KPZ latent evolution block / bench_report.json
Beats tuned baseline
1{
2 "bench_version": 1,
3 "track": "custom_tracks/kpz_field_forecast",
4 "model": "kpz_cell_shared",
5 "metric_direction": "lower is better",
6 "n_seeds": 8,
7 "baseline": {
8 "best_cfg": {
9 "lr": 0.006,
10 "lam": 0.0
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "lam": 0.0
17 },
18 "mean": 0.001372527563944459
19 },
20 {
21 "cfg": {
22 "lr": 0.003,
23 "lam": 0.0
24 },
25 "mean": 0.0012717426288872957
26 },
27 {
28 "cfg": {
29 "lr": 0.006,
30 "lam": 0.0
31 },
32 "mean": 0.0011299708858132362
33 }
34 ],
35 "full": {
36 "mean": 0.001126505361753516,
37 "std": 1.5728558641491704e-05,
38 "per_seed": [
39 0.0011347912950441241,
40 0.0011231173994019628,
41 0.0011269556125625968,
42 0.0011350192362442613,
43 0.0010915156453847885,
44 0.0011320485500618815,
45 0.001149294781498611,
46 0.0011193003738299012
47 ],
48 "n": 8
49 }
50 },
51 "idea": {
52 "mean": 0.000540067398105748,
53 "std": 2.4417121177500956e-06,
54 "per_seed": [
55 0.0005460529937408864,
56 0.0005373440217226744,
57 0.00053932867012918,
58 0.0005404861294664443,
59 0.0005400783265940845,
60 0.0005387364653870463,
61 0.0005386892007663846,
62 0.0005398233770392835
63 ],
64 "n": 8,
65 "cfg": {
66 "lr": 0.006,
67 "lam": 2.4
68 },
69 "sweep": [
70 {
71 "cfg": {
72 "lr": 0.001,
73 "lam": 1.2
74 },
75 "mean": 0.0007497890837839805
76 },
77 {
78 "cfg": {
79 "lr": 0.003,
80 "lam": 1.8
81 },
82 "mean": 0.0005883654812350869
83 },
84 {
85 "cfg": {
86 "lr": 0.006,
87 "lam": 2.4
88 },
89 "mean": 0.000540067398105748
90 }
91 ],
92 "baseline_best_lr": 0.006
93 },
94 "comparison": {
95 "delta_mean": -0.0005864379636477679,
96 "idea_wins": 8,
97 "n_pairs": 8,
98 "per_seed_diffs": [
99 -0.0005887383013032377,
100 -0.0005857733776792884,
101 -0.0005876269424334168,
102 -0.000594533106777817,
103 -0.000551437318790704,
104 -0.0005933120846748352,
105 -0.0006106055807322264,
106 -0.0005794769967906177
107 ],
108 "p_value": 0.0081,
109 "mde": 1.4053872303936461e-05,
110 "mde_rel_pct": 1.2475637294846278,
111 "verdict": "idea better (significant)",
112 "system_worked": true
113 },
114 "mechanism_signature": {
115 "prediction": "mean increment slope = dt*lambda/2 vs mean |gradient|^2",
116 "rows": [
117 {
118 "system": "baseline",
119 "observed_slope": 0.0002200861940903433,
120 "predicted_slope": 0.0,
121 "relative_error": 220.08619409034333,
122 "test_mse": 0.0011347912950441241
123 },
124 {
125 "system": "idea",
126 "observed_slope": 0.031170539640519964,
127 "predicted_slope": 0.03,
128 "relative_error": 0.03901668746108347,
129 "test_mse": 0.0005460529937408864
130 }
131 ],
132 "confirmed": true
133 },
134 "custom_track": {
135 "name": "kpz_field_forecast",
136 "file": "kpz_track.py",
137 "domain": "pde"
138 },
139 "protocol": {
140 "paired_seeds": [
141 0,
142 1,
143 2,
144 3,
145 4,
146 5,
147 6,
148 7
149 ],
150 "baseline_sweep_seeds": [
151 0,
152 1,
153 2,
154 3
155 ],
156 "baseline_grid": [
157 {
158 "lr": 0.001,
159 "lam": 0.0
160 },
161 {
162 "lr": 0.003,
163 "lam": 0.0
164 },
165 {
166 "lr": 0.006,
167 "lam": 0.0
168 }
169 ],
170 "idea_grid": [
171 {
172 "lr": 0.001,
173 "lam": 1.2
174 },
175 {
176 "lr": 0.003,
177 "lam": 1.8
178 },
179 {
180 "lr": 0.006,
181 "lam": 2.4
182 }
183 ],
184 "epochs": 15,
185 "n_train": 400,
186 "n_test": 120
187 }
188}