Cut-Aware Augmentation Filtering / bench_report.json
Mechanism confirmed, baseline not beaten
1{
2 "bench_version": 1,
3 "track": "vision",
4 "model": "cnn_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": 0.8149999976158142
17 },
18 {
19 "cfg": {
20 "lr": 0.003
21 },
22 "mean": 0.7900000065565109
23 },
24 {
25 "cfg": {
26 "lr": 0.005
27 },
28 "mean": 0.8599999994039536
29 }
30 ],
31 "full": {
32 "mean": 0.8212500065565109,
33 "std": 0.04414110526102245,
34 "per_seed": [
35 0.8050000071525574,
36 0.7250000238418579,
37 0.8149999976158142,
38 0.8149999976158142,
39 0.8650000095367432,
40 0.875,
41 0.8550000190734863,
42 0.8149999976158142
43 ],
44 "n": 8
45 }
46 },
47 "idea": {
48 "mean": 0.8218750059604645,
49 "std": 0.042788540285862334,
50 "per_seed": [
51 0.8050000071525574,
52 0.7300000190734863,
53 0.8149999976158142,
54 0.8149999976158142,
55 0.8650000095367432,
56 0.875,
57 0.8550000190734863,
58 0.8149999976158142
59 ],
60 "n": 8
61 },
62 "comparison": {
63 "delta_mean": 0.0006249994039535522,
64 "idea_wins": 0,
65 "n_pairs": 8,
66 "per_seed_diffs": [
67 0.0,
68 0.004999995231628418,
69 0.0,
70 0.0,
71 0.0,
72 0.0,
73 0.0,
74 0.0
75 ],
76 "p_value": 1.0,
77 "mde": 0.001478123590350151,
78 "mde_rel_pct": 0.17998460621606582,
79 "verdict": "no measurable effect",
80 "system_worked": false
81 },
82 "math_check": {
83 "scale_error": 0.0,
84 "q_order": false
85 },
86 "idea_sweep": [
87 {
88 "cfg": {
89 "lr": 0.001
90 },
91 "result": {
92 "mean": 0.8312499970197678,
93 "std": 0.04655305473804935,
94 "per_seed": [
95 0.7850000262260437,
96 0.824999988079071,
97 0.8949999809265137,
98 0.7549999952316284,
99 0.8799999952316284,
100 0.8799999952316284,
101 0.8149999976158142,
102 0.8149999976158142
103 ],
104 "n": 8
105 }
106 },
107 {
108 "cfg": {
109 "lr": 0.003
110 },
111 "result": {
112 "mean": 0.8218750059604645,
113 "std": 0.042788540285862334,
114 "per_seed": [
115 0.8050000071525574,
116 0.7300000190734863,
117 0.8149999976158142,
118 0.8149999976158142,
119 0.8650000095367432,
120 0.875,
121 0.8550000190734863,
122 0.8149999976158142
123 ],
124 "n": 8
125 }
126 },
127 {
128 "cfg": {
129 "lr": 0.005
130 },
131 "result": {
132 "mean": 0.8475000038743019,
133 "std": 0.04847678961382153,
134 "per_seed": [
135 0.800000011920929,
136 0.8299999833106995,
137 0.9100000262260437,
138 0.8949999809265137,
139 0.8600000143051147,
140 0.7900000214576721,
141 0.7900000214576721,
142 0.9049999713897705
143 ],
144 "n": 8
145 }
146 }
147 ],
148 "structural_match": "CIFAR image augmentations and CNN prediction graph",
149 "mechanism_signature": {
150 "prediction": "higher estimated cut policy receives lower q",
151 "observed": [
152 {
153 "seed": 0,
154 "baseline": {
155 "metric": 0.8050000071525574,
156 "cuts": [
157 0.8347027159886934,
158 0.8347027159886934,
159 0.9102364030599416
160 ],
161 "q": [
162 0.4159778406556032,
163 0.4159778406556032,
164 0.16804431868879358
165 ],
166 "graph_reg": 0.016758395358920097
167 },
168 "idea": {
169 "metric": 0.8050000071525574,
170 "cuts": [
171 0.8347027159886934,
172 0.8347027159886934,
173 0.9102364030599416
174 ],
175 "q": [
176 0.4159778406556032,
177 0.4159778406556032,
178 0.16804431868879358
179 ],
180 "graph_reg": 0.014007863588631153
181 }
182 },
183 {
184 "seed": 1,
185 "baseline": {
186 "metric": 0.7250000238418579,
187 "cuts": [
188 0.8353256392943406,
189 0.8353256392943406,
190 0.908369077739702
191 ],
192 "q": [
193 0.41386816845062135,
194 0.41386816845062135,
195 0.17226366309875726
196 ],
197 "graph_reg": 0.005923776887357235
198 },
199 "idea": {
200 "metric": 0.7300000190734863,
201 "cuts": [
202 0.8353256392943406,
203 0.8353256392943406,
204 0.908369077739702
205 ],
206 "q": [
207 0.41386816845062135,
208 0.41386816845062135,
209 0.17226366309875726
210 ],
211 "graph_reg": 0.005063352175056934
212 }
213 }
214 ],
215 "confirmed": true
216 },
217 "custom_track": null
218}