Second-order fusion prior for point-set diffusion / bench_report.json
Failed on benchmark
1{
2 "bench_version": 1,
3 "track": "unordered_pointset_denoising",
4 "model": "mlp_tiny",
5 "metric_direction": "lower is better",
6 "n_seeds": 8,
7 "baseline": {
8 "best_cfg": {
9 "lr": 0.003,
10 "weight_decay": 0.0001
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "weight_decay": 0.0
17 },
18 "mean": 0.09564068168401718
19 },
20 {
21 "cfg": {
22 "lr": 0.001,
23 "weight_decay": 0.0001
24 },
25 "mean": 0.09560250677168369
26 },
27 {
28 "cfg": {
29 "lr": 0.002,
30 "weight_decay": 0.0
31 },
32 "mean": 0.084096509963274
33 },
34 {
35 "cfg": {
36 "lr": 0.002,
37 "weight_decay": 0.0001
38 },
39 "mean": 0.0841062506660819
40 },
41 {
42 "cfg": {
43 "lr": 0.003,
44 "weight_decay": 0.0
45 },
46 "mean": 0.07198812533169985
47 },
48 {
49 "cfg": {
50 "lr": 0.003,
51 "weight_decay": 0.0001
52 },
53 "mean": 0.07177905738353729
54 }
55 ],
56 "full": {
57 "mean": 0.07177905738353729,
58 "std": 0.0038503444468855823,
59 "per_seed": [
60 0.07469823211431503,
61 0.06881649047136307,
62 0.06867200881242752,
63 0.0654606819152832,
64 0.0715550109744072,
65 0.07852364331483841,
66 0.07229366153478622,
67 0.07421272993087769
68 ],
69 "n": 8,
70 "collision_per_seed": [
71 0.054999999701976776,
72 0.042500000447034836,
73 0.05374999716877937,
74 0.06499999761581421,
75 0.0637499988079071,
76 0.0625,
77 0.04625000059604645,
78 0.07750000059604645
79 ],
80 "collision_mean": 0.05828124936670065,
81 "mean_gap": 0.9999973326921463,
82 "cfg": {
83 "lr": 0.003,
84 "weight_decay": 0.0001
85 }
86 }
87 },
88 "idea": {
89 "mean": 0.08495018165558577,
90 "std": 0.0036923579732418346,
91 "per_seed": [
92 0.08852718770503998,
93 0.08314542472362518,
94 0.08323974907398224,
95 0.0792478397488594,
96 0.0855635404586792,
97 0.09225999563932419,
98 0.08457005023956299,
99 0.08304766565561295
100 ],
101 "n": 8,
102 "collision_per_seed": [
103 0.07499999552965164,
104 0.05624999850988388,
105 0.05624999850988388,
106 0.07374999672174454,
107 0.05999999865889549,
108 0.07000000029802322,
109 0.0625,
110 0.06875000149011612
111 ],
112 "collision_mean": 0.06531249871477485,
113 "mean_gap": 0.9999975040555,
114 "cfg": {
115 "lr": 0.003,
116 "weight_decay": 0.0001,
117 "lam": 0.1,
118 "beta": 2.0,
119 "delta": 0.08
120 }
121 },
122 "comparison": {
123 "delta_mean": 0.013171124272048473,
124 "idea_wins": 0,
125 "n_pairs": 8,
126 "per_seed_diffs": [
127 0.013828955590724945,
128 0.014328934252262115,
129 0.014567740261554718,
130 0.013787157833576202,
131 0.014008529484272003,
132 0.013736352324485779,
133 0.012276388704776764,
134 0.00883493572473526
135 ],
136 "p_value": 0.0081,
137 "verdict": "idea worse (significant)",
138 "system_worked": false
139 },
140 "custom_track": {
141 "name": "unordered_pointset_denoising",
142 "file": "pointset_fusion_track.py",
143 "domain": "point-set-diffusion"
144 },
145 "idea_settings": [
146 {
147 "mean": 0.08495018165558577,
148 "std": 0.0036923579732418346,
149 "per_seed": [
150 0.08852718770503998,
151 0.08314542472362518,
152 0.08323974907398224,
153 0.0792478397488594,
154 0.0855635404586792,
155 0.09225999563932419,
156 0.08457005023956299,
157 0.08304766565561295
158 ],
159 "n": 8,
160 "collision_per_seed": [
161 0.07499999552965164,
162 0.05624999850988388,
163 0.05624999850988388,
164 0.07374999672174454,
165 0.05999999865889549,
166 0.07000000029802322,
167 0.0625,
168 0.06875000149011612
169 ],
170 "collision_mean": 0.06531249871477485,
171 "mean_gap": 0.9999975040555,
172 "cfg": {
173 "lr": 0.003,
174 "weight_decay": 0.0001,
175 "lam": 0.1,
176 "beta": 2.0,
177 "delta": 0.08
178 }
179 },
180 {
181 "mean": 0.0936331320554018,
182 "std": 0.0037630771536154607,
183 "per_seed": [
184 0.09448400139808655,
185 0.09308362752199173,
186 0.08933884650468826,
187 0.0864451453089714,
188 0.09648385643959045,
189 0.09782309830188751,
190 0.09371592104434967,
191 0.09769055992364883
192 ],
193 "n": 8,
194 "collision_per_seed": [
195 0.07999999821186066,
196 0.07000000029802322,
197 0.06749999523162842,
198 0.07124999910593033,
199 0.0612499974668026,
200 0.07249999791383743,
201 0.0637499988079071,
202 0.06499999761581421
203 ],
204 "collision_mean": 0.0689062480814755,
205 "mean_gap": 0.9999975264072418,
206 "cfg": {
207 "lr": 0.002,
208 "weight_decay": 0.0001,
209 "lam": 0.1,
210 "beta": 2.0,
211 "delta": 0.08
212 }
213 },
214 {
215 "mean": 0.09946490358561277,
216 "std": 0.0037473580173809696,
217 "per_seed": [
218 0.09996496886014938,
219 0.10018042474985123,
220 0.09583928436040878,
221 0.09132290631532669,
222 0.1030997484922409,
223 0.10119754076004028,
224 0.1009417474269867,
225 0.10317260771989822
226 ],
227 "n": 8,
228 "collision_per_seed": [
229 0.07750000059604645,
230 0.08374999463558197,
231 0.05874999985098839,
232 0.06499999761581421,
233 0.06749999523162842,
234 0.07249999791383743,
235 0.08249999582767487,
236 0.07625000178813934
237 ],
238 "collision_mean": 0.07296874793246388,
239 "mean_gap": 0.9999975264072418,
240 "cfg": {
241 "lr": 0.001,
242 "weight_decay": 0.0001,
243 "lam": 0.1,
244 "beta": 2.0,
245 "delta": 0.08
246 }
247 }
248 ],
249 "mechanism_signature": {
250 "quantity": "normalized predicted short-gap collision rate",
251 "baseline_predicted_collision": 0.05828124936670065,
252 "idea_predicted_collision": 0.06531249871477485,
253 "predicted_delta": 0.007031249348074198,
254 "observed_test_mse_delta": 0.013171124272048473,
255 "confirmed": false
256 }
257}