Mean-Reverting Levy-Jump Optimizer / bench_report.json
Mechanism confirmed, baseline not beaten
1{
2 "bench_version": 1,
3 "track": "tabular",
4 "model": "mlp_tiny",
5 "metric_direction": "lower is better",
6 "n_seeds": 8,
7 "baseline": {
8 "best_cfg": {
9 "lr": 0.01,
10 "wd": 0.0
11 },
12 "sweep": [
13 {
14 "cfg": {
15 "lr": 0.001,
16 "wd": 0.0
17 },
18 "mean": 14.000884532928467
19 },
20 {
21 "cfg": {
22 "lr": 0.003,
23 "wd": 0.0
24 },
25 "mean": 7.579635977745056
26 },
27 {
28 "cfg": {
29 "lr": 0.01,
30 "wd": 0.0
31 },
32 "mean": 6.920138120651245
33 }
34 ],
35 "full": {
36 "mean": 6.9205562472343445,
37 "std": 0.5716103719461259,
38 "per_seed": [
39 7.096155643463135,
40 6.974558353424072,
41 6.666646480560303,
42 6.943192005157471,
43 5.7007622718811035,
44 7.199983596801758,
45 6.8888444900512695,
46 7.8943071365356445
47 ],
48 "n": 8
49 }
50 },
51 "idea": {
52 "mean": 6.810217976570129,
53 "std": 0.34114415830274575,
54 "per_seed": [
55 6.926016807556152,
56 6.900168418884277,
57 6.7102837562561035,
58 6.759965896606445,
59 6.147101879119873,
60 6.569408893585205,
61 7.09395170211792,
62 7.374846458435059
63 ],
64 "n": 8
65 },
66 "comparison": {
67 "delta_mean": -0.11033827066421509,
68 "idea_wins": 5,
69 "n_pairs": 8,
70 "per_seed_diffs": [
71 -0.17013883590698242,
72 -0.07438993453979492,
73 0.04363727569580078,
74 -0.1832261085510254,
75 0.44633960723876953,
76 -0.6305747032165527,
77 0.2051072120666504,
78 -0.5194606781005859
79 ],
80 "p_value": 0.39255,
81 "mde": 0.29668947868390905,
82 "mde_rel_pct": 4.2870756061331745,
83 "verdict": "no significant win",
84 "system_worked": false
85 },
86 "mechanism_signature": {
87 "signature_type": "trained_nn_optimizer_dynamics",
88 "alpha": 1.5,
89 "lambda": 1.0,
90 "lr": 0.01,
91 "q": 0.01,
92 "predicted_drift_multiplier": 0.99,
93 "observed_median_drift_multiplier": 0.9900000095367432,
94 "observed_nn_jump_cf_power": 1.6607667261616632,
95 "jump_mad": 0.0009690423030406237,
96 "confirmed": true,
97 "n_observations": 5914800,
98 "training_metric_for_signature": {
99 "mean": 6.810217976570129,
100 "std": 0.34114415830274575,
101 "per_seed": [
102 6.926016807556152,
103 6.900168418884277,
104 6.7102837562561035,
105 6.759965896606445,
106 6.147101879119873,
107 6.569408893585205,
108 7.09395170211792,
109 7.374846458435059
110 ],
111 "n": 8
112 }
113 },
114 "idea_sweep": [
115 {
116 "cfg": {
117 "lr": 0.01,
118 "wd": 0.0,
119 "lambda": 1.0,
120 "beta": 0.95,
121 "noise": 0.001
122 },
123 "mean": 6.824108719825745
124 },
125 {
126 "cfg": {
127 "lr": 0.001,
128 "wd": 0.0,
129 "lambda": 1.0,
130 "beta": 0.95,
131 "noise": 0.001
132 },
133 "mean": 13.783732652664185
134 },
135 {
136 "cfg": {
137 "lr": 0.003,
138 "wd": 0.0,
139 "lambda": 1.0,
140 "beta": 0.95,
141 "noise": 0.001
142 },
143 "mean": 7.443580269813538
144 }
145 ],
146 "protocol": {
147 "paired_seeds": [
148 0,
149 1,
150 2,
151 3,
152 4,
153 5,
154 6,
155 7
156 ],
157 "epochs": 15,
158 "n_train": 1200,
159 "n_test": 400,
160 "structural_match": "optimizer intervention on tabular regression"
161 }
162}