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

Preference Alignment on UFB (test)

81.05Win Rate

CW-DPO

56.152462.616269.0875.5438Mar 5, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.03
81.05-
2026.03
80.41-
2026.03
80.17-
2026.03
78.12-
2026.03
77.1-
2026.03
76.15-
2026.03
75.96-
2026.03
74.03-
2026.03
73.12-
2026.03
65.17-
2026.03
63.95-
2026.03
63.72-
2026.03
63.61-
2026.03
63.41-
2026.03
62.63-
2026.03
62.12-
2026.03
60.96-
2026.03
57.11-
2026.03
-61.3
2026.03
-63.4
2026.03
-63.1
2026.03
-63.4
2026.03
-61.3
2026.03
-66.4
2026.03
-58.9
2026.03
-61.2
2026.03
-63.7
2026.03
-78.1
2026.03
-78.3
2026.03
-80.1
2026.03
-78.5
2026.03
-77.2
2026.03
-80.7
2026.03
-72.4
2026.03
-75.1
2026.03
-76.8