Black-box LLM Jailbreaking on AdvBench Gemma 3-12B target (test)
100ASRFlipAttack
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FlipAttackTrial setting=50-trial2026.02 | 100 | 1.41 | |
| TrailBlazerTrial setting=50-trial2026.02 | 100 | 3.54 | |
| RLbreakerTrial setting=50-trial2026.02 | 98.71 | 1.58 | |
| PAIRTrial setting=50-trial2026.02 | 83.97 | 10.41 | |
| AutoDAN-TurboTrial setting=50-trial2026.02 | 49.36 | 17.57 | |
| Adaptive AttacksTrial setting=50-trial2026.02 | 35.26 | 13.28 |