Differentially Passive Neural Blocks / 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 },
11 "sweep": [
12 {
13 "cfg": {
14 "lr": 0.001
15 },
16 "mean": 0.02264283667318523
17 },
18 {
19 "cfg": {
20 "lr": 0.003
21 },
22 "mean": 0.006244247197173536
23 },
24 {
25 "cfg": {
26 "lr": 0.01
27 },
28 "mean": 0.00514151350944303
29 }
30 ],
31 "full": {
32 "mean": 0.004185360565315932,
33 "std": 0.00262482957347662,
34 "per_seed": [
35 0.0021960907615721226,
36 0.008501192554831505,
37 0.0017922910628840327,
38 0.008076479658484459,
39 0.0030377130024135113,
40 0.0011907307198271155,
41 0.0035283213946968317,
42 0.005160065367817879
43 ],
44 "n": 8
45 },
46 "union_grid": [
47 {
48 "lr": 0.001
49 },
50 {
51 "lr": 0.003
52 },
53 {
54 "lr": 0.01
55 }
56 ]
57 },
58 "idea": {
59 "mean": 0.002211095365055371,
60 "std": 0.001581263928471004,
61 "per_seed": [
62 0.0014746395172551274,
63 0.002074356423690915,
64 0.000778969086240977,
65 0.0025305193848907948,
66 0.002626171102747321,
67 0.0014305157819762826,
68 0.006008145399391651,
69 0.0007654462242498994
70 ],
71 "n": 8
72 },
73 "comparison": {
74 "delta_mean": -0.001974265200260561,
75 "idea_wins": 6,
76 "n_pairs": 8,
77 "per_seed_diffs": [
78 -0.0007214512443169951,
79 -0.00642683613114059,
80 -0.0010133219766430557,
81 -0.005545960273593664,
82 -0.00041154189966619015,
83 0.00023978506214916706,
84 0.0024798240046948195,
85 -0.004394619143567979
86 ],
87 "p_value": 0.14225,
88 "mde": 0.0026083344497749737,
89 "mde_rel_pct": 62.32042398904009,
90 "verdict": "no significant win",
91 "system_worked": false
92 },
93 "idea_sweep": [
94 {
95 "cfg": {
96 "lr": 0.001
97 },
98 "result": {
99 "mean": 0.06869563926011324,
100 "std": 0.06865088808304035,
101 "per_seed": [
102 0.0694878026843071,
103 0.22950862348079681,
104 0.026874704286456108,
105 0.011694572865962982,
106 0.01833035983145237,
107 0.047722410410642624,
108 0.11673897504806519,
109 0.029207665473222733
110 ],
111 "n": 8
112 }
113 },
114 {
115 "cfg": {
116 "lr": 0.003
117 },
118 "result": {
119 "mean": 0.013020415848586708,
120 "std": 0.007406346915347002,
121 "per_seed": [
122 0.017435848712921143,
123 0.02717023342847824,
124 0.010481353849172592,
125 0.005967922043055296,
126 0.004026471171528101,
127 0.011689518578350544,
128 0.019995147362351418,
129 0.007396831642836332
130 ],
131 "n": 8
132 }
133 },
134 {
135 "cfg": {
136 "lr": 0.01
137 },
138 "result": {
139 "mean": 0.002211095365055371,
140 "std": 0.001581263928471004,
141 "per_seed": [
142 0.0014746395172551274,
143 0.002074356423690915,
144 0.000778969086240977,
145 0.0025305193848907948,
146 0.002626171102747321,
147 0.0014305157819762826,
148 0.006008145399391651,
149 0.0007654462242498994
150 ],
151 "n": 8
152 }
153 }
154 ],
155 "structural_match": "Dynamics track directly tests recurrent trajectory sensitivity and long-horizon stability.",
156 "mechanism_signature": {
157 "predicted_ratio_bound": 0.9591663046625439,
158 "observed_mean_ratio": NaN,
159 "predicted_lmi_bound": 0.0,
160 "observed_max_lmi": NaN,
161 "confirmed": false
162 },
163 "selected_idea_cfg": {
164 "lr": 0.01
165 }
166}