Diversity-Weighted Leave-One-Out Policy Baseline / bench_report.json
Mechanism confirmed, baseline not beaten
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 "epochs": 6,
11 "p": 0.0
12 },
13 "sweep": [
14 {
15 "cfg": {
16 "lr": 0.001,
17 "epochs": 6,
18 "p": 0.0
19 },
20 "mean": 0.1710126642137766
21 },
22 {
23 "cfg": {
24 "lr": 0.003,
25 "epochs": 6,
26 "p": 0.0
27 },
28 "mean": 0.024177383398637176
29 },
30 {
31 "cfg": {
32 "lr": 0.01,
33 "epochs": 6,
34 "p": 0.0
35 },
36 "mean": 0.012901934329420328
37 }
38 ],
39 "full": {
40 "mean": 0.014406590606085956,
41 "std": 0.005962791282569369,
42 "per_seed": [
43 0.010699832811951637,
44 0.012548977509140968,
45 0.009432518854737282,
46 0.018926408141851425,
47 0.02494051493704319,
48 0.01899670995771885,
49 0.004924832843244076,
50 0.014782929793000221
51 ],
52 "n": 8
53 }
54 },
55 "idea": {
56 "mean": 0.014913259132299572,
57 "std": 0.006629419112933174,
58 "per_seed": [
59 0.008315389044582844,
60 0.016753537580370903,
61 0.00829843245446682,
62 0.019920941442251205,
63 0.02598617970943451,
64 0.019167941063642502,
65 0.005189504008740187,
66 0.015674147754907608
67 ],
68 "n": 8
69 },
70 "comparison": {
71 "delta_mean": 0.0005066685262136161,
72 "idea_wins": 2,
73 "n_pairs": 8,
74 "per_seed_diffs": [
75 -0.0023844437673687935,
76 0.004204560071229935,
77 -0.001134086400270462,
78 0.0009945333003997803,
79 0.0010456647723913193,
80 0.00017123110592365265,
81 0.0002646711654961109,
82 0.0008912179619073868
83 ],
84 "p_value": 0.5343,
85 "mde": 0.0015981726750217166,
86 "mde_rel_pct": 11.09334414171928,
87 "verdict": "no significant win",
88 "system_worked": false
89 },
90 "mechanism_signature": {
91 "custom_track": null,
92 "signature": {
93 "p": 1.0,
94 "predicted_weight_ratio": 16.0,
95 "observed_weight_ratio": 16.0,
96 "relative_error": 0.0,
97 "confirmed": true
98 },
99 "idea_grid": [
100 {
101 "lr": 0.01,
102 "epochs": 6,
103 "p": 1.0
104 },
105 {
106 "lr": 0.001,
107 "epochs": 6,
108 "p": 1.0
109 },
110 {
111 "lr": 0.01,
112 "epochs": 6,
113 "p": 1.0
114 }
115 ],
116 "baseline_union_lrs": [
117 0.001,
118 0.003,
119 0.01
120 ]
121 }
122}