Harmful prompt detection on adv-ViSU
100PrecisionDetoxify (Orig)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Detoxify (Orig)2026.04 | 100 | 7 | 13 | |
| DiffGuard2026.04 | 97 | 40 | 65 | |
| HyPE2026.04 | 97 | 67 | 80 | |
| NSFW-Classifier2026.04 | 62 | 65 | 64 | |
| Latent Guard2026.04 | 49 | 18 | 27 | |
| GuardT2I2026.04 | 42 | 71 | 53 |