Jailbreak Attack Defense on ReNeLLM languageMix
34ASRIntention Analysis
Evaluation Results
| Method | Links | |
|---|---|---|
| Intention AnalysisBase LLM=Vicuna2025.08 | 34 | |
| Context FilteringBase LLM=Vicuna2025.08 | 42 | |
| Self-ExaminationBase LLM=Vicuna2025.08 | 86 | |
| Safe DecodingBase LLM=Vicuna2025.08 | 90 | |
| Self-ReminderBase LLM=Vicuna2025.08 | 94 | |
| ICDBase LLM=Vicuna2025.08 | 96 | |
| No DefenseBase LLM=Vicuna2025.08 | 100 |