Passivity-Regularized Sequence Layer / bench_report.json

Mechanism confirmed, baseline not beaten

Raw ⬇ ZIP
 1{
 2  "bench_version": 1,
 3  "track": "dynamics",
 4  "model": "rnn_small",
 5  "metric_direction": "lower is better",
 6  "n_seeds": 8,
 7  "baseline": {
 8    "best_cfg": {
 9      "lr": 0.01
10    },
11    "sweep": [
12      {
13        "cfg": {
14          "lr": 0.001
15        },
16        "mean": 0.007516150246374309
17      },
18      {
19        "cfg": {
20          "lr": 0.003
21        },
22        "mean": 0.004790205741301179
23      },
24      {
25        "cfg": {
26          "lr": 0.01
27        },
28        "mean": 0.0030207290838006884
29      }
30    ],
31    "full": {
32      "mean": 0.002415244802250527,
33      "std": 0.0011322891072914308,
34      "per_seed": [
35        0.004476976580917835,
36        0.002256670966744423,
37        0.0014001723611727357,
38        0.00394909642636776,
39        0.002509008627384901,
40        0.0020131270866841078,
41        0.0016695521771907806,
42        0.0010473541915416718
43      ],
44      "n": 8
45    },
46    "selected_cfg": {
47      "lr": 0.01
48    }
49  },
50  "idea": {
51    "mean": 0.002238515968201682,
52    "std": 0.0010780951878567476,
53    "per_seed": [
54      0.0014792155707255006,
55      0.0021166419610381126,
56      0.001531714922748506,
57      0.0042628017254173756,
58      0.001253232709132135,
59      0.0011274017160758376,
60      0.0025555682368576527,
61      0.0035815509036183357
62    ],
63    "n": 8
64  },
65  "comparison": {
66    "delta_mean": -0.00017672883404884487,
67    "idea_wins": 4,
68    "n_pairs": 8,
69    "per_seed_diffs": [
70      -0.0029977610101923347,
71      -0.00014002900570631027,
72      0.00013154256157577038,
73      0.00031370529904961586,
74      -0.0012557759182527661,
75      -0.0008857253706082702,
76      0.000886016059666872,
77      0.002534196712076664
78    ],
79    "p_value": 0.72575,
80    "mde": 0.0013571432898663574,
81    "mde_rel_pct": 56.19071361220901,
82    "verdict": "no significant win",
83    "system_worked": false
84  },
85  "selected_idea_cfg": {
86    "lr": 0.01,
87    "lam": 0.001,
88    "paired": true
89  },
90  "mechanism_signature": {
91    "baseline_cumulative_residual": -15.145476698875427,
92    "idea_cumulative_residual": -15.241462588310242,
93    "baseline_amplification": 0.1478598192334175,
94    "idea_amplification": 0.14565712958574295,
95    "predicted_direction": "passivity reduces residual and amplification",
96    "confirmed": true
97  }
98}