Laplace-Heterogeneous MoE Routing / results.json
Failed on benchmark
1{
2 "seed": 46,
3 "math_verification": {
4 "p1_derivative_variance_max_abs_error": 2.6500089234104962e-08,
5 "p1_max_variance_scale": 27.044748354034365,
6 "p2_m_at_zero": 5.3125,
7 "p2_weighted_mean_prediction": 5.3125,
8 "p2_initial_abs_error": 0.0,
9 "p2_min_lambda": 0.25,
10 "p2_m_at_pressure_20": 0.25000022942667016,
11 "p2_observed_within_1pct_pressure": 7.65,
12 "p2_pairwise_dominance_prediction": 7.600591432023894,
13 "p3_log_q_derivative_hazard_max_abs_error": 4.1232674163893535e-07,
14 "heterogeneity_sweep": [
15 {
16 "high_lambda": 1.0,
17 "m0": 0.625,
18 "curvature_at_0": 0.140625,
19 "curvature_at_0.5": 0.1357947569547857
20 },
21 {
22 "high_lambda": 2.0,
23 "m0": 1.125,
24 "curvature_at_0": 0.765625,
25 "curvature_at_0.5": 0.6359358495992439
26 },
27 {
28 "high_lambda": 4.0,
29 "m0": 2.125,
30 "curvature_at_0": 3.515625,
31 "curvature_at_0.5": 1.621191811309358
32 },
33 {
34 "high_lambda": 8.0,
35 "m0": 4.125,
36 "curvature_at_0": 15.015625,
37 "curvature_at_0.5": 1.1963818593093694
38 },
39 {
40 "high_lambda": 16.0,
41 "m0": 8.125,
42 "curvature_at_0": 62.015625,
43 "curvature_at_0.5": 0.09422409138497
44 },
45 {
46 "high_lambda": 32.0,
47 "m0": 16.125,
48 "curvature_at_0": 252.015625,
49 "curvature_at_0.5": 0.00012854714877870432
50 }
51 ]
52 },
53 "routing": {
54 "baseline": {
55 "mean_cv_last_50": 2.3968464796216757,
56 "mean_overflow_last_50": 0.77734375,
57 "mean_router_entropy_last_50": 1.499848845300241,
58 "final_load": [
59 469,
60 32,
61 3,
62 1,
63 0,
64 2,
65 4,
66 1
67 ],
68 "max_load_last_50": 469.0
69 },
70 "mixture": {
71 "mean_cv_last_50": 0.18015315762314132,
72 "mean_overflow_last_50": 0.014140625,
73 "mean_router_entropy_last_50": 1.9113420128496639,
74 "final_load": [
75 36,
76 78,
77 71,
78 68,
79 68,
80 64,
81 63,
82 64
83 ],
84 "max_load_last_50": 79.0
85 },
86 "single": {
87 "mean_cv_last_50": 0.49192256897439757,
88 "mean_overflow_last_50": 0.115546875,
89 "mean_router_entropy_last_50": 1.8726956006710163,
90 "final_load": [
91 1,
92 125,
93 76,
94 62,
95 65,
96 57,
97 66,
98 60
99 ],
100 "max_load_last_50": 127.0
101 },
102 "linear": {
103 "mean_cv_last_50": 0.7869943572429986,
104 "mean_overflow_last_50": 0.3047265625,
105 "mean_router_entropy_last_50": 1.8817180149088584,
106 "final_load": [
107 128,
108 170,
109 45,
110 31,
111 30,
112 36,
113 34,
114 38
115 ],
116 "max_load_last_50": 170.0
117 }
118 }
119}