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

Overrefusal Detection on OR-Bench

93.42AUROC

T3+OCSVM

39.620853.587967.55581.5221Feb 4, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.02
93.4225.1796.6292.93
2026.02
93.1143.597.2990.63
2026.02
91.0822.1792.6594.05
2026.02
89.0514.8392.493.46
2026.02
87.1753.1791.8188.33
2026.02
85.9432.6790.2291.53
2026.02
85.814191.1490.6
2026.02
85.778994.7890.82
2026.02
84.9827.4898.3997.96
2026.02
7971.1790.1985.36
2026.02
75.368.8386.7985.47
2026.02
74.7475.586.7484.91
2026.02
71.6272.583.8684.93
2026.02
62.292.8378.6283.33
2026.02
55.0997.8374.1683.33
2026.02
54.3894.1774.283.33
2026.02
47.198.1768.7183.33
2026.02
41.6995.1766.6283.33