{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006,
      "epochs": 12
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 12
        },
        "mean": 0.004298779065720737
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 12
        },
        "mean": 0.0034615940821822733
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 12
        },
        "mean": 0.0018282315868418664
      }
    ],
    "full": {
      "mean": 0.002105819425196387,
      "std": 0.000389436566226016,
      "per_seed": [
        0.002004395006224513,
        0.001437882543541491,
        0.0015748648438602686,
        0.002295783953741193,
        0.0024020641576498747,
        0.0024898042902350426,
        0.0020851334556937218,
        0.0025566271506249905
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.0018193616560893133,
    "std": 0.00037736014800368497,
    "per_seed": [
      0.0017986779566854239,
      0.0012207203544676304,
      0.0014683871995657682,
      0.0016677540261298418,
      0.0024448649492114782,
      0.002230470534414053,
      0.0020495401695370674,
      0.0016744780587032437
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.00028645776910707355,
    "idea_wins": 7,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.0002057170495390892,
      -0.00021716218907386065,
      -0.00010647764429450035,
      -0.000628029927611351,
      4.2800791561603546e-05,
      -0.0002593337558209896,
      -3.559328615665436e-05,
      -0.0008821490919217467
    ],
    "p_value": 0.0224,
    "mde": 0.0002620079869585246,
    "mde_rel_pct": 12.442091844322785,
    "verdict": "idea better (significant)",
    "system_worked": true
  },
  "mechanism_signature": {
    "signature": {
      "q": 0.03,
      "Gamma": 2.0,
      "epsilon": 0.13,
      "predicted_uniform_radius": 0.17285714285714285,
      "observed_proxy_radius_mean": 0.1817476600408554,
      "observed_proxy_radius_std": 0.02882220596075058,
      "predicted_vs_observed_correlation": NaN,
      "controller_refines_when_bound_exceeds_epsilon": true,
      "confirmed": false,
      "note": "trained benchmark models use adaptive resampling, but the conservative fixed-gap law is not empirically calibrated by this track"
    },
    "idea_sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 12
        },
        "result": {
          "mean": 0.0052960661123506725,
          "std": 0.0015572022789016526,
          "per_seed": [
            0.008267243392765522,
            0.005108773708343506,
            0.0028678099624812603,
            0.004508262034505606,
            0.005263541359454393,
            0.004577877465635538,
            0.0047037601470947266,
            0.007071260828524828
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 12
        },
        "result": {
          "mean": 0.002939960832009092,
          "std": 0.0017693752721639644,
          "per_seed": [
            0.0019986729603260756,
            0.002653212519362569,
            0.0012997412122786045,
            0.007327801547944546,
            0.0022197901271283627,
            0.0036344309337437153,
            0.0022852604743093252,
            0.002100776880979538
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "lr": 0.006,
          "epochs": 12
        },
        "result": {
          "mean": 0.0018193616560893133,
          "std": 0.00037736014800368497,
          "per_seed": [
            0.0017986779566854239,
            0.0012207203544676304,
            0.0014683871995657682,
            0.0016677540261298418,
            0.0024448649492114782,
            0.002230470534414053,
            0.0020495401695370674,
            0.0016744780587032437
          ],
          "n": 8
        }
      }
    ]
  },
  "idea_selected_cfg": {
    "lr": 0.006,
    "epochs": 12
  },
  "protocol_note": "Baseline and idea share rnn_small, data, seeds, epochs and learning-rate union; adaptive temporal resampling is the sole intervention."
}