Adversarial Attack Diversity Analysis on Llama 70B 3.3
0.266Average Attack SimilarityWildTeaming
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| WildTeamingAttack Method=WT2026.04 | 0.266 | 0.164 | |
| AICAttack Method=AIC Subtle2026.04 | 0.269 | 0.155 | |
| AICAttack Method=AIC Aggr.2026.04 | 0.353 | 0.24 |