List your company

Overrefusal Detection on OR-Bench

93.42AUROCT3+OCSVM
39.620853.587967.55581.5221Feb 4, 2026

Evaluation results

MethodLinks
T3+OCSVMaugmented=false2026.0293.4225.1796.6292.93
DUOGUARD2026.0293.1143.597.2990.63
T3+GMMaugmented=false2026.0291.0822.1792.6594.05
LLAMAGUARD3-1Bparameters=1B2026.0289.0514.8392.493.46
POLYGUARD2026.0287.1753.1791.8188.33
T3+GMMaugmented=true, augmentation_source=GPT-OSS-20B2026.0285.9432.6790.2291.53
T3+OCSVMaugmented=true, augmentation_source=GPT-OSS-20B2026.0285.814191.1490.6
MDJUDGE2026.0285.778994.7890.82
LLAMAGUARD4-12Bparameters=12B2026.0284.9827.4898.3997.96
CIDER2026.027971.1790.1985.36
GMM2026.0275.368.8386.7985.47
RMD2026.0274.7475.586.7484.91
VIM2026.0271.6272.583.8684.93
NNGUIDE2026.0262.292.8378.6283.33
ADASCALE2026.0255.0997.8374.1683.33
REACT2026.0254.3894.1774.283.33
OPENMAX2026.0247.198.1768.7183.33
FDBD2026.0241.6995.1766.6283.33