Feasibility-Preserving Error Compensator / 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": 0.01,
10 "epochs": 12
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "epochs": 12
17 },
18 "mean": 0.005020207085181028
19 },
20 {
21 "cfg": {
22 "lr": 0.003,
23 "epochs": 12
24 },
25 "mean": 0.0029489120497601107
26 },
27 {
28 "cfg": {
29 "lr": 0.01,
30 "epochs": 12
31 },
32 "mean": 0.0014198302378645167
33 }
34 ],
35 "full": {
36 "mean": 0.0014198302378645167,
37 "std": 0.00039839114901047844,
38 "per_seed": [
39 0.0009327387670055032,
40 0.0018641706556081772,
41 0.0010785290505737066,
42 0.0020708823576569557,
43 0.0009260809747502208,
44 0.0014055331703275442,
45 0.0016443432541564107,
46 0.001436363672837615
47 ],
48 "n": 8
49 }
50 },
51 "idea": {
52 "per_seed": [
53 0.4626535177230835,
54 0.45927271246910095,
55 0.5099033713340759,
56 0.5794274210929871,
57 0.4526548683643341,
58 0.45540836453437805,
59 0.4523857831954956,
60 0.4395509958267212
61 ],
62 "mean": 0.47640712931752205
63 },
64 "comparison": {
65 "delta_mean": 0.47498729907965753,
66 "idea_wins": 0,
67 "n_pairs": 8,
68 "per_seed_diffs": [
69 0.461720778956078,
70 0.4574085418134928,
71 0.5088248422835022,
72 0.5773565387353301,
73 0.4517287873895839,
74 0.4540028313640505,
75 0.4507414399413392,
76 0.4381146321538836
77 ],
78 "p_value": 0.0081,
79 "mde": 0.03875567283474413,
80 "mde_rel_pct": 2729.599060591516,
81 "verdict": "idea worse (significant)",
82 "system_worked": false
83 },
84 "mechanism_signature": {
85 "math_and_trained_behavior": {
86 "spectral_radius_predicted": 0.8822698000045112,
87 "decay_factor_observed": 0.8815778323329195,
88 "decay_relative_error": 0.00078430393014483,
89 "ordinary_max_compensator": 0.708185009625288,
90 "antiwindup_max_compensator": 0.1603214553812824,
91 "windup_reduction_ratio": 0.22638357661102013,
92 "constraint_violation": 0.0,
93 "prediction_confirmed": true,
94 "trained_nominal_abs_mean": 0.6836502552032471,
95 "trained_shaped_abs_mean": 0.21633578836917877,
96 "trained_residual_abs_mean": 0.2636767625808716,
97 "trained_compensator_abs_mean": 0.22096218168735504,
98 "confirmed": true
99 },
100 "idea_sweep": [
101 {
102 "cfg": {
103 "lr": 0.01,
104 "epochs": 12,
105 "ki": 0.05,
106 "ka": 0.7
107 },
108 "mean": 0.47640712931752205,
109 "std": 0.04352936722428276,
110 "per_seed": [
111 0.4626535177230835,
112 0.45927271246910095,
113 0.5099033713340759,
114 0.5794274210929871,
115 0.4526548683643341,
116 0.45540836453437805,
117 0.4523857831954956,
118 0.4395509958267212
119 ],
120 "n": 8
121 },
122 {
123 "cfg": {
124 "lr": 0.01,
125 "epochs": 12,
126 "ki": 0.075,
127 "ka": 0.9
128 },
129 "mean": 0.4816158339381218,
130 "std": 0.04396653306668558,
131 "per_seed": [
132 0.46778514981269836,
133 0.46430763602256775,
134 0.5159854292869568,
135 0.5854425430297852,
136 0.45754313468933105,
137 0.46042028069496155,
138 0.45683005452156067,
139 0.44461244344711304
140 ],
141 "n": 8
142 },
143 {
144 "cfg": {
145 "lr": 0.01,
146 "epochs": 12,
147 "ki": 0.1,
148 "ka": 1.1
149 },
150 "mean": 0.48671987280249596,
151 "std": 0.044320348769445794,
152 "per_seed": [
153 0.4729180335998535,
154 0.46918609738349915,
155 0.5218073129653931,
156 0.5911740660667419,
157 0.46244245767593384,
158 0.4652194082736969,
159 0.4616015553474426,
160 0.4494100511074066
161 ],
162 "n": 8
163 }
164 ]
165 }
166}