Complementary-Channel Switched Latent Observer / bench_report.json

Failed on benchmark

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "dynamics",
  4  "model": "custom_switched_observer_rnn",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.006,
 10      "residual_scale": 0.0
 11    },
 12    "sweep": [
 13      {
 14        "cfg": {
 15          "lr": 0.0015,
 16          "residual_scale": 0.0
 17        },
 18        "mean": 0.013166054850444198
 19      },
 20      {
 21        "cfg": {
 22          "lr": 0.0015,
 23          "residual_scale": 0.45
 24        },
 25        "mean": 0.014295766479335725
 26      },
 27      {
 28        "cfg": {
 29          "lr": 0.0015,
 30          "residual_scale": 0.9
 31        },
 32        "mean": 0.0171925718896091
 33      },
 34      {
 35        "cfg": {
 36          "lr": 0.003,
 37          "residual_scale": 0.0
 38        },
 39        "mean": 0.004933564574457705
 40      },
 41      {
 42        "cfg": {
 43          "lr": 0.003,
 44          "residual_scale": 0.45
 45        },
 46        "mean": 0.0055945796193555
 47      },
 48      {
 49        "cfg": {
 50          "lr": 0.003,
 51          "residual_scale": 0.9
 52        },
 53        "mean": 0.00661358906654641
 54      },
 55      {
 56        "cfg": {
 57          "lr": 0.006,
 58          "residual_scale": 0.0
 59        },
 60        "mean": 0.002542233618441969
 61      },
 62      {
 63        "cfg": {
 64          "lr": 0.006,
 65          "residual_scale": 0.45
 66        },
 67        "mean": 0.002950671303551644
 68      },
 69      {
 70        "cfg": {
 71          "lr": 0.006,
 72          "residual_scale": 0.9
 73        },
 74        "mean": 0.0036307289847172797
 75      }
 76    ],
 77    "full": {
 78      "mean": 0.0026320953911636025,
 79      "std": 0.0005285747590703118,
 80      "per_seed": [
 81        0.0022362906020134687,
 82        0.0019099744968116283,
 83        0.002377047436311841,
 84        0.0036456219386309385,
 85        0.0032889561261981726,
 86        0.002569406060501933,
 87        0.0024848924949765205,
 88        0.002544573973864317
 89      ],
 90      "n": 8
 91    }
 92  },
 93  "idea": {
 94    "mean": 0.0031172324815997854,
 95    "std": 0.0013738510956345731,
 96    "per_seed": [
 97      0.003699983935803175,
 98      0.002306041307747364,
 99      0.0014978268882259727,
100      0.004068186040967703,
101      0.0020008618012070656,
102      0.0021134966518729925,
103      0.0032527041621506214,
104      0.005998759064823389
105    ],
106    "n": 8
107  },
108  "comparison": {
109    "delta_mean": 0.0004851370904361829,
110    "idea_wins": 3,
111    "n_pairs": 8,
112    "per_seed_diffs": [
113      0.0014636933337897062,
114      0.0003960668109357357,
115      -0.0008792205480858684,
116      0.00042256410233676434,
117      -0.001288094324991107,
118      -0.0004559094086289406,
119      0.0007678116671741009,
120      0.003454185090959072
121    ],
122    "p_value": 0.43845,
123    "mde": 0.0012544680223890213,
124    "mde_rel_pct": 47.6604315558048,
125    "verdict": "no significant win",
126    "system_worked": false
127  },
128  "mechanism_signature": {
129    "prediction": "cycle sensitivity should be below individual branch sensitivity",
130    "observed_mean_branch_norms": [
131      1.3430568054318428,
132      1.3305060217777889
133    ],
134    "observed_cycle_norm_product": 1.786945167216707,
135    "confirmed": false
136  },
137  "idea_sweep": [
138    {
139      "cfg": {
140        "lr": 0.006,
141        "gain": 0.45
142      },
143      "mean_on_sweep_seeds": 0.0028930095431860536
144    },
145    {
146      "cfg": {
147        "lr": 0.0015,
148        "gain": 0.45
149      },
150      "mean_on_sweep_seeds": 0.01692552154418081
151    },
152    {
153      "cfg": {
154        "lr": 0.006,
155        "gain": 0.45
156      },
157      "mean_on_sweep_seeds": 0.0028930095431860536
158    }
159  ]
160}