Benign Query Discrimination on Benign Queries
0.019False Positive Rate (FPR)Guard Few-shot
Evaluation Results
| Method | Links | |
|---|---|---|
| Guard Few-shotTarget Model=qwen plus2025.02 | 0.019 | |
| EDDFTarget Model=qwen plus2025.02 | 0.0218 | |
| Llama3-GuardTarget Model=qwen plus2025.02 | 0.083 | |
| RegexTarget Model=qwen plus2025.02 | 0.0871 | |
| EmbeddingTarget Model=qwen plus2025.02 | 0.1227 | |
| Self-ReminderTarget Model=qwen plus2025.02 | 0.1246 | |
| Defense PromptTarget Model=qwen plus2025.02 | 0.1975 | |
| Intention AnalysisTarget Model=qwen plus2025.02 | 0.3489 |