Safety Generalization on i2p (test)
85.31Self-Harm ScoreSAeUron
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| SAeUronModel=Stable Diffusion 3.5 Large, Evaluation Level=Prompt-level2026.05 | 85.31 | 94.17 | 81.55 | 89.51 | 89.22 | 83.74 | 75.78 | 84.11 | -6.32 | |
| BaseModel=Stable Diffusion 3.5 Large, Evaluation Level=Prompt-level2026.05 | 79.06 | 82.59 | 77.92 | 86.01 | 83.05 | 76.56 | 68 | 77.79 | — | |
| EraseDiffModel=Stable Diffusion 3.5 Large, Evaluation Level=Prompt-level2026.05 | 77.89 | 82.96 | 73.79 | 83.72 | 78.71 | 73.5 | 65.89 | 75.3 | 2.49 | |
| SAFREEModel=Stable Diffusion 3.5 Large, Evaluation Level=Prompt-level2026.05 | 68.46 | 80.37 | 68.48 | 77.4 | 70.99 | 66.75 | 52.34 | 67 | 10.79 | |
| SAFREEModel=FLUX.1 Dev, Evaluation Level=Prompt-level2026.05 | 67.26 | 65.77 | 60.36 | 73.7 | 55.52 | 57.08 | 51.87 | 61.34 | -7.36 | |
| ErasingModel=Stable Diffusion 3.5 Large, Evaluation Level=Prompt-level2026.05 | 64.24 | 78.03 | 67.79 | 76.6 | 69.61 | 65.3 | 48.44 | 64.81 | 12.98 | |
| ErasingModel=FLUX.1 Dev, Evaluation Level=Prompt-level2026.05 | 63.46 | 66.82 | 57.54 | 69.12 | 55.18 | 56.85 | 49.22 | 58.47 | -4.5 | |
| EraseDiffModel=FLUX.1 Dev, Evaluation Level=Prompt-level2026.05 | 61.64 | 57.4 | 55.2 | 65.26 | 51.96 | 49.17 | 42.44 | 53.81 | 0.16 | |
| SAeUronModel=FLUX.1 Dev, Evaluation Level=Prompt-level2026.05 | 60.6 | 67.71 | 56.81 | 66.95 | 54.48 | 54.67 | 48.56 | 57.31 | -3.33 | |
| SafeDIGModel=Stable Diffusion 3.5 Large, Evaluation Level=Prompt-level2026.05 | 60.05 | 57.14 | 50.29 | 68.06 | 59.54 | 52.08 | 44.71 | 55.36 | 22.43 | |
| BaseModel=FLUX.1 Dev, Evaluation Level=Prompt-level2026.05 | 59.43 | 57.14 | 55.26 | 65.74 | 51.96 | 48.79 | 44.56 | 53.97 | — | |
| BaseModel=Stable Diffusion 3.5 Large, Evaluation Level=Line-level2026.05 | 48.38 | 57.4 | 43.41 | 57.99 | 49.26 | 46.25 | 29.63 | 45.71 | — | |
| EraseDiffModel=Stable Diffusion 3.5 Large, Evaluation Level=Line-level2026.05 | 48.21 | 56.8 | 43.3 | 58.15 | 47.09 | 45.29 | 29.76 | 45.05 | 0.65 | |
| SafeDIGModel=FLUX.1 Dev, Evaluation Level=Prompt-level2026.05 | 44.81 | 51.34 | 32.43 | 43.9 | 36.1 | 36.82 | 30.29 | 38.01 | 15.96 | |
| SAeUronModel=Stable Diffusion 3.5 Large, Evaluation Level=Line-level2026.05 | 40.74 | 49.57 | 36.31 | 49.24 | 45.21 | 40.66 | 26.06 | 39.64 | 6.07 | |
| SAFREEModel=Stable Diffusion 3.5 Large, Evaluation Level=Line-level2026.05 | 40.2 | 46.23 | 33.86 | 49.6 | 39.79 | 35.01 | 22.28 | 36.76 | 8.95 | |
| SAFREEModel=FLUX.1 Dev, Evaluation Level=Line-level2026.05 | 38.1 | 28.12 | 28.04 | 41.61 | 27.22 | 22.14 | 20.47 | 30.06 | -5.32 | |
| ErasingModel=Stable Diffusion 3.5 Large, Evaluation Level=Line-level2026.05 | 37.03 | 48.74 | 34.26 | 48.8 | 39.14 | 36.37 | 19.91 | 35.9 | 9.8 | |
| BaseModel=FLUX.1 Dev, Evaluation Level=Line-level2026.05 | 29 | 25.8 | 22.56 | 35.67 | 23.65 | 21.24 | 16.49 | 24.74 | — | |
| EraseDiffModel=FLUX.1 Dev, Evaluation Level=Line-level2026.05 | 28.79 | 25.41 | 21.02 | 34.05 | 22.22 | 20.32 | 15.86 | 23.66 | 1.09 | |
| SafeDIGModel=Stable Diffusion 3.5 Large, Evaluation Level=Line-level2026.05 | 28.16 | 33.9 | 22.7 | 33.9 | 27.24 | 25.47 | 16.23 | 25.69 | 20.02 | |
| ErasingModel=FLUX.1 Dev, Evaluation Level=Line-level2026.05 | 25.27 | 31.17 | 19.71 | 32.23 | 22 | 22.84 | 15.36 | 23.05 | 1.69 | |
| SAeUronModel=FLUX.1 Dev, Evaluation Level=Line-level2026.05 | 23.1 | 26.54 | 18.34 | 28.64 | 19.07 | 20.89 | 14.93 | 21.02 | 3.72 | |
| SafeDIGModel=FLUX.1 Dev, Evaluation Level=Line-level2026.05 | 13.48 | 17.49 | 9.96 | 14.52 | 9.58 | 11.47 | 7.98 | 11.51 | 13.23 |