Centered-Geometry Projection Loss / bench_report.json
Failed on benchmark
1{
2 "bench_version": 1,
3 "track": "tabular",
4 "model": "mlp",
5 "metric_direction": "lower is better",
6 "n_seeds": 8,
7 "baseline": {
8 "best_cfg": {
9 "lr": 0.01
10 },
11 "sweep": [
12 {
13 "cfg": {
14 "lr": 0.001
15 },
16 "mean": 17.006587505340576
17 },
18 {
19 "cfg": {
20 "lr": 0.003
21 },
22 "mean": 11.705099105834961
23 },
24 {
25 "cfg": {
26 "lr": 0.01
27 },
28 "mean": 7.426158666610718
29 }
30 ],
31 "full": {
32 "mean": 7.655009210109711,
33 "std": 0.7549298959102074,
34 "per_seed": [
35 7.125377178192139,
36 6.509033203125,
37 7.528863430023193,
38 8.541360855102539,
39 7.81464958190918,
40 6.914392948150635,
41 8.888213157653809,
42 7.918183326721191
43 ],
44 "n": 8
45 }
46 },
47 "idea": {
48 "mean": 7.631358027458191,
49 "std": 0.8269721813703816,
50 "per_seed": [
51 7.135181903839111,
52 6.160409450531006,
53 7.647706031799316,
54 8.540335655212402,
55 7.844282150268555,
56 6.913455486297607,
57 8.89346694946289,
58 7.916026592254639
59 ],
60 "n": 8
61 },
62 "comparison": {
63 "delta_mean": -0.023651182651519775,
64 "idea_wins": 4,
65 "n_pairs": 8,
66 "per_seed_diffs": [
67 0.009804725646972656,
68 -0.34862375259399414,
69 0.11884260177612305,
70 -0.0010251998901367188,
71 0.029632568359375,
72 -0.0009374618530273438,
73 0.005253791809082031,
74 -0.0021567344665527344
75 ],
76 "p_value": 0.9445,
77 "mde": 0.1149067142918632,
78 "mde_rel_pct": 1.5010656569832703,
79 "verdict": "no measurable effect",
80 "system_worked": false
81 },
82 "track_reason": "tabular is the registered optimizer/regularizer track; both systems share the same MLP and learned bottleneck",
83 "idea_sweep": [
84 {
85 "cfg": {
86 "lr": 0.001,
87 "lambda": 0.03
88 },
89 "mean_first4": 17.134891271591187
90 },
91 {
92 "cfg": {
93 "lr": 0.001,
94 "lambda": 0.1
95 },
96 "mean_first4": 17.241438150405884
97 },
98 {
99 "cfg": {
100 "lr": 0.001,
101 "lambda": 0.3
102 },
103 "mean_first4": 17.401116609573364
104 },
105 {
106 "cfg": {
107 "lr": 0.003,
108 "lambda": 0.03
109 },
110 "mean_first4": 11.699760913848877
111 },
112 {
113 "cfg": {
114 "lr": 0.003,
115 "lambda": 0.1
116 },
117 "mean_first4": 11.717785120010376
118 },
119 {
120 "cfg": {
121 "lr": 0.003,
122 "lambda": 0.3
123 },
124 "mean_first4": 11.66649079322815
125 },
126 {
127 "cfg": {
128 "lr": 0.01,
129 "lambda": 0.03
130 },
131 "mean_first4": 7.370908260345459
132 },
133 {
134 "cfg": {
135 "lr": 0.01,
136 "lambda": 0.1
137 },
138 "mean_first4": 7.376788854598999
139 },
140 {
141 "cfg": {
142 "lr": 0.01,
143 "lambda": 0.3
144 },
145 "mean_first4": 7.378063797950745
146 }
147 ],
148 "selected_idea_cfg": {
149 "lr": 0.01,
150 "lambda": 0.03
151 },
152 "mechanism_signature": {
153 "predicted_centered_distance_correlation": "no universal value for learned encoder",
154 "observed_centered_distance_correlation": 0.9999756813049316,
155 "observed_centered_variance_ratio": 98.05452728271484,
156 "predicted_variance_ceiling": 1.0,
157 "confirmed": false
158 },
159 "budget": {
160 "epochs": 24,
161 "n_train": 400,
162 "n_test": 200,
163 "batch": 128,
164 "hidden_dim": 32,
165 "bottleneck_dim": 8,
166 "baseline_grid": [
167 0.001,
168 0.003,
169 0.01
170 ],
171 "idea_lambdas": [
172 0.03,
173 0.1,
174 0.3
175 ],
176 "seeds": [
177 0,
178 1,
179 2,
180 3,
181 4,
182 5,
183 6,
184 7
185 ]
186 }
187}