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

Safety Detection on Alpaca + XSTest full (val-selected)

0.991AUROC

wopt

0.49180.62140.7510.8806Apr 20, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
0.9910.894
2026.04
0.9890.809
2026.04
0.9890.869
2026.04
0.9890.816
2026.04
0.9880.857
2026.04
0.9880.857
2026.04
0.9810.766
2026.04
0.980.819
2026.04
0.980.742
2026.04
0.9790.76
2026.04
0.9770.718
2026.04
0.9770.718
2026.04
0.9740.676
2026.04
0.9730.61
2026.04
0.9710.672
2026.04
0.970.713
2026.04
0.9690.708
2026.04
0.9630.538
2026.04
0.7560.021
2026.04
0.7450.037
2026.04
0.7410.078
2026.04
0.7210.015
2026.04
0.7110.024
2026.04
0.7020.07
2026.04
0.6850.01
2026.04
0.6770
2026.04
0.6670.031
2026.04
0.6420.019
2026.04
0.6030.004
2026.04
0.590.003
2026.04
0.5750.034
2026.04
0.570.01
2026.04
0.5690.013
2026.04
0.5680.003
2026.04
0.5570.012
2026.04
0.5530.004
2026.04
0.540.001
2026.04
0.5340.003
2026.04
0.5170.004
2026.04
0.5170
2026.04
0.5150
2026.04
0.5110.001