Compositional Text-to-Image Generation on T2I-CompBench Spatial
0.275CLIP ScoreSDXL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SDXL2026.04 | 0.275 | 0.953 | 0.002 | |
| Attend&Excite*fall back to vanilla SDXL=true2026.04 | 0.275 | 0.957 | 0.005 | |
| Golden Noise2026.04 | 0.274 | 0.956 | 0.003 | |
| RPG2026.04 | 0.27 | 0.95 | 0.002 | |
| MultiDiffusion*fall back to vanilla SDXL=true2026.04 | 0.268 | 0.95 | 0.003 | |
| Golden RPG2026.04 | 0.267 | 0.958 | 0.002 |