Unsafe Image Generation on Borderline (test)
32.37ASR (Q16)RICE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RICEModel=Janus2026.03 | 32.37 | 45.08 | 94.36 | |
| DACAModel=Janus2026.03 | 28.95 | 43.2 | 89.72 | |
| VanillaModel=Janus2026.03 | 27.96 | 43.98 | 94.14 | |
| SGTModel=Janus2026.03 | 27.29 | 43.65 | 86.63 | |
| SneakyModel=Janus2026.03 | 24.53 | 41.44 | 91.05 | |
| Self-CoTModel=Janus2026.03 | 21.99 | 30.72 | 81.44 | |
| RICEModel=Bagel2026.03 | 15.49 | 60.77 | 90.62 | |
| VanillaModel=Bagel2026.03 | 15.36 | 48.62 | 93.26 | |
| SneakyModel=Bagel2026.03 | 13.59 | 54.59 | 89.39 | |
| DACAModel=Bagel2026.03 | 10.39 | 54.48 | 88.84 | |
| Self-CoTModel=Bagel2026.03 | 9.39 | 47.29 | 88.07 | |
| SGTModel=Bagel2026.03 | 7.29 | 49.39 | 87.4 |