Consistent Text-to-Image Generation on ConsiStory+
0.8889CLIP-TSDXL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SDXLType=Unconstrained baseline2025.12 | 0.8889 | 0.8972 | 0.7298 | 0.1989 | |
| Geometric Disentanglement of Text EmbeddingsMode=Dual-subspace orthogonal projection, Training-free=true2025.12 | 0.8766 | 0.9168 | 0.79 | 0.157 | |
| 1P1SFull Name=1Prompt1Story2025.12 | 0.8252 | 0.8826 | 0.6898 | 0.2195 |