Attack Success Rate on AutoDAN
10.5ASRQwen2.5-14B-IT + SELF-REDTEAM
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-14B-IT + SELF-REDTEAMModel Size=14B, Training=SELF-REDTEAM2025.06 | 10.5 | |
| Qwen2.5-14B-ITModel Size=14B2025.06 | 17 | |
| Qwen2.5-7B-IT + SELF-REDTEAMModel Size=7B, Training=SELF-REDTEAM2025.06 | 33.6 | |
| Qwen2.5-3B-IT + SELF-REDTEAMModel Size=3B, Training=SELF-REDTEAM2025.06 | 45.2 | |
| Qwen2.5-3B-ITModel Size=3B2025.06 | 51.8 | |
| Qwen2.5-7B-ITModel Size=7B2025.06 | 55.5 |