Hidden-Diffusion Irreversibility Monitor / bench_report.json
Failed on benchmark
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": 0.006,
10 "weight_decay": 0.0001
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "weight_decay": 0.0
17 },
18 "mean": 0.17645552940666676
19 },
20 {
21 "cfg": {
22 "lr": 0.001,
23 "weight_decay": 0.0001
24 },
25 "mean": 0.17659775167703629
26 },
27 {
28 "cfg": {
29 "lr": 0.003,
30 "weight_decay": 0.0
31 },
32 "mean": 0.028942419216036797
33 },
34 {
35 "cfg": {
36 "lr": 0.003,
37 "weight_decay": 0.0001
38 },
39 "mean": 0.028870720881968737
40 },
41 {
42 "cfg": {
43 "lr": 0.006,
44 "weight_decay": 0.0
45 },
46 "mean": 0.013678546762093902
47 },
48 {
49 "cfg": {
50 "lr": 0.006,
51 "weight_decay": 0.0001
52 },
53 "mean": 0.013676861766725779
54 }
55 ],
56 "full": {
57 "mean": 0.013490461744368076,
58 "std": 0.0036455379141356633,
59 "per_seed": [
60 0.01656705141067505,
61 0.01134894136339426,
62 0.013825791887938976,
63 0.012965662404894829,
64 0.02068343572318554,
65 0.014116058126091957,
66 0.010448829270899296,
67 0.007967923767864704
68 ],
69 "n": 8
70 }
71 },
72 "idea": {
73 "mean": 0.013573068077675998,
74 "std": 0.0036332788661090037,
75 "per_seed": [
76 0.016753576695919037,
77 0.011368772946298122,
78 0.013876390643417835,
79 0.012907849624752998,
80 0.020729539915919304,
81 0.014167544431984425,
82 0.010720721445977688,
83 0.008060148917138577
84 ],
85 "n": 8,
86 "cfg": {
87 "lr": 0.006,
88 "weight_decay": 0.0,
89 "lam": 0.001
90 },
91 "sweep": [
92 {
93 "cfg": {
94 "lr": 0.001,
95 "weight_decay": 0.0,
96 "lam": 0.001
97 },
98 "mean": 0.16410381672903895,
99 "per_seed": [
100 0.08069120347499847,
101 0.20044010877609253,
102 0.1486465185880661,
103 0.2762965261936188,
104 0.060551416128873825,
105 0.1775653064250946,
106 0.17477571964263916,
107 0.19386373460292816
108 ]
109 },
110 {
111 "cfg": {
112 "lr": 0.003,
113 "weight_decay": 0.0,
114 "lam": 0.001
115 },
116 "mean": 0.02484322735108435,
117 "per_seed": [
118 0.03566475957632065,
119 0.02596493996679783,
120 0.02446393296122551,
121 0.02981584146618843,
122 0.026317916810512543,
123 0.01899126172065735,
124 0.01727476716041565,
125 0.020252399146556854
126 ]
127 },
128 {
129 "cfg": {
130 "lr": 0.006,
131 "weight_decay": 0.0,
132 "lam": 0.001
133 },
134 "mean": 0.013573068077675998,
135 "per_seed": [
136 0.016753576695919037,
137 0.011368772946298122,
138 0.013876390643417835,
139 0.012907849624752998,
140 0.020729539915919304,
141 0.014167544431984425,
142 0.010720721445977688,
143 0.008060148917138577
144 ]
145 }
146 ]
147 },
148 "comparison": {
149 "delta_mean": 8.260633330792189e-05,
150 "idea_wins": 1,
151 "n_pairs": 8,
152 "per_seed_diffs": [
153 0.00018652528524398804,
154 1.9831582903862e-05,
155 5.059875547885895e-05,
156 -5.7812780141830444e-05,
157 4.610419273376465e-05,
158 5.14863058924675e-05,
159 0.00027189217507839203,
160 9.222514927387238e-05
161 ],
162 "p_value": 0.04575,
163 "mde": 8.579928619368267e-05,
164 "mde_rel_pct": 0.635999625657748,
165 "verdict": "idea worse (significant)",
166 "system_worked": false
167 },
168 "protocol_note": "Baseline and idea use identical rnn_small systems; only latent diffusion penalty differs.",
169 "idea_grid": [
170 {
171 "cfg": {
172 "lr": 0.001,
173 "weight_decay": 0.0,
174 "lam": 0.001
175 },
176 "mean": 0.16410381672903895,
177 "per_seed": [
178 0.08069120347499847,
179 0.20044010877609253,
180 0.1486465185880661,
181 0.2762965261936188,
182 0.060551416128873825,
183 0.1775653064250946,
184 0.17477571964263916,
185 0.19386373460292816
186 ]
187 },
188 {
189 "cfg": {
190 "lr": 0.003,
191 "weight_decay": 0.0,
192 "lam": 0.001
193 },
194 "mean": 0.02484322735108435,
195 "per_seed": [
196 0.03566475957632065,
197 0.02596493996679783,
198 0.02446393296122551,
199 0.02981584146618843,
200 0.026317916810512543,
201 0.01899126172065735,
202 0.01727476716041565,
203 0.020252399146556854
204 ]
205 },
206 {
207 "cfg": {
208 "lr": 0.006,
209 "weight_decay": 0.0,
210 "lam": 0.001
211 },
212 "mean": 0.013573068077675998,
213 "per_seed": [
214 0.016753576695919037,
215 0.011368772946298122,
216 0.013876390643417835,
217 0.012907849624752998,
218 0.020729539915919304,
219 0.014167544431984425,
220 0.010720721445977688,
221 0.008060148917138577
222 ]
223 }
224 ],
225 "mechanism_signature": {
226 "predicted": "hidden cross diffusion can vary while observed output power is comparatively stable",
227 "baseline_observed_output_power_std": 0.15846109952602266,
228 "idea_observed_output_power_std": 0.15854744148375174,
229 "baseline_hidden_dxy_std": 0.00042474186371430176,
230 "idea_hidden_dxy_std": 2.490041131097002e-05,
231 "idea_mean_reverse_forward_gap": 0.018812500281407442,
232 "confirmed": false
233 }
234}