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

Target behavior attack on MMLU bio

98.4Attack Success Rate (ASR)

AiTM

38.39253.97169.5585.129Feb 20, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.02
98.4
2025.02
96.5
2025.02
96.3
2025.02
96.3
2025.02
94.9
2025.02
93.7
2025.02
93.3
2025.02
93.1
2025.02
89.2
2025.02
81.7
2025.02
68.2
2025.02
52.4
2025.02
51.5
2025.02
48.7
2025.02
43.9
2025.02
40.7