Black-box LLM Jailbreaking on AdvBench GPT-oss-20B target (test)
85.3ASRTrailBlazer
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TrailBlazerTrial setting=50-trial2026.02 | 85.3 | 21.77 | |
| FlipAttackTrial setting=50-trial2026.02 | 51.92 | 14.27 | |
| PAIRTrial setting=50-trial2026.02 | 46.79 | 17.81 | |
| RLbreakerTrial setting=50-trial2026.02 | 4.48 | 26 | |
| AutoDAN-TurboTrial setting=50-trial2026.02 | 3.21 | 19.6 | |
| Adaptive AttacksTrial setting=50-trial2026.02 | 0 | — |