Harmful prompt detection on ViSU
98PrecisionDetoxify (Orig)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Detoxify (Orig)2026.04 | 98 | 26 | 40 | |
| HyPE2026.04 | 98 | 98 | 98 | |
| HyPEAttack Type=Style Attack, Attack Strength (p)=0.42026.04 | 97 | 67 | 80 | |
| Latent Guard2026.04 | 79 | 52 | 63 | |
| NSFW-Classifier2026.04 | 70 | 80 | 75 | |
| NSFW-ClassifierAttack Type=Style Attack, Attack Strength (p)=0.42026.04 | 65 | 65 | 65 | |
| GuardT2I2026.04 | 48 | 77 | 59 | |
| DiffGuard2026.04 | 27 | 36 | 31 | |
| DiffGuardAttack Type=Style Attack, Attack Strength (p)=0.42026.04 | 0 | 0 | 0 | |
| Detoxify (Orig)Attack Type=Style Attack, Attack Strength (p)=0.42026.04 | 0 | 0 | 0 | |
| Latent GuardAttack Type=Style Attack, Attack Strength (p)=0.42026.04 | 0 | 0 | 0 |