Research map / Image generation / Text-to-image diffusion
CoMPaSS: Enhancing Spatial Understanding in Text-to-Image Diffusion Models
Gaoyang Zhang, Bingtao Fu, Qingnan Fan, Qi Zhang, Runxing Liu, Hong Gu, Huaqi Zhang, Xinguo 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 |
|---|---|---|---|
| add CoMPaSS (SCOP data + TENOR) | FLUX.1 · VISOR uncond | accuracy: 37.96 -> 75.17 | Improved |
| add CoMPaSS | FLUX.1 · T2I-CompBench Spatial | score: 0.18 -> 0.30 | Improved |
| add CoMPaSS | FLUX.1 · GenEval Position | score: 0.26 -> 0.60 | Improved |
| add CoMPaSS | SD1.5 · GenEval Position | score: 0.04 -> 0.54 | Improved |
| add CoMPaSS | FLUX.1 · DPG-Bench overall | score: 80.63 -> 84.42 | Improved |
| add CoMPaSS | SD2.1 · FID | FID: 21.65 -> 16.96 | Improved |
| add CoMPaSS | SD1.5 · FID | FID: 12.82 -> 10.89 | Hurt |
| add SCOP dataset fine-tuning (ablation) | SD1.5 · GenEval Position | score: 0.04 -> 0.39 | Improved |
| add SCOP dataset fine-tuning (ablation) | FLUX.1 · GenEval Position | score: 0.26 -> 0.56 | Improved |
| add TENOR on top of SCOP (ablation) | SD1.5 · GenEval Position | score: 0.39 -> 0.54 | Improved |
| add TENOR on top of SCOP (ablation) | FLUX.1 · GenEval Position | score: 0.56 -> 0.60 | Improved |
| add TENOR on top of SCOP (ablation) | FLUX.1 · T2I-CompBench Spatial | score: 0.29 -> 0.30 | Improved |
Closest papers by meaning
- Spatial Transport Optimization by Repositioning Attention Map for Training-Free Text-to-Image Synthesis (CVPR 2025 · similarity 0.90)
- Synergistic Dual Spatial-aware Generation of Image-to-text and Text-to-image (NeurIPS 2024 · similarity 0.89)
- Prompt-Free Diffusion: Taking "Text" out of Text-to-Image Diffusion Models (CVPR 2024 · similarity 0.88)
- Stitch and Tell: A Structured Data Augmentation Method for Spatial Understanding (NeurIPS 2025 · similarity 0.88)
- Magnet: We Never Know How Text-to-Image Diffusion Models Work, Until We Learn How Vision-Language Models Function (NeurIPS 2024 · similarity 0.88)
- SSMG: Spatial-Semantic Map Guided Diffusion Model for Free-Form Layout-to-Image Generation (AAAI 2024 · similarity 0.88)
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.
