Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Safety Alignment on PKU-SafeRLHF (IID)

89.26WR

ShaPO-T

21.233638.894356.55574.2157Feb 7, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.02
89.260.750.13
2026.02
89.231.130
2026.02
88.932.010.13
2026.02
87.451.130.13
2026.02
85.881.250.13
2026.02
85.452.760.13
2026.02
85.291.750.5
2026.02
83.772.380.5
2026.02
82.744.512.01
2026.02
82.54.390.5
2026.02
82.136.643.26
2026.02
81.34.141.75
2026.02
75.8120.557.27
2026.02
73.9721.38.02
2026.02
73.711.286.14
2026.02
72.9125.4412.78
2026.02
72.711.786.14
2026.02
70.2226.8213.53
2026.02
69.1926.5714.41
2026.02
69.0932.2113.16
2026.02
67.6531.8313.66
2026.02
59.4837.9423.43
2026.02
57.1411.42.76
2026.02
56.1415.161.76
2026.02
52.6629.9518.17
2026.02
46.5457.7723.43
2026.02
45.2761.6522.68
2026.02
45.2760.2816.92
2026.02
44.9366.7918.42
2026.02
44.2357.1419.17
2026.02
42.8267.5424.69
2026.02
41.0571.326.32
2026.02
40.1873.3126.94
2026.02
35.9368.6745.36
2026.02
25.0970.5552.88
2026.02
23.8569.9252.51