{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01,
      "epochs": 10,
      "sigma": 0.0
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "epochs": 10,
          "sigma": 0.0
        },
        "mean": 0.009392368956469
      },
      {
        "cfg": {
          "lr": 0.003,
          "epochs": 10,
          "sigma": 0.0
        },
        "mean": 0.005014403606764972
      },
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 10,
          "sigma": 0.0
        },
        "mean": 0.004590500087942928
      }
    ],
    "full": {
      "mean": 0.003632939886301756,
      "std": 0.001261873819762896,
      "per_seed": [
        0.005352384876459837,
        0.005787275265902281,
        0.0038195690140128136,
        0.003402771195396781,
        0.0029491649474948645,
        0.00237746792845428,
        0.003468934213742614,
        0.0019059516489505768
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.004008013929706067,
    "std": 0.0023638783550923134,
    "per_seed": [
      0.005741244647651911,
      0.003205370157957077,
      0.0026500194799154997,
      0.0029542287811636925,
      0.0027808116283267736,
      0.0022908966057002544,
      0.0027672694995999336,
      0.009674270637333393
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": 0.00037507404340431094,
    "idea_wins": 6,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.0003888597711920738,
      -0.0025819051079452038,
      -0.0011695495340973139,
      -0.0004485424142330885,
      -0.00016835331916809082,
      -8.657132275402546e-05,
      -0.0007016647141426802,
      0.007768318988382816
    ],
    "p_value": 0.9688,
    "mde": 0.002608707656005132,
    "mde_rel_pct": 71.80706914093018,
    "verdict": "no significant win",
    "system_worked": false
  },
  "mechanism_signature": {
    "mechanism_signature": {
      "prediction": "trained recurrent hidden temporal energy is nonzero and increases with sigma^2",
      "trained_model_observed": [
        {
          "sigma": 0.01,
          "sigma2": 0.0001,
          "observed_temporal_energy": 0.018960949033498764
        },
        {
          "sigma": 0.03,
          "sigma2": 0.0009,
          "observed_temporal_energy": 0.018547870218753815
        },
        {
          "sigma": 0.1,
          "sigma2": 0.010000000000000002,
          "observed_temporal_energy": 0.018667757511138916
        }
      ],
      "baseline_sigma0": {
        "metric": 0.005352384876459837,
        "hidden_variance": 0.19951528310775757,
        "temporal_energy": 0.020083513110876083,
        "sigma": 0.0
      },
      "best_sigma": {
        "metric": 0.005741244647651911,
        "hidden_variance": 0.19854561984539032,
        "temporal_energy": 0.018960949033498764,
        "sigma": 0.01
      },
      "through_origin_slope": 2.035972557147564,
      "through_origin_r2": -7019.474509912291,
      "confirmed": false,
      "note": "measured from trained benchmark model hidden trajectories"
    },
    "idea_sweep": [
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 10,
          "sigma": 0.01
        },
        "result": {
          "mean": 0.004008013929706067,
          "std": 0.0023638783550923134,
          "per_seed": [
            0.005741244647651911,
            0.003205370157957077,
            0.0026500194799154997,
            0.0029542287811636925,
            0.0027808116283267736,
            0.0022908966057002544,
            0.0027672694995999336,
            0.009674270637333393
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 10,
          "sigma": 0.03
        },
        "result": {
          "mean": 0.004176608053967357,
          "std": 0.002662518875846031,
          "per_seed": [
            0.005716613493859768,
            0.0032172678038477898,
            0.0026425563264638186,
            0.0028939577750861645,
            0.002898444654420018,
            0.0025497509632259607,
            0.0027517646085470915,
            0.010742508806288242
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 10,
          "sigma": 0.1
        },
        "result": {
          "mean": 0.0049640603247098625,
          "std": 0.0029402421909946234,
          "per_seed": [
            0.005744725465774536,
            0.0041050356812775135,
            0.0026244467590004206,
            0.0034257906954735518,
            0.004549694713205099,
            0.0031880675815045834,
            0.0036924462765455246,
            0.01238227542489767
          ],
          "n": 8
        }
      }
    ],
    "track_match": "stability/control and recurrent state evolution -> dynamics"
  },
  "custom_track": null
}