Text-to-Image Generation on InterActing Compositional Relation scenario sampled
4.283Human Likert ScoreDetailScribe
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DetailScribeBackbone=Stable Diffusion 3.52025.04 | 4.283 | 5 | 1.545 | 92.3 | 48.5 | |
| SD + GPT RefineBackbone=Stable Diffusion 3.5, Prompt Strategy=GPT Refine2025.04 | 3.667 | 4.8 | 1.39 | 88.9 | 38.9 | |
| SD + Inf ScaleBackbone=Stable Diffusion 3.5, Strategy=Inference Scaling2025.04 | 3.65 | 4.8 | 1.538 | 91.2 | 42.3 | |
| SDBackbone=Stable Diffusion 3.52025.04 | 3.583 | 4.867 | 1.415 | 89.8 | 37.9 | |
| DALL·E 32025.04 | 3.433 | 4.867 | 1.382 | 83.8 | 36.7 | |
| SD + GPT RewriteBackbone=Stable Diffusion 3.5, Prompt Strategy=GPT Rewrite2025.04 | 3.167 | 4.467 | 1.283 | 86.7 | 32.5 |