List your company

Adversarial Attack on Claude Opus 4.1 (test)

46.7Bin-ASRPLAGUE
7.90817.97928.0538.121Oct 20, 2025

Evaluation results

MethodLinks
PLAGUEInteraction budget=equal budget (six queries per behavior), Configuration=Best2025.1046.767.3
FITDAttacker model=DeepSeek-R1, Interaction budget=six queries per behavior2025.1010.556
X-TeamingAttacker model=DeepSeek-R1, Interaction budget=six queries per behavior2025.109.422.8