Human Preference on StableDiffusion Evaluation Set 3
1.9088ImageRewardGradient Ascent
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Gradient AscentBackbone=StableDiffusion 32025.09 | 1.9088 | 0.2247 | 0.2877 | 5.5775 | 0.1474 | 0.9195 | |
| Trajectory Optimization ControlBackbone=StableDiffusion 32025.09 | 1.8529 | 0.24 | 0.289 | 6.173 | 0.2475 | 0.9013 | |
| Inversion+FreeDoMBackbone=StableDiffusion 32025.09 | 1.5887 | 0.2288 | 0.2305 | 5.7446 | 0.546 | 0.6893 | |
| Inversion+TFGBackbone=StableDiffusion 32025.09 | 1.5162 | 0.2216 | 0.2745 | 5.6072 | 0.3083 | 0.8537 | |
| Inversion+DPSBackbone=StableDiffusion 32025.09 | 1.4169 | 0.2189 | 0.2552 | 5.7227 | 0.3767 | 0.7896 | |
| NoneBackbone=StableDiffusion 32025.09 | 0.1542 | 0.2385 | 0.2887 | 6.0516 | 0 | 1 |