Firmly Nonexpansive Convex-Gradient Denoiser / bench_report.json
Failed on benchmark
1{
2 "bench_version": 1,
3 "track": "dynamics",
4 "model": "custom_matched_refinement_rnn",
5 "metric_direction": "lower is better",
6 "n_seeds": 8,
7 "baseline": {
8 "best_cfg": {
9 "lr": 0.003,
10 "step": 0.25
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "step": 0.25
17 },
18 "mean": 0.00020656592459999956
19 },
20 {
21 "cfg": {
22 "lr": 0.001,
23 "step": 0.5
24 },
25 "mean": 0.00035531837966118474
26 },
27 {
28 "cfg": {
29 "lr": 0.001,
30 "step": 1.0
31 },
32 "mean": 0.0009092524851439521
33 },
34 {
35 "cfg": {
36 "lr": 0.001,
37 "step": 2.0
38 },
39 "mean": 0.008616556180641055
40 },
41 {
42 "cfg": {
43 "lr": 0.003,
44 "step": 0.25
45 },
46 "mean": 0.00010479371485416777
47 },
48 {
49 "cfg": {
50 "lr": 0.003,
51 "step": 0.5
52 },
53 "mean": 0.0001718404637358617
54 },
55 {
56 "cfg": {
57 "lr": 0.003,
58 "step": 1.0
59 },
60 "mean": 0.0007685718519496731
61 },
62 {
63 "cfg": {
64 "lr": 0.003,
65 "step": 2.0
66 },
67 "mean": 0.010642364271916449
68 },
69 {
70 "cfg": {
71 "lr": 0.01,
72 "step": 0.25
73 },
74 "mean": 0.00014638340144301765
75 },
76 {
77 "cfg": {
78 "lr": 0.01,
79 "step": 0.5
80 },
81 "mean": 0.0006321593828033656
82 },
83 {
84 "cfg": {
85 "lr": 0.01,
86 "step": 1.0
87 },
88 "mean": 0.0017989037442021072
89 },
90 {
91 "cfg": {
92 "lr": 0.01,
93 "step": 2.0
94 },
95 "mean": 0.035143551183864474
96 }
97 ],
98 "full": {
99 "mean": 0.00010781512173707597,
100 "std": 2.4695774008571815e-05,
101 "per_seed": [
102 6.690840382361785e-05,
103 0.00013107550330460072,
104 0.0001228735054610297,
105 9.831744682742283e-05,
106 0.0001292095403186977,
107 0.00013566877169068903,
108 0.00010378748993389308,
109 7.468031253665686e-05
110 ],
111 "n": 8
112 }
113 },
114 "idea": {
115 "mean": 0.00011978308180005115,
116 "std": 0.00014333295352723913,
117 "per_seed": [
118 4.686686224886216e-05,
119 9.109569509746507e-05,
120 0.00011464366252766922,
121 8.399749276577495e-06,
122 7.34728091629222e-05,
123 6.693405157420784e-05,
124 0.000491157581564039,
125 6.569424294866621e-05
126 ],
127 "n": 8,
128 "cfg": {
129 "lr": 0.003,
130 "step": 0.25
131 },
132 "sweep": [
133 {
134 "cfg": {
135 "lr": 0.003,
136 "step": 0.25
137 },
138 "mean": 0.00011978308180005115,
139 "per_seed": [
140 4.686686224886216e-05,
141 9.109569509746507e-05,
142 0.00011464366252766922,
143 8.399749276577495e-06,
144 7.34728091629222e-05,
145 6.693405157420784e-05,
146 0.000491157581564039,
147 6.569424294866621e-05
148 ]
149 },
150 {
151 "cfg": {
152 "lr": 0.003,
153 "step": 0.5
154 },
155 "mean": 0.00017802219872464775,
156 "per_seed": [
157 5.6299137213500217e-05,
158 0.0002794665051624179,
159 9.379914990859106e-05,
160 0.0001387570664519444,
161 0.0003630504652392119,
162 0.00018886031466536224,
163 0.00017176337132696062,
164 0.00013218157982919365
165 ]
166 }
167 ]
168 },
169 "comparison": {
170 "delta_mean": 1.196796006297518e-05,
171 "idea_wins": 7,
172 "n_pairs": 8,
173 "per_seed_diffs": [
174 -2.0041541574755684e-05,
175 -3.997980820713565e-05,
176 -8.229842933360487e-06,
177 -8.991769755084533e-05,
178 -5.573673115577549e-05,
179 -6.873472011648118e-05,
180 0.0003873700916301459,
181 -8.986069587990642e-06
182 ],
183 "p_value": 1.0,
184 "mde": 0.0001291404226093749,
185 "mde_rel_pct": 119.77950822548252,
186 "verdict": "no significant win",
187 "system_worked": false
188 },
189 "mechanism_signature": {
190 "track_choice": "dynamics matches stability/control structure",
191 "idea_sweep": [
192 {
193 "cfg": {
194 "lr": 0.003,
195 "step": 0.25
196 },
197 "mean": 0.00011978308180005115,
198 "per_seed": [
199 4.686686224886216e-05,
200 9.109569509746507e-05,
201 0.00011464366252766922,
202 8.399749276577495e-06,
203 7.34728091629222e-05,
204 6.693405157420784e-05,
205 0.000491157581564039,
206 6.569424294866621e-05
207 ]
208 },
209 {
210 "cfg": {
211 "lr": 0.003,
212 "step": 0.5
213 },
214 "mean": 0.00017802219872464775,
215 "per_seed": [
216 5.6299137213500217e-05,
217 0.0002794665051624179,
218 9.379914990859106e-05,
219 0.0001387570664519444,
220 0.0003630504652392119,
221 0.00018886031466536224,
222 0.00017176337132696062,
223 0.00013218157982919365
224 ]
225 }
226 ],
227 "mechanism_signature": {
228 "prediction": "convex-gradient refinement should have non-amplifying local response",
229 "idea_observed_mean_output_sensitivity": 0.21240035444498062,
230 "baseline_observed_mean_output_sensitivity": 0.2104792781174183,
231 "idea_observed_max_output_sensitivity": 0.2549852430820465,
232 "baseline_observed_max_output_sensitivity": 0.25559741258621216,
233 "idea_mean_firm_gap_proxy": 2.065331489120581e-05,
234 "baseline_mean_firm_gap_proxy": 2.03489205432561e-05,
235 "confirmed": false
236 }
237 },
238 "baseline_union_grid": [
239 {
240 "lr": 0.001,
241 "step": 0.25
242 },
243 {
244 "lr": 0.001,
245 "step": 0.5
246 },
247 {
248 "lr": 0.001,
249 "step": 1.0
250 },
251 {
252 "lr": 0.001,
253 "step": 2.0
254 },
255 {
256 "lr": 0.003,
257 "step": 0.25
258 },
259 {
260 "lr": 0.003,
261 "step": 0.5
262 },
263 {
264 "lr": 0.003,
265 "step": 1.0
266 },
267 {
268 "lr": 0.003,
269 "step": 2.0
270 },
271 {
272 "lr": 0.01,
273 "step": 0.25
274 },
275 {
276 "lr": 0.01,
277 "step": 0.5
278 },
279 {
280 "lr": 0.01,
281 "step": 1.0
282 },
283 {
284 "lr": 0.01,
285 "step": 2.0
286 }
287 ],
288 "protocol_notes": {
289 "epochs": 12,
290 "batch": 64,
291 "n_train": 400,
292 "n_test": 200,
293 "paired_seeds": [
294 0,
295 1,
296 2,
297 3,
298 4,
299 5,
300 6,
301 7
302 ],
303 "baseline_sweep_seeds": [
304 0,
305 1,
306 2,
307 3
308 ]
309 }
310}