Red-teaming on Llama-3.3-70B
0.558Attack Success Rate (ASR)Adaptive Instruction Composition
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Adaptive Instruction CompositionBandit Type=Aggressive, lambda=0.012026.04 | 0.558 | 5,579 | |
| Adaptive Instruction CompositionBandit Type=Subtle, lambda=0.012026.04 | 0.247 | 2,465 | |
| WildTeamingAttack Strategy=WildTeaming2026.04 | 0.183 | 1,830 |