Harmful prompt detection on NSFW56k
99AccuracyHyPE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HyPE2026.04 | 99 | — | |
| NSFW-Classifier2026.04 | 95 | — | |
| DiffGuard2026.04 | 89 | — | |
| Latent Guard2026.04 | 52 | — | |
| Detoxify (Orig)2026.04 | 34 | — | |
| GuardT2I2026.04 | 9 | — | |
| Detoxify (Orig)Attack Type=Style Attack, Attack Strength (p)=0.42026.04 | — | 91 | |
| DiffGuardAttack Type=Style Attack, Attack Strength (p)=0.42026.04 | — | 92 | |
| HyPEAttack Type=Style Attack, Attack Strength (p)=0.42026.04 | — | 27 | |
| Latent GuardAttack Type=Style Attack, Attack Strength (p)=0.42026.04 | — | 94 | |
| NSFW-ClassifierAttack Type=Style Attack, Attack Strength (p)=0.42026.04 | — | 82 |