": 0.0006552643263981685,
    "per_seed": [
      0.001565085374750197,
      0.001931023900397122,
      0.003373584244400263,
      0.002406168496236205,
      0.0027771489694714546,
      0.0014064890565350652,
      0.002705090679228306,
      0.002989846747368574
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.00015398264804389328,
    "idea_wins": 4,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.0011616948759183288,
      -0.001212596078403294,
      0.001866740989498794,
      -0.0005206430796533823,
      0.0002767972182482481,
      -6.817153189331293e-05,
      0.0001833457499742508,
      0.0018680827924981713
    ],
    "p_value": 0.69925,
    "mde": 0.0009989272382128146,
    "mde_rel_pct": 44.58855568002585,
    "verdict": "no significant win",
    "system_worked": false
  },
  "math_sanity": {
    "a": 0.8,
    "sigma": [
      [
        0.001565085374750197,
        1.0062328577041626,
        1.0018754711727258
      ],
      [
        0.001931023900397122,
        1.0306692123413086,
        1.001999129195911
      ],
      [
        0.003373584244400263,
        1.0256351232528687,
        1.0008405618216547
      ],
      [
        0.002406168496236205,
        1.0819001197814941,
        1.000112452055684
      ],
      [
        0.0027771489694714546,
        0.9877385497093201,
        1.001914544541687
      ],
      [
        0.0014064890565350652,
        1.1168031692504883,
        1.0022314286996865
      ],
      [
        0.002705090679228306,
        1.0017019510269165,
        0.9949402147998869
      ],
      [
        0.002989846747368574,
        0.9434056282043457,
        0.9980177719281845
      ]
    ],
    "K_scalar": 1.0
  },
  "idea_sweep": {
    "configs": [
      {
        "lr": 0.01,
        "penalty": 1.0
      },
      {
        "lr": 0.001,
        "penalty": 3.0
      },
      {
        "lr": 0.003,
        "penalty": 10.0
      }
    ],
    "best_cfg": {
      "lr": 0.01,
      "penalty": 1.0
    }
  },
  "mechanism_signature": {
    "noise_sigma": 0.2,
    "predicted_lifted_rho_mean": 1.024260826408863,
    "observed_hidden_second_moment_ratio_mean": 1.0002414467769276,
    "relative_error": 0.02345045227996197,
    "confirmed": true,
    "source": "trained idea GRU models on dynamics test rollout"
  }
}