Goal Hijacking on Safety-Prompts
92Mean Accuracy
Scenario-Tailored PC-Inj
Updated 1mo ago
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
2024.10 | 92 | 0.7 | 13.3 | |
2024.10 | 90.6 | 0.6 | 13.1 | |
2024.10 | 89.3 | 0.8 | 10.6 | |
2024.10 | 88 | 0.5 | 10.5 | |
2024.10 | 85.3 | 0.4 | 6.6 | |
2024.10 | 84.5 | 0.4 | 15.2 | |
2024.10 | 83.8 | 0.4 | 6.3 | |
| 78.7 | 0.6 | - | ||
2024.10 | 78.2 | 0.6 | 8.9 | |
| 77.5 | 0.5 | - | ||
2024.10 | 76.5 | 0.7 | 7.2 | |
| 69.3 | 2.3 | - |