Input-Aware Contracting Neural ODE / 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 "rho": 0.0
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "rho": 0.0
17 },
18 "mean": 0.5870515555143356
19 },
20 {
21 "cfg": {
22 "lr": 0.003,
23 "rho": 0.0
24 },
25 "mean": 0.19268612749874592
26 },
27 {
28 "cfg": {
29 "lr": 0.006,
30 "rho": 0.0
31 },
32 "mean": 0.0888435747474432
33 }
34 ],
35 "full": {
36 "mean": 0.08503814833238721,
37 "std": 0.01852758683212337,
38 "per_seed": [
39 0.08398815989494324,
40 0.09531869739294052,
41 0.0957954153418541,
42 0.08027202636003494,
43 0.06067569926381111,
44 0.11256278306245804,
45 0.09799148142337799,
46 0.053700923919677734
47 ],
48 "n": 8
49 },
50 "union_grid": [
51 {
52 "lr": 0.001,
53 "rho": 0.0
54 },
55 {
56 "lr": 0.003,
57 "rho": 0.0
58 },
59 {
60 "lr": 0.006,
61 "rho": 0.0
62 }
63 ]
64 },
65 "idea": {
66 "mean": 0.08122727181762457,
67 "std": 0.021878604412979827,
68 "per_seed": [
69 0.06868152320384979,
70 0.08553639054298401,
71 0.10347606986761093,
72 0.057179175317287445,
73 0.059706274420022964,
74 0.11696434020996094,
75 0.0995219275355339,
76 0.05875247344374657
77 ],
78 "n": 8,
79 "cfg": {
80 "lr": 0.006,
81 "rho": 0.05
82 },
83 "details": [
84 {
85 "metric": 0.06868152320384979,
86 "mu_mean": 0.7426044940948486,
87 "mu_max": 1.4340018033981323,
88 "violation_fraction": 1.0,
89 "metric_rate_effect": 0.17669588327407837,
90 "metric_min_eig": 0.47089532017707825
91 },
92 {
93 "metric": 0.08553639054298401,
94 "mu_mean": 0.8353224992752075,
95 "mu_max": 1.6954039335250854,
96 "violation_fraction": 1.0,
97 "metric_rate_effect": 0.2537629008293152,
98 "metric_min_eig": 0.6923100352287292
99 },
100 {
101 "metric": 0.10347606986761093,
102 "mu_mean": 0.8890057802200317,
103 "mu_max": 1.7825677394866943,
104 "violation_fraction": 1.0,
105 "metric_rate_effect": 0.17414918541908264,
106 "metric_min_eig": 0.4535291790962219
107 },
108 {
109 "metric": 0.057179175317287445,
110 "mu_mean": 0.6833391785621643,
111 "mu_max": 2.0641016960144043,
112 "violation_fraction": 0.9453125,
113 "metric_rate_effect": 0.3635365962982178,
114 "metric_min_eig": 0.4815308451652527
115 },
116 {
117 "metric": 0.059706274420022964,
118 "mu_mean": 0.7647638320922852,
119 "mu_max": 1.741713523864746,
120 "violation_fraction": 0.9375,
121 "metric_rate_effect": 0.2081649750471115,
122 "metric_min_eig": 0.5761778354644775
123 },
124 {
125 "metric": 0.11696434020996094,
126 "mu_mean": 0.6851489543914795,
127 "mu_max": 0.8102754354476929,
128 "violation_fraction": 1.0,
129 "metric_rate_effect": 0.06493743509054184,
130 "metric_min_eig": 0.6551467776298523
131 },
132 {
133 "metric": 0.0995219275355339,
134 "mu_mean": 0.9402241110801697,
135 "mu_max": 1.841058373451233,
136 "violation_fraction": 0.98828125,
137 "metric_rate_effect": 0.31226134300231934,
138 "metric_min_eig": 0.7394733428955078
139 },
140 {
141 "metric": 0.05875247344374657,
142 "mu_mean": 0.7863765358924866,
143 "mu_max": 2.171452760696411,
144 "violation_fraction": 0.77734375,
145 "metric_rate_effect": 0.44044044613838196,
146 "metric_min_eig": 0.47293055057525635
147 }
148 ],
149 "sweep_summary": [
150 {
151 "cfg": {
152 "lr": 0.001,
153 "rho": 0.05
154 },
155 "mean": 0.5566666722297668
156 },
157 {
158 "cfg": {
159 "lr": 0.003,
160 "rho": 0.05
161 },
162 "mean": 0.1854265844449401
163 },
164 {
165 "cfg": {
166 "lr": 0.006,
167 "rho": 0.05
168 },
169 "mean": 0.08122727181762457
170 }
171 ]
172 },
173 "comparison": {
174 "delta_mean": -0.00381087651476264,
175 "idea_wins": 4,
176 "n_pairs": 8,
177 "per_seed_diffs": [
178 -0.015306636691093445,
179 -0.009782306849956512,
180 0.007680654525756836,
181 -0.023092851042747498,
182 -0.000969424843788147,
183 0.004401557147502899,
184 0.0015304461121559143,
185 0.005051549524068832
186 ],
187 "p_value": 0.3924,
188 "mde": 0.00923697491590414,
189 "mde_rel_pct": 10.862154335486855,
190 "verdict": "no significant win",
191 "system_worked": false
192 },
193 "mechanism_signature": {
194 "math_check": {
195 "frozen_mu": -1.2000000000000002,
196 "total_mu": 0.2999999999999998,
197 "predicted_rate_slope": 1.5,
198 "observed_rate_slope": 1.5,
199 "confirmed": true
200 },
201 "trained_model": {
202 "mu_mean": 0.7908481732010841,
203 "baseline_mu_mean": "not_measured",
204 "rate_effect": 0.24924359563738108,
205 "positive_definite_min_eig": 0.4535291790962219
206 },
207 "confirmed": false,
208 "note": "The total-derivative contribution was observed in trained adaptive models, but no quantitative NN-scale predicted-vs-observed rate law was established; analytical scalar verification is reported separately and is not used as confirmation."
209 },
210 "math_check": {
211 "frozen_mu": -1.2000000000000002,
212 "total_mu": 0.2999999999999998,
213 "predicted_rate_slope": 1.5,
214 "observed_rate_slope": 1.5,
215 "confirmed": true
216 }
217}