Over-refusal Accuracy on XSTest and OKTest
99.2Over-refusal Accuracy (XSTest)
ReSA-RL (Ours)
Updated 1mo ago
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
2025.09 | 99.2 | 95.33 | 97.27 | |
| 97.2 | 87.67 | 92.44 | ||
2025.09 | 97.2 | 88.67 | 92.94 | |
2025.09 | 94.8 | 91 | 92.9 | |
2025.09 | 94.4 | 89.33 | 91.87 | |
2025.09 | 93.2 | 84 | 88.6 | |
2025.09 | 90.8 | 90.33 | 90.57 | |
2025.09 | 90 | 91 | 90.5 | |
2025.09 | 89.6 | 74.33 | 81.97 | |
2025.09 | 81.2 | 80.33 | 80.77 | |
2025.09 | 80.8 | 66 | 73.4 | |
2025.09 | 51.6 | 59.67 | 55.65 |