Event-triggered phase desynchronisation for recurrent hidden states / bench_report.json
Unverified
1{
2 "bench_version": 1,
3 "track": "dynamics",
4 "model": "rnn_small_phase_gru",
5 "metric_direction": "lower is better",
6 "n_seeds": 8,
7 "baseline": {
8 "best_cfg": {
9 "lr": 0.006,
10 "weight_decay": 0.0
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "weight_decay": 0.0
17 },
18 "mean": 0.002156580056180246,
19 "std": 0.0005337133383512335,
20 "per_seed": [
21 0.0014283749042078853,
22 0.0019986501429229975,
23 0.0017361249774694443,
24 0.0022385285701602697,
25 0.0016417494043707848,
26 0.002327029127627611,
27 0.0030278568156063557,
28 0.002854326507076621
29 ]
30 },
31 {
32 "cfg": {
33 "lr": 0.003,
34 "weight_decay": 0.0
35 },
36 "mean": 0.0011022633989341557,
37 "std": 0.00039374711633484005,
38 "per_seed": [
39 0.0009436382679268718,
40 0.0008971625356934965,
41 0.0006927313515916467,
42 0.001675509149208665,
43 0.0007120227674022317,
44 0.0008096328820101917,
45 0.0013870575930923223,
46 0.00170035264454782
47 ]
48 },
49 {
50 "cfg": {
51 "lr": 0.006,
52 "weight_decay": 0.0
53 },
54 "mean": 0.0007801908141118474,
55 "std": 0.000154764005468824,
56 "per_seed": [
57 0.0007548112771473825,
58 0.0008833256433717906,
59 0.0005147535703144968,
60 0.0009292386239394546,
61 0.0005692947888746858,
62 0.0008575883111916482,
63 0.0007604749407619238,
64 0.0009720393572933972
65 ]
66 }
67 ],
68 "full": {
69 "mean": 0.0007801908141118474,
70 "std": 0.000154764005468824,
71 "per_seed": [
72 0.0007548112771473825,
73 0.0008833256433717906,
74 0.0005147535703144968,
75 0.0009292386239394546,
76 0.0005692947888746858,
77 0.0008575883111916482,
78 0.0007604749407619238,
79 0.0009720393572933972
80 ],
81 "n": 8,
82 "signatures": [
83 {},
84 {},
85 {},
86 {},
87 {},
88 {},
89 {},
90 {}
91 ]
92 }
93 },
94 "idea": {
95 "best_cfg": {
96 "lr": 0.006,
97 "weight_decay": 0.0,
98 "k": 1.0,
99 "delta": 0.12
100 },
101 "sweep": [
102 {
103 "cfg": {
104 "lr": 0.006,
105 "weight_decay": 0.0,
106 "k": 1.0,
107 "delta": 0.02
108 },
109 "mean": 0.000786285956564825,
110 "std": 0.00016373570475228794
111 },
112 {
113 "cfg": {
114 "lr": 0.006,
115 "weight_decay": 0.0,
116 "k": 1.0,
117 "delta": 0.05
118 },
119 "mean": 0.0007858229582780041,
120 "std": 0.0001619668831625854
121 },
122 {
123 "cfg": {
124 "lr": 0.006,
125 "weight_decay": 0.0,
126 "k": 1.0,
127 "delta": 0.12
128 },
129 "mean": 0.0007801908141118474,
130 "std": 0.000154764005468824
131 }
132 ],
133 "full": {
134 "cfg": {
135 "lr": 0.006,
136 "weight_decay": 0.0,
137 "k": 1.0,
138 "delta": 0.12
139 },
140 "mean": 0.0007801908141118474,
141 "std": 0.000154764005468824,
142 "per_seed": [
143 0.0007548112771473825,
144 0.0008833256433717906,
145 0.0005147535703144968,
146 0.0009292386239394546,
147 0.0005692947888746858,
148 0.0008575883111916482,
149 0.0007604749407619238,
150 0.0009720393572933972
151 ],
152 "n": 8,
153 "signatures": [
154 {
155 "mean_V": 0.014306977391242981,
156 "mean_exact_u_norm": 0.026661492884159088,
157 "mean_applied_u_norm": 0.0,
158 "mean_hold_error": 0.026661492884159088,
159 "event_rate_per_sequence_step": 0.0,
160 "max_hold_error": 0.073698490858078
161 },
162 {
163 "mean_V": 0.04791915416717529,
164 "mean_exact_u_norm": 0.051865410059690475,
165 "mean_applied_u_norm": 0.0,
166 "mean_hold_error": 0.051865410059690475,
167 "event_rate_per_sequence_step": 0.0,
168 "max_hold_error": 0.10102833807468414
169 },
170 {
171 "mean_V": 0.019514720886945724,
172 "mean_exact_u_norm": 0.032075561583042145,
173 "mean_applied_u_norm": 0.0,
174 "mean_hold_error": 0.032075561583042145,
175 "event_rate_per_sequence_step": 0.0,
176 "max_hold_error": 0.08094639331102371
177 },
178 {
179 "mean_V": 0.01886623166501522,
180 "mean_exact_u_norm": 0.030544035136699677,
181 "mean_applied_u_norm": 0.0,
182 "mean_hold_error": 0.030544035136699677,
183 "event_rate_per_sequence_step": 0.0,
184 "max_hold_error": 0.08747458457946777
185 },
186 {
187 "mean_V": 0.021111171692609787,
188 "mean_exact_u_norm": 0.03352213650941849,
189 "mean_applied_u_norm": 0.0,
190 "mean_hold_error": 0.03352213650941849,
191 "event_rate_per_sequence_step": 0.0,
192 "max_hold_error": 0.07995948940515518
193 },
194 {
195 "mean_V": 0.026953421533107758,
196 "mean_exact_u_norm": 0.038697995245456696,
197 "mean_applied_u_norm": 0.0,
198 "mean_hold_error": 0.038697995245456696,
199 "event_rate_per_sequence_step": 0.0,
200 "max_hold_error": 0.07766351848840714
201 },
202 {
203 "mean_V": 0.013531758449971676,
204 "mean_exact_u_norm": 0.027224672958254814,
205 "mean_applied_u_norm": 0.0,
206 "mean_hold_error": 0.027224672958254814,
207 "event_rate_per_sequence_step": 0.0,
208 "max_hold_error": 0.07655420154333115
209 },
210 {
211 "mean_V": 0.0649743378162384,
212 "mean_exact_u_norm": 0.060073141008615494,
213 "mean_applied_u_norm": 0.0,
214 "mean_hold_error": 0.060073141008615494,
215 "event_rate_per_sequence_step": 0.0,
216 "max_hold_error": 0.09835705906152725
217 }
218 ]
219 }
220 },
221 "comparison": {
222 "delta_mean": 0.0,
223 "idea_wins": 0,
224 "n_pairs": 8,
225 "per_seed_diffs": [
226 0.0,
227 0.0,
228 0.0,
229 0.0,
230 0.0,
231 0.0,
232 0.0,
233 0.0
234 ],
235 "p_value": 1.0,
236 "mde": 0.0,
237 "mde_rel_pct": 0.0,
238 "verdict": "no measurable effect",
239 "system_worked": false,
240 "paired_diffs": [
241 0.0,
242 0.0,
243 0.0,
244 0.0,
245 0.0,
246 0.0,
247 0.0,
248 0.0
249 ],
250 "permutation_pvalue": 1.0
251 },
252 "mechanism_signature": {
253 "prediction": "event rotation should preserve pair norms and bound hold error by delta",
254 "observed": [
255 {
256 "mean_V": 0.014306977391242981,
257 "mean_exact_u_norm": 0.026661492884159088,
258 "mean_applied_u_norm": 0.0,
259 "mean_hold_error": 0.026661492884159088,
260 "event_rate_per_sequence_step": 0.0,
261 "max_hold_error": 0.073698490858078
262 },
263 {
264 "mean_V": 0.04791915416717529,
265 "mean_exact_u_norm": 0.051865410059690475,
266 "mean_applied_u_norm": 0.0,
267 "mean_hold_error": 0.051865410059690475,
268 "event_rate_per_sequence_step": 0.0,
269 "max_hold_error": 0.10102833807468414
270 },
271 {
272 "mean_V": 0.019514720886945724,
273 "mean_exact_u_norm": 0.032075561583042145,
274 "mean_applied_u_norm": 0.0,
275 "mean_hold_error": 0.032075561583042145,
276 "event_rate_per_sequence_step": 0.0,
277 "max_hold_error": 0.08094639331102371
278 },
279 {
280 "mean_V": 0.01886623166501522,
281 "mean_exact_u_norm": 0.030544035136699677,
282 "mean_applied_u_norm": 0.0,
283 "mean_hold_error": 0.030544035136699677,
284 "event_rate_per_sequence_step": 0.0,
285 "max_hold_error": 0.08747458457946777
286 },
287 {
288 "mean_V": 0.021111171692609787,
289 "mean_exact_u_norm": 0.03352213650941849,
290 "mean_applied_u_norm": 0.0,
291 "mean_hold_error": 0.03352213650941849,
292 "event_rate_per_sequence_step": 0.0,
293 "max_hold_error": 0.07995948940515518
294 },
295 {
296 "mean_V": 0.026953421533107758,
297 "mean_exact_u_norm": 0.038697995245456696,
298 "mean_applied_u_norm": 0.0,
299 "mean_hold_error": 0.038697995245456696,
300 "event_rate_per_sequence_step": 0.0,
301 "max_hold_error": 0.07766351848840714
302 },
303 {
304 "mean_V": 0.013531758449971676,
305 "mean_exact_u_norm": 0.027224672958254814,
306 "mean_applied_u_norm": 0.0,
307 "mean_hold_error": 0.027224672958254814,
308 "event_rate_per_sequence_step": 0.0,
309 "max_hold_error": 0.07655420154333115
310 },
311 {
312 "mean_V": 0.0649743378162384,
313 "mean_exact_u_norm": 0.060073141008615494,
314 "mean_applied_u_norm": 0.0,
315 "mean_hold_error": 0.060073141008615494,
316 "event_rate_per_sequence_step": 0.0,
317 "max_hold_error": 0.09835705906152725
318 }
319 ],
320 "mean_hold_error": 0.03758305567316711,
321 "delta": 0.12,
322 "confirmed": true
323 },
324 "math_check": {
325 "Vdot_observed": -0.006759319770566208,
326 "Vdot_predicted": -0.0067593166955923385,
327 "ratio": 1.0000004549237753,
328 "passed": true
329 },
330 "custom_track": null
331}