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

Human Preference Alignment on HPD v2

0.379HPS-v2.1

BranchGRPO

0.268240.2969950.325750.354505Nov 21, 2025Nov 24, 2025Nov 27, 2025Nov 30, 2025Dec 3, 2025Dec 6, 2025Dec 9, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.11
0.379-6.5710.3623.138
2025.11
0.379-6.5030.3683.132
2025.11
0.372-6.6260.3713.166
2025.11
0.371-6.3650.3673.125
2025.12
0.3641.34266.42370.383-
2025.11
0.356-6.40.3663.114
2025.12
0.35211.20566.04880.3812-
2025.12
0.34851.3936.32240.3862-
2025.12
0.27481.00615.850.3884-
2025.12
0.27250.8875.95190.3996-