Text-to-Image Generation on Webster 500 Memorized Prompts 2023 v1.4 (430 with available target images)
0.6364SSCD (Target)No mitigation
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| No mitigationDiffusion Model=SD v1.4, Denoising Sampler=DDIM Sampling2026.02 | 0.6364 | 0.491 | 0.0956 | 42.1447 | 0.3129 | |
| No mitigationModel=Realistic Vision, Sampler=DDIM2026.02 | 0.5839 | 0.4901 | 0.0865 | 36.5499 | 0.3183 | |
| RADS (without constraint)Diffusion Model=SD v1.4, Denoising Sampler=DDIM Sampling, Constraint=None2026.02 | 0.4998 | 0.3449 | 0.0786 | 36.3543 | 0.301 | |
| Ren et al.Model=Realistic Vision, Sampler=DDIM2026.02 | 0.4318 | 0.295 | 0.0826 | 34.8126 | 0.3014 | |
| Wen et al. (2024)Diffusion Model=SD v1.4, Denoising Sampler=DDIM Sampling2026.02 | 0.4187 | 0.2132 | 0.0613 | 31.7825 | 0.3056 | |
| Hintersdorf et al. (2024)Diffusion Model=SD v1.4, Denoising Sampler=DDIM Sampling2026.02 | 0.4149 | 0.2534 | 0.0573 | 32.4977 | 0.3119 | |
| Hintersdorf et al.Model=Realistic Vision, Sampler=DDIM2026.02 | 0.4035 | 0.3154 | 0.0514 | 29.4091 | 0.3175 | |
| Wen et al.Model=Realistic Vision, Sampler=DDIM2026.02 | 0.3827 | 0.2136 | 0.0596 | 28.2042 | 0.3107 | |
| Ren et al. (2024)Diffusion Model=SD v1.4, Denoising Sampler=DDIM Sampling2026.02 | 0.3781 | 0.2332 | 0.0695 | 32.7631 | 0.294 | |
| RADS (Ours)Diffusion Model=SD v1.4, Denoising Sampler=DDIM Sampling2026.02 | 0.2303 | 0.1553 | 0.0409 | 31.5678 | 0.2917 | |
| RADSModel=Realistic Vision, Sampler=DDIM2026.02 | 0.1992 | 0.1834 | 0.0364 | 33.0128 | 0.2727 | |
| Jain et al. (2025)Diffusion Model=SD v1.4, Denoising Sampler=DDIM Sampling2026.02 | 0.1816 | 0.062 | 0.2724 | 63.9765 | 0.2266 | |
| Jain et al.Model=Realistic Vision, Sampler=DDIM2026.02 | 0.1544 | 0.0489 | 0.3199 | 84.8658 | 0.2139 |