Nonlinear Noise-Tightening Drift / 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": 0.003
10 },
11 "sweep": [
12 {
13 "cfg": {
14 "lr": 0.001
15 },
16 "mean": 1.5209165837859473e-07
17 },
18 {
19 "cfg": {
20 "lr": 0.003
21 },
22 "mean": 7.913097554990145e-08
23 },
24 {
25 "cfg": {
26 "lr": 0.01
27 },
28 "mean": 8.186839473012242e-08
29 }
30 ],
31 "full": {
32 "mean": 7.923793088338016e-08,
33 "std": 7.46882980049606e-09,
34 "per_seed": [
35 7.157926518175373e-08,
36 7.663354750775397e-08,
37 7.570882587515371e-08,
38 9.260226363494439e-08,
39 7.609644114836556e-08,
40 7.839068416615191e-08,
41 9.065027484211896e-08,
42 7.224214471079904e-08
43 ],
44 "n": 8
45 }
46 },
47 "idea": {
48 "per_seed": [
49 1.0914992998323214e-07,
50 1.3094300754801225e-07,
51 1.2386591663471336e-07,
52 1.4415415705570922e-07,
53 1.1842626435054626e-07,
54 1.1481838413374135e-07,
55 1.2749221411922917e-07,
56 1.0909479897236452e-07
57 ],
58 "mean": 1.2224308409969353e-07,
59 "config": {
60 "beta": 0.05,
61 "lr": 0.01
62 }
63 },
64 "comparison": {
65 "delta_mean": 4.3005153216313374e-08,
66 "idea_wins": 0,
67 "n_pairs": 8,
68 "per_seed_diffs": [
69 3.7570664801478415e-08,
70 5.430946004025827e-08,
71 4.815709075955965e-08,
72 5.155189342076483e-08,
73 4.2329823202180705e-08,
74 3.642769996758943e-08,
75 3.6841939277110214e-08,
76 3.6852654261565476e-08
77 ],
78 "p_value": 0.0081,
79 "mde": 6.131983352471457e-09,
80 "mde_rel_pct": 7.738696965088996,
81 "verdict": "idea worse (significant)",
82 "system_worked": false
83 },
84 "mechanism_signature": {
85 "quantity": "effective theta restoring rate from trained one-step map",
86 "predicted_baseline_lower_bound": 0.981,
87 "observed_mean_rate": 0.03263185149990022,
88 "observed_median_rate": 0.030955076217651367,
89 "observed_10th_percentile": 0.02489030361175537,
90 "observed_fraction_above_lambda": 0.0,
91 "beta": 0.05,
92 "confirmed": false
93 },
94 "idea_sweep": [
95 {
96 "config": {
97 "beta": 0.05,
98 "lr": 0.001
99 },
100 "mean": 2.14023083699999e-07,
101 "per_seed": [
102 2.0379644638524042e-07,
103 1.9002254703082144e-07,
104 2.5476279574832006e-07,
105 2.034364285918855e-07,
106 1.6803994640213205e-07,
107 2.7017307502319454e-07,
108 2.43161309754214e-07,
109 1.7879212066418404e-07
110 ]
111 },
112 {
113 "config": {
114 "beta": 0.05,
115 "lr": 0.003
116 },
117 "mean": 1.2451453113015987e-07,
118 "per_seed": [
119 1.0692676966073122e-07,
120 1.221927590222549e-07,
121 1.1927961907076678e-07,
122 1.5280032528153242e-07,
123 1.207399833447198e-07,
124 1.1392528875830976e-07,
125 1.4401997816548828e-07,
126 1.1623152573747575e-07
127 ]
128 },
129 {
130 "config": {
131 "beta": 0.05,
132 "lr": 0.01
133 },
134 "mean": 1.2224308409969353e-07,
135 "per_seed": [
136 1.0914992998323214e-07,
137 1.3094300754801225e-07,
138 1.2386591663471336e-07,
139 1.4415415705570922e-07,
140 1.1842626435054626e-07,
141 1.1481838413374135e-07,
142 1.2749221411922917e-07,
143 1.0909479897236452e-07
144 ]
145 },
146 {
147 "config": {
148 "beta": 0.15,
149 "lr": 0.001
150 },
151 "mean": 4.032821507848894e-07,
152 "per_seed": [
153 3.502042602576694e-07,
154 3.693708094942849e-07,
155 4.762017624670989e-07,
156 4.272998808119155e-07,
157 3.5077033544439473e-07,
158 4.4331363824312575e-07,
159 4.6180804247342166e-07,
160 3.4728847708720423e-07
161 ]
162 },
163 {
164 "config": {
165 "beta": 0.15,
166 "lr": 0.003
167 },
168 "mean": 2.739564326503796e-07,
169 "per_seed": [
170 2.2414128864056693e-07,
171 2.758638970590255e-07,
172 2.6184392254435807e-07,
173 3.4700917694863165e-07,
174 2.725748231569014e-07,
175 2.2845142666483298e-07,
176 3.138804345326207e-07,
177 2.678864916560997e-07
178 ]
179 },
180 {
181 "config": {
182 "beta": 0.15,
183 "lr": 0.01
184 },
185 "mean": 2.634781353094695e-07,
186 "per_seed": [
187 2.325318462226278e-07,
188 2.7763206844610977e-07,
189 2.6512879003348644e-07,
190 3.1356714202956937e-07,
191 2.6467418479114713e-07,
192 2.3143238081502204e-07,
193 2.7973300120720523e-07,
194 2.431256689305883e-07
195 ]
196 },
197 {
198 "config": {
199 "beta": 0.3,
200 "lr": 0.001
201 },
202 "mean": 8.732421434842763e-07,
203 "per_seed": [
204 7.096875833667582e-07,
205 8.606693882029504e-07,
206 9.6827398010646e-07,
207 9.847033197729615e-07,
208 8.193193821171008e-07,
209 8.333087180290022e-07,
210 1.0446385658724466e-06,
211 7.653362104065309e-07
212 ]
213 },
214 {
215 "config": {
216 "beta": 0.3,
217 "lr": 0.003
218 },
219 "mean": 6.4876725502927e-07,
220 "per_seed": [
221 5.215468377173238e-07,
222 6.694369290016766e-07,
223 6.181929279591714e-07,
224 8.246532274824858e-07,
225 6.660657163592987e-07,
226 5.123378628013597e-07,
227 7.337577585531108e-07,
228 6.441467803597334e-07
229 ]
230 },
231 {
232 "config": {
233 "beta": 0.3,
234 "lr": 0.01
235 },
236 "mean": 6.068020255156625e-07,
237 "per_seed": [
238 5.414903512246383e-07,
239 6.367587843669753e-07,
240 6.005964792166196e-07,
241 7.084224193931732e-07,
242 6.152868081699125e-07,
243 5.057927410234697e-07,
244 6.651686703662563e-07,
245 5.808999503642553e-07
246 ]
247 }
248 ],
249 "protocol_notes": {
250 "n_train": 1200,
251 "n_test": 400,
252 "epochs": 12,
253 "baseline_lr_union": [
254 0.001,
255 0.003,
256 0.01
257 ],
258 "structural_match": "controlled pendulum rollout; nonlinear restoring force is applied in the recurrent state update",
259 "baseline_best_lr": 0.003
260 }
261}