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

Malicious Query Safety Evaluation on AdvBench

100Refusal Rate

SSD

91.201693.485895.7798.0542Apr 18, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
100
2026.04
99.81
2026.04
99.81
2026.04
99.81
2026.04
99.62
2026.04
99.62
2026.04
99.42
2026.04
99.23
2026.04
99.04
2026.04
98.85
2026.04
98.08
2026.04
97.5
2026.04
97.31
2026.04
97.12
2026.04
95.96
2026.04
94.62
2026.04
93.65
2026.04
91.54