List your company

Red-teaming on Llama-3-70B

45Attack Success Rate (ASR)Adaptive Instruction Composition
7.35217.12626.936.674Apr 22, 2026

Evaluation results

MethodLinks
Adaptive Instruction CompositionBandit Type=Aggressive, lambda=0.012026.04454,500
Adaptive Instruction CompositionBandit Type=Subtle, lambda=0.012026.0415.51,552
WildTeamingAttack Strategy=WildTeaming2026.048.8876