Jailbreaking Attack (Crescendo) on Jailbreak Bench (ASR)
64ASRw/o Defense
Evaluation results
| Method | Links | |
|---|---|---|
| w/o DefenseTarget Model=GPT-4o2026.04 | 64 | |
| Self ReminderTarget Model=GPT-4o2026.04 | 58 | |
| Avg. ASRTarget Model=GPT-4o2026.04 | 54.3 | |
| Smooth LLMTarget Model=GPT-4o2026.04 | 53 | |
| Input FilterTarget Model=GPT-4o2026.04 | 42 |