Joint Modeling for Stochastic Interventions / bench_report.json
Mechanism confirmed, baseline not beaten
1{
2 "bench_version": 1,
3 "track": "stochastic_intervention_scm",
4 "model": "mlp_tiny",
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": 0.003
15 },
16 "mean": 2.497592270374298
17 },
18 {
19 "cfg": {
20 "lr": 0.01
21 },
22 "mean": 2.5648475885391235
23 },
24 {
25 "cfg": {
26 "lr": 0.03
27 },
28 "mean": 2.6781041622161865
29 }
30 ],
31 "full": {
32 "mean": 2.414719134569168,
33 "std": 0.23689432610640407,
34 "per_seed": [
35 2.342486619949341,
36 2.4945199489593506,
37 2.7365739345550537,
38 2.4167885780334473,
39 2.2174108028411865,
40 2.058196783065796,
41 2.258479356765747,
42 2.793297052383423
43 ],
44 "n": 8
45 }
46 },
47 "idea": {
48 "mean": 2.2138066589832306,
49 "std": 0.5331661449712295,
50 "per_seed": [
51 2.4092166423797607,
52 2.4376542568206787,
53 2.0347108840942383,
54 1.5852525234222412,
55 1.881182074546814,
56 1.8407953977584839,
57 3.4355263710021973,
58 2.0861151218414307
59 ],
60 "n": 8,
61 "best_cfg": {
62 "lr": 0.03
63 },
64 "sweep": [
65 {
66 "cfg": {
67 "lr": 0.003
68 },
69 "mean": 7.919645547866821,
70 "std": 2.5149207407136784,
71 "per_seed": [
72 10.004054069519043,
73 5.846197605133057,
74 7.366078853607178,
75 4.32836389541626,
76 7.836812496185303,
77 5.4305877685546875,
78 11.452704429626465,
79 11.092365264892578
80 ],
81 "n": 8
82 },
83 {
84 "cfg": {
85 "lr": 0.01
86 },
87 "mean": 3.1929454803466797,
88 "std": 0.8334051410010931,
89 "per_seed": [
90 3.184743642807007,
91 3.0707268714904785,
92 3.08829665184021,
93 2.054521322250366,
94 2.8313469886779785,
95 2.428166151046753,
96 4.886856555938721,
97 3.998905658721924
98 ],
99 "n": 8
100 },
101 {
102 "cfg": {
103 "lr": 0.03
104 },
105 "mean": 2.2138066589832306,
106 "std": 0.5331661449712295,
107 "per_seed": [
108 2.4092166423797607,
109 2.4376542568206787,
110 2.0347108840942383,
111 1.5852525234222412,
112 1.881182074546814,
113 1.8407953977584839,
114 3.4355263710021973,
115 2.0861151218414307
116 ],
117 "n": 8
118 }
119 ]
120 },
121 "comparison": {
122 "delta_mean": -0.2009124755859375,
123 "idea_wins": 6,
124 "n_pairs": 8,
125 "per_seed_diffs": [
126 0.06673002243041992,
127 -0.056865692138671875,
128 -0.7018630504608154,
129 -0.831536054611206,
130 -0.33622872829437256,
131 -0.217401385307312,
132 1.1770470142364502,
133 -0.7071819305419922
134 ],
135 "p_value": 0.41835,
136 "mde": 0.5395513694218982,
137 "mde_rel_pct": 22.34427025891624,
138 "verdict": "no significant win",
139 "system_worked": false
140 },
141 "custom_track": {
142 "name": "stochastic_intervention_scm",
143 "file": "custom_stochastic_intervention_track.py",
144 "domain": "causal-world-model"
145 },
146 "mechanism_signature": {
147 "baseline_predicted_selected_mean": 1.6279271692037582,
148 "idea_predicted_selected_mean": 1.8218559473752975,
149 "observed_selected_mean": 1.7572368532419205,
150 "baseline_abs_error": 0.12930968403816223,
151 "idea_abs_error": 0.06461909413337708,
152 "confirmed": true
153 }
154}