Jailbreak Defense on Original Dataset
5.82ASREDDF
Evaluation Results
| Method | Links | |
|---|---|---|
| EDDFTarget Model=qwen plus2025.02 | 5.82 | |
| Defense PromptTarget Model=qwen plus2025.02 | 9.93 | |
| Intention AnalysisTarget Model=qwen plus2025.02 | 12.58 | |
| Self-ReminderTarget Model=qwen plus2025.02 | 16.37 | |
| EmbeddingTarget Model=qwen plus2025.02 | 36.4 | |
| RegexTarget Model=qwen plus2025.02 | 46.03 | |
| Llama3-GuardTarget Model=qwen plus2025.02 | 55 | |
| Guard Few-shotTarget Model=qwen plus2025.02 | 71.47 |