Safety Evaluation on Harmful Prompts
8.3Harmful ScoreSurgery
Evaluation Results
| Method | Links | |
|---|---|---|
| Surgeryn (sample number)=2000, p (harmful ratio)=0.12026.02 | 8.3 | |
| Surgeryn (sample number)=2500, p (harmful ratio)=0.12026.02 | 8.4 | |
| Surgeryn (sample number)=1000, p (harmful ratio)=0.12026.02 | 8.9 | |
| Surgeryn (sample number)=1500, p (harmful ratio)=0.12026.02 | 9.4 | |
| Surgeryn (sample number)=500, p (harmful ratio)=0.12026.02 | 9.7 | |
| Lisan (sample number)=500, p (harmful ratio)=0.12026.02 | 13.4 | |
| Lisan (sample number)=2500, p (harmful ratio)=0.12026.02 | 13.9 | |
| Lisan (sample number)=1000, p (harmful ratio)=0.12026.02 | 14.8 | |
| DSSn (sample number)=500, p (harmful ratio)=0.12026.02 | 15.6 | |
| AsFTn (sample number)=1000, p (harmful ratio)=0.12026.02 | 16.1 | |
| AsFTn (sample number)=2500, p (harmful ratio)=0.12026.02 | 16.2 | |
| Lisan (sample number)=1500, p (harmful ratio)=0.12026.02 | 16.4 | |
| AsFTn (sample number)=1500, p (harmful ratio)=0.12026.02 | 16.4 | |
| Lisan (sample number)=2000, p (harmful ratio)=0.12026.02 | 16.5 | |
| SafeGradn (sample number)=500, p (harmful ratio)=0.12026.02 | 16.5 | |
| AsFTn (sample number)=500, p (harmful ratio)=0.12026.02 | 16.5 | |
| SafeGradn (sample number)=1500, p (harmful ratio)=0.12026.02 | 16.6 | |
| AsFTn (sample number)=17.00, p (harmful ratio)=0.12026.02 | 17 | |
| SafeGradn (sample number)=2000, p (harmful ratio)=0.12026.02 | 17.6 | |
| ConstrainedSFTn (sample number)=1000, p (harmful ratio)=0.12026.02 | 17.6 | |
| ConstrainedSFTn (sample number)=500, p (harmful ratio)=0.12026.02 | 17.7 | |
| DSSn (sample number)=1000, p (harmful ratio)=0.12026.02 | 17.8 | |
| SafeGradn (sample number)=2500, p (harmful ratio)=0.12026.02 | 18 | |
| DSSn (sample number)=1500, p (harmful ratio)=0.12026.02 | 18 | |
| DSSn (sample number)=2500, p (harmful ratio)=0.12026.02 | 18.3 | |
| SafeGradn (sample number)=1000, p (harmful ratio)=0.12026.02 | 18.5 | |
| DSSn (sample number)=2000, p (harmful ratio)=0.12026.02 | 18.6 | |
| SPARDn (sample number)=1500, p (harmful ratio)=0.12026.02 | 20.1 | |
| ConstrainedSFTn (sample number)=2000, p (harmful ratio)=0.12026.02 | 20.6 | |
| SPARDn (sample number)=1000, p (harmful ratio)=0.12026.02 | 20.6 | |
| ConstrainedSFTn (sample number)=1500, p (harmful ratio)=0.12026.02 | 20.9 | |
| SPARDn (sample number)=500, p (harmful ratio)=0.12026.02 | 21.1 | |
| ConstrainedSFTn (sample number)=2500, p (harmful ratio)=0.12026.02 | 22 | |
| SPARDn (sample number)=2500, p (harmful ratio)=0.12026.02 | 22.4 | |
| SPARDn (sample number)=2000, p (harmful ratio)=0.12026.02 | 23.5 | |
| SFTn (sample number)=2500, p (harmful ratio)=0.12026.02 | 24.1 | |
| SFTn (sample number)=500, p (harmful ratio)=0.12026.02 | 24.4 | |
| SFTn (sample number)=1500, p (harmful ratio)=0.12026.02 | 24.7 | |
| SFTn (sample number)=1000, p (harmful ratio)=0.12026.02 | 27.6 | |
| SFTn (sample number)=2000, p (harmful ratio)=0.12026.02 | 28.4 |