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

Reward Modeling on Unified Feedback (UF)

78.9Accuracy

GRM-SFT

68.08470.89273.776.508Feb 11, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.02
78.9
2026.02
78.8
2026.02
78.7
2026.02
77.3
2026.02
77.3
2026.02
77.3
2026.02
77.2
2026.02
77
2026.02
76.9
2026.02
76.9
2026.02
76.7
2026.02
76.6
2026.02
76.6
2026.02
76.6
2026.02
76.2
2026.02
75.8
2026.02
75.5
2026.02
75.2
2026.02
75.1
2026.02
74.7
2026.02
74.7
2026.02
74.5
2026.02
74.2
2026.02
74.1
2026.02
73.9
2026.02
73.9
2026.02
73.8
2026.02
73.2
2026.02
72.8
2026.02
72.1
2026.02
72
2026.02
71.6
2026.02
71.5
2026.02
71.5
2026.02
71.4
2026.02
70.2
2026.02
69.9
2026.02
69.6
2026.02
68.8
2026.02
68.5