Følner-Gated Message Passing / 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 "weight_decay": 0.0001
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "weight_decay": 0.0
17 },
18 "mean": 0.001921573159052059
19 },
20 {
21 "cfg": {
22 "lr": 0.001,
23 "weight_decay": 0.0001
24 },
25 "mean": 0.0018964317569043487
26 },
27 {
28 "cfg": {
29 "lr": 0.003,
30 "weight_decay": 0.0
31 },
32 "mean": 0.0011179699067724869
33 },
34 {
35 "cfg": {
36 "lr": 0.003,
37 "weight_decay": 0.0001
38 },
39 "mean": 0.0010905505041591823
40 },
41 {
42 "cfg": {
43 "lr": 0.01,
44 "weight_decay": 0.0
45 },
46 "mean": 0.0006299250526353717
47 },
48 {
49 "cfg": {
50 "lr": 0.01,
51 "weight_decay": 0.0001
52 },
53 "mean": 0.0005868283187737688
54 }
55 ],
56 "full": {
57 "mean": 0.0005577069205173757,
58 "std": 0.00016235484867295895,
59 "per_seed": [
60 0.00046681062667630613,
61 0.0004913305747322738,
62 0.0004835708241444081,
63 0.0009056012495420873,
64 0.0005242845509201288,
65 0.0003316242655273527,
66 0.0006971154361963272,
67 0.0005613178364001215
68 ],
69 "n": 8
70 }
71 },
72 "idea": {
73 "mean": 0.0005104744304844644,
74 "std": 0.00016259431597973005,
75 "per_seed": [
76 0.0005603914032690227,
77 0.0004172833578195423,
78 0.00043428907520137727,
79 0.0008996041724458337,
80 0.00048453747876919806,
81 0.00045123344170860946,
82 0.000313605327391997,
83 0.0005228511872701347
84 ],
85 "n": 8,
86 "chosen_cfg": {
87 "lr": 0.01,
88 "weight_decay": 0.0001,
89 "delta": 0.5,
90 "beta": 0.7,
91 "slope": 8.0
92 },
93 "sweep": [
94 {
95 "cfg": {
96 "lr": 0.001,
97 "weight_decay": 0.0001,
98 "delta": 0.25,
99 "beta": 0.7,
100 "slope": 8.0
101 },
102 "mean": 0.001631094521144405
103 },
104 {
105 "cfg": {
106 "lr": 0.001,
107 "weight_decay": 0.0001,
108 "delta": 0.35,
109 "beta": 0.7,
110 "slope": 8.0
111 },
112 "mean": 0.0015781476977281272
113 },
114 {
115 "cfg": {
116 "lr": 0.001,
117 "weight_decay": 0.0001,
118 "delta": 0.5,
119 "beta": 0.7,
120 "slope": 8.0
121 },
122 "mean": 0.0017034276097547263
123 },
124 {
125 "cfg": {
126 "lr": 0.003,
127 "weight_decay": 0.0001,
128 "delta": 0.25,
129 "beta": 0.7,
130 "slope": 8.0
131 },
132 "mean": 0.0012407805188558996
133 },
134 {
135 "cfg": {
136 "lr": 0.003,
137 "weight_decay": 0.0001,
138 "delta": 0.35,
139 "beta": 0.7,
140 "slope": 8.0
141 },
142 "mean": 0.0012196671304991469
143 },
144 {
145 "cfg": {
146 "lr": 0.003,
147 "weight_decay": 0.0001,
148 "delta": 0.5,
149 "beta": 0.7,
150 "slope": 8.0
151 },
152 "mean": 0.0011900454992428422
153 },
154 {
155 "cfg": {
156 "lr": 0.01,
157 "weight_decay": 0.0001,
158 "delta": 0.25,
159 "beta": 0.7,
160 "slope": 8.0
161 },
162 "mean": 0.0006524444761453196
163 },
164 {
165 "cfg": {
166 "lr": 0.01,
167 "weight_decay": 0.0001,
168 "delta": 0.35,
169 "beta": 0.7,
170 "slope": 8.0
171 },
172 "mean": 0.0006070969975553453
173 },
174 {
175 "cfg": {
176 "lr": 0.01,
177 "weight_decay": 0.0001,
178 "delta": 0.5,
179 "beta": 0.7,
180 "slope": 8.0
181 },
182 "mean": 0.000577892002183944
183 }
184 ]
185 },
186 "comparison": {
187 "delta_mean": -4.72324900329113e-05,
188 "idea_wins": 6,
189 "n_pairs": 8,
190 "per_seed_diffs": [
191 9.358077659271657e-05,
192 -7.404721691273153e-05,
193 -4.9281748943030834e-05,
194 -5.9970770962536335e-06,
195 -3.974707215093076e-05,
196 0.00011960917618125677,
197 -0.00038351010880433023,
198 -3.846664912998676e-05
199 ],
200 "p_value": 0.5225,
201 "mde": 0.00012771297744744513,
202 "mde_rel_pct": 22.89965800118956,
203 "verdict": "no significant win",
204 "system_worked": false
205 },
206 "mechanism_signature": {
207 "prediction": "expansive temporal receptive fields should yield r>1+delta and gate<0.5",
208 "observed_mean_ratio": 1.284388935357143,
209 "observed_mean_gate": 0.8395185023546219,
210 "predicted_threshold": 1.35,
211 "confirmed": false
212 },
213 "protocol_note": "Baseline sweep and idea sweep use shared lr union; 8 paired seeds; dynamics is structurally matched to stability/control."
214}