Research map / Image generation / Text-to-image diffusion
Compositional Text-to-Image Generation Via Region-aware Bimodal Direct Preference Optimization
Zhuohan Liu, Wujian Peng, Yitong Chen, Zuxuan Wu
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 |
|---|---|---|---|
| add region-level guidance to BiDPO (ablation) | SDXL-BiDPO · T2I-CompBench | overall score: 53.10 -> 54.37 | Improved |
| add region-level guidance to BiDPO (ablation) | SDXL-BiDPO · GenEval | overall score: 60.71 -> 62.14 | Improved |
| add region-level guidance to BiDPO (ablation) | SDXL-BiDPO · DPG-Bench | overall score: 77.53 -> 78.84 | Improved |
| use bimodal DPO instead of image-only DPO (ablation) | SDXL · T2I-CompBench | overall score: 45.58 -> 53.10 | Improved |
| use bimodal DPO instead of image-only DPO (ablation) | SDXL · GenEval | overall score: 53.00 -> 60.71 | Improved |
| use text-only DPO instead of bimodal DPO (ablation) | SDXL · T2I-CompBench | overall score: 53.10 -> 13.48 | Hurt |
| use text-only DPO instead of bimodal DPO (ablation) | SDXL · GenEval | overall score: 60.71 -> 4.71 | Hurt |
| fine-tune with supervised fine-tuning on the composition dataset instead of DPO (ablation) | SDXL · T2I-CompBench | overall score: 43.57 -> 43.34 | No clear effect |
| fine-tune with supervised fine-tuning on the composition dataset instead of DPO (ablation) | SDXL · DPG-Bench | overall score: 73.38 -> 73.23 | No clear effect |
| apply image-only DPO to SDXL (ablation) | SDXL · DPG-Bench | overall score: 73.38 -> 75.70 | Improved |
| apply image-only DPO to SDXL (ablation) | SDXL · T2I-CompBench Numeracy | numeracy score: 50.08 -> 39.34 | Hurt |
| apply BiDPO to SDXL | SDXL · T2I-CompBench Color | color score: 58.90 -> 79.35 | Improved |
Closest papers by meaning
- SILMM: Self-Improving Large Multimodal Models for Compositional Text-to-Image Generation (CVPR 2025 · similarity 0.90)
- Multimodal Large Language Models Make Text-to-Image Generative Models Align Better (NeurIPS 2024 · similarity 0.89)
- Improving Long-Text Alignment for Text-to-Image Diffusion Models (ICLR 2025 · similarity 0.89)
- Mastering Text-to-Image Diffusion: Recaptioning, Planning, and Generating with Multimodal LLMs (ICML 2024 · similarity 0.89)
- Discriminative Probing and Tuning for Text-to-Image Generation (CVPR 2024 · similarity 0.89)
- ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty (NeurIPS 2024 · 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.
