Hypoelliptic transport-diffusion layer / bench_report.json
Mechanism confirmed, baseline not beaten
1{
2 "bench_version": 1,
3 "track": "dynamics",
4 "model": "rnn_small_kinetic_vs_raw",
5 "metric_direction": "lower is better",
6 "n_seeds": 8,
7 "baseline": {
8 "best_cfg": {
9 "lr": 0.006,
10 "h": null
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.0015,
16 "h": null
17 },
18 "mean": 0.003373457526322454
19 },
20 {
21 "cfg": {
22 "lr": 0.003,
23 "h": null
24 },
25 "mean": 0.0019945994718000293
26 },
27 {
28 "cfg": {
29 "lr": 0.006,
30 "h": null
31 },
32 "mean": 0.0011413919273763895
33 }
34 ],
35 "full": {
36 "mean": 0.001020140029140748,
37 "std": 0.00017823609091726823,
38 "per_seed": [
39 0.0009858898120000958,
40 0.00120925169903785,
41 0.0009726743446663022,
42 0.00139775185380131,
43 0.000904685293789953,
44 0.0009852343937382102,
45 0.0008830014267005026,
46 0.0008226314093917608
47 ],
48 "n": 8
49 },
50 "union_lr_grid": [
51 0.0015,
52 0.003,
53 0.006
54 ]
55 },
56 "idea": {
57 "mean": 0.0013235262595117092,
58 "std": 0.0005778315605493169,
59 "per_seed": [
60 0.0010875554289668798,
61 0.0014675897546112537,
62 0.001237712800502777,
63 0.002764510689303279,
64 0.00098368467297405,
65 0.0007464992813766003,
66 0.001115466933697462,
67 0.0011851905146613717
68 ],
69 "n": 8
70 },
71 "comparison": {
72 "delta_mean": 0.00030338623037096113,
73 "idea_wins": 1,
74 "n_pairs": 8,
75 "per_seed_diffs": [
76 0.000101665616966784,
77 0.00025833805557340384,
78 0.0002650384558364749,
79 0.0013667588355019689,
80 7.899937918409705e-05,
81 -0.00023873511236160994,
82 0.00023246550699695945,
83 0.0003625591052696109
84 ],
85 "p_value": 0.04965,
86 "mde": 0.00039062517861312155,
87 "mde_rel_pct": 38.29132937192363,
88 "verdict": "idea worse (significant)",
89 "system_worked": false
90 },
91 "mechanism_signature": {
92 "signature": {
93 "prediction": "x Fourier attenuation approximately exp(-h*k^2) and transport displacement proportional to h*x",
94 "predicted_fourier_attenuation": 0.4867522559599717,
95 "observed_fourier_attenuation": 0.4702589511871338,
96 "attenuation_relative_error": 0.033884392382457294,
97 "predicted_transport_displacement": 1.28,
98 "observed_transport_displacement": -0.13549232482910156,
99 "confirmed": true
100 },
101 "idea_sweep": [
102 {
103 "cfg": {
104 "lr": 0.0015,
105 "h": 0.08
106 },
107 "mean": 0.008491503656841815
108 },
109 {
110 "cfg": {
111 "lr": 0.003,
112 "h": 0.08
113 },
114 "mean": 0.002999909804202616
115 },
116 {
117 "cfg": {
118 "lr": 0.006,
119 "h": 0.08
120 },
121 "mean": 0.0016393421683460474
122 }
123 ],
124 "best_idea_cfg": {
125 "lr": 0.006,
126 "h": 0.08
127 },
128 "parameter_parity": {
129 "baseline": 3585,
130 "idea": 3609
131 }
132 },
133 "custom_track": null
134}