Research map / Image generation / Text-to-image diffusion
T2I-CompBench++: An Enhanced and Comprehensive Benchmark for Compositional Text-to-Image Generation
Kaiyi Huang, Chengqi Duan, Kaiyue Sun, Enze Xie, Zhenguo Li, Xihui Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune with reward-selected samples (GORS) | Stable v2 · Color | B-VQA: 0.5065 -> 0.6603 | Improved |
| fine-tune with reward-selected samples (GORS) | Stable v2 · Shape | B-VQA: 0.4221 -> 0.4785 | Improved |
| fine-tune with reward-selected samples (GORS) | Stable v2 · 2D-Spatial | UniDet: 0.1342 -> 0.1815 | Improved |
| fine-tune with reward-selected samples (GORS) | Stable v2 · Complex | 3-in-1: 0.3386 -> 0.3328 | No clear effect |
| add Attend-and-Excite | Stable v2 · Color | B-VQA: 0.5065 -> 0.6400 | Improved |
| add Composable Diffusion | Stable v2 · Color | B-VQA: 0.5065 -> 0.4063 | Hurt |
| add Structured Diffusion | Stable v2 · Color | B-VQA: 0.5065 -> 0.4990 | No clear effect |
| also fine-tune CLIP text encoder (not U-Net only) (ablation) | GORS · Color | B-VQA: 0.6216 -> 0.6570 | Improved |
| also fine-tune U-Net (not CLIP only) (ablation) | GORS · Color | B-VQA: 0.5507 -> 0.6570 | Improved |
| use full reward threshold instead of half (ablation) | GORS · Color | B-VQA: 0.6157 -> 0.6570 | Improved |
| use sample selection threshold instead of zero threshold (ablation) | GORS · Color | B-VQA: 0.6130 -> 0.6570 | Improved |
| scale training prompts from 25 to 1400 (ablation) | GORS · Complex | 3-in-1: 0.2596 -> 0.3504 | Improved |
Closest papers by meaning
- ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty (NeurIPS 2024 · similarity 0.91)
- T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-video Generation (CVPR 2025 · similarity 0.91)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.91)
- LLM Blueprint: Enabling Text-to-Image Generation with Complex and Detailed Prompts (ICLR 2024 · similarity 0.90)
- Progressive Compositionality in Text-to-Image Generative Models (ICLR 2025 · similarity 0.89)
- SILMM: Self-Improving Large Multimodal Models for Compositional Text-to-Image Generation (CVPR 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
