Label-Free Bayesian Truth Serum Reward / results.json
Mechanism confirmed, baseline not beaten
1{
2 "prediction_1_log_scaling": [
3 {
4 "rho": 0.5,
5 "observed": -0.689171032180306,
6 "predicted": -0.6891710321803058,
7 "abs_error": 1.1102230246251565e-16
8 },
9 {
10 "rho": 0.8,
11 "observed": -0.2221480310727081,
12 "predicted": -0.22214803107270806,
13 "abs_error": 5.551115123125783e-17
14 },
15 {
16 "rho": 1.0,
17 "observed": 0.0,
18 "predicted": 0.0,
19 "abs_error": 0.0
20 },
21 {
22 "rho": 1.5,
23 "observed": 0.4041362042581106,
24 "predicted": 0.40413620425811064,
25 "abs_error": 5.551115123125783e-17
26 },
27 {
28 "rho": 2.0,
29 "observed": 0.691153161953081,
30 "predicted": 0.691153161953081,
31 "abs_error": 0.0
32 }
33 ],
34 "prediction_2_monotonicity": {
35 "min_adjacent_delta": 0.041629319979983426,
36 "p_at_max": 0.98,
37 "r_at_min": -2.816263785742443,
38 "r_at_max": 1.0277862360994974
39 },
40 "prediction_3_variance_1_over_G": [
41 {
42 "G": 8,
43 "observed": 1.98586541769121,
44 "predicted": 0.2897319014138917,
45 "observed_times_G": 15.88692334152968,
46 "predicted_times_G": 2.3178552113111337,
47 "ratio": 6.854148293647287
48 },
49 {
50 "G": 16,
51 "observed": 0.30585307390337574,
52 "predicted": 0.14486595070694586,
53 "observed_times_G": 4.893649182454012,
54 "predicted_times_G": 2.3178552113111337,
55 "ratio": 2.1112833789500756
56 },
57 {
58 "G": 32,
59 "observed": 0.08655735952915722,
60 "predicted": 0.07243297535347293,
61 "observed_times_G": 2.769835504933031,
62 "predicted_times_G": 2.3178552113111337,
63 "ratio": 1.194999364678274
64 },
65 {
66 "G": 64,
67 "observed": 0.039064428805029146,
68 "predicted": 0.036216487676736464,
69 "observed_times_G": 2.5001234435218653,
70 "predicted_times_G": 2.3178552113111337,
71 "ratio": 1.0786365909834499
72 },
73 {
74 "G": 128,
75 "observed": 0.01876288762881845,
76 "predicted": 0.018108243838368232,
77 "observed_times_G": 2.4016496164887617,
78 "predicted_times_G": 2.3178552113111337,
79 "ratio": 1.036151699540468
80 },
81 {
82 "G": 256,
83 "observed": 0.009226378730077934,
84 "predicted": 0.009054121919184116,
85 "observed_times_G": 2.361952954899951,
86 "predicted_times_G": 2.3178552113111337,
87 "ratio": 1.0190252365090022
88 }
89 ],
90 "policy_simulation": {
91 "bts_neutral": {
92 "accuracy": 0.86225,
93 "pressure_flip_rate": 0.0,
94 "mean_reward_variance": 0.20875399578457615
95 },
96 "bts_pressure": {
97 "accuracy": 0.50325,
98 "pressure_flip_rate": 0.49675,
99 "mean_reward_variance": 0.2278711124689051
100 },
101 "agreement_neutral": {
102 "accuracy": 0.9975,
103 "pressure_flip_rate": 0.0,
104 "mean_reward_variance": 0.046902008056640625
105 },
106 "agreement_pressure": {
107 "accuracy": 0.5105,
108 "pressure_flip_rate": 0.4895,
109 "mean_reward_variance": 0.01296771240234375
110 }
111 }
112}