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

Refusal Detection on Alpaca + XSTest (val)

0.988AUROC

wopt

Updated 1mo ago

Evaluation Results

Method	Links
wopt 2026.04		0.988	83.4
theta two-class 2026.04		0.988	81.5
wopt 2026.04		0.987	83.6
wLDA 2026.04		0.987	83.3
wopt 2026.04		0.987	84
wopt 2026.04		0.981	71.8
wopt 2026.04		0.979	79.4
wLDA 2026.04		0.975	65.4
wLDA 2026.04		0.974	65.7
wopt 2026.04		0.974	75.4
theta two-class 2026.04		0.966	52.5
theta two-class 2026.04		0.966	54.2
wLDA 2026.04		0.964	64.2
wLDA 2026.04		0.961	53
wLDA 2026.04		0.958	57.5
theta two-class 2026.04		0.945	36
theta two-class 2026.04		0.915	45.7
theta two-class 2026.04		0.902	42.2
theta-norm 2026.04		0.77	0.1
wPC1 2026.04		0.76	2.2
theta-norm 2026.04		0.758	0.3
Random 2026.04		0.694	2.7
Random 2026.04		0.691	2.7
theta-norm 2026.04		0.679	0.3
Random 2026.04		0.675	3.3
Perplexity 2026.04		0.667	0.1
theta-norm 2026.04		0.661	0.3
Perplexity 2026.04		0.618	0.3
Perplexity 2026.04		0.611	0.3
Perplexity 2026.04		0.609	0
wPC1 2026.04		0.607	0.3
theta-norm 2026.04		0.605	0
Random 2026.04		0.604	2.2
theta-norm 2026.04		0.602	0
wPC1 2026.04		0.6	0.3
Random 2026.04		0.589	1
Perplexity 2026.04		0.589	0.1
Perplexity 2026.04		0.584	0.1
wPC1 2026.04		0.578	0.1
wPC1 2026.04		0.568	0
wPC1 2026.04		0.567	0.1
Random 2026.04		0.542	2.1