Research map / Image generation / Text-to-image diffusion
LayoutGPT: Compositional Visual Planning and Generation with Large Language Models
Weixi Feng, Wanrong Zhu, Tsu-Jui Fu, Varun Jampani, Arjun Reddy Akula, Xuehai He, S Basu, Xin Eric Wang, William Yang Wang
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 |
|---|---|---|---|
| Represent layouts as CSS-style structured programs rather than comma-separated values (ablation) | LayoutGPT · NSR-1K spatial reasoning prompts | layout accuracy: 55.12 -> 80.82 | Improved |
| Add task instructions (ablation) | LayoutGPT · NSR-1K spatial reasoning prompts | layout accuracy: 55.12 -> 78.23 | Improved |
| Add normalization when used alone (ablation) | LayoutGPT · NSR-1K spatial reasoning prompts | layout accuracy: 55.12 -> 44.10 | Hurt |
| Add normalization to task instructions and CSS formatting (ablation) | LayoutGPT · NSR-1K spatial reasoning prompts | layout accuracy: 81.84 -> 82.54 | Improved |
| Add normalization to the 3D layout prompt (ablation) | LayoutGPT · 3D-FRONT bedroom test | out-of-bounds rate: 55.32 -> 46.57 | Improved |
| Add CSS structure to the 3D layout prompt (ablation) | LayoutGPT · 3D-FRONT bedroom test | out-of-bounds rate: 55.32 -> 51.77 | Improved |
| Use retrieved exemplars similar to the test room's floor plan instead of random bedroom exemplars (ablation) | LayoutGPT · 3D-FRONT bedroom test | out-of-bounds rate: 85.58 -> 43.26 | Improved |
| Generate layouts with LayoutGPT instead of LayoutTransformer | 2D layout planning · NSR-1K numerical reasoning | layout accuracy: 22.26 -> 86.33 | Improved |
| Use a layout-planning stage with LayoutGPT rather than direct Stable Diffusion text-to-image generation | Numerical image generation · NSR-1K numerical reasoning | GLIP image accuracy: 42.44 -> 55.64 | Improved |
| Use LayoutGPT layouts rather than supervised ATISS layouts | 3D indoor scene synthesis · 3D-FRONT bedroom test | KL divergence: 0.0113 -> 0.0995 | Hurt |
| Use LayoutGPT instead of ATISS | 3D indoor scene synthesis · 3D-FRONT living room test | FID: 85.40 -> 76.34 | Improved |
| Use LayoutGPT + ReCo rather than Attend-and-Excite | Color attribute binding · HRS-Bench color prompts, overall | attribute binding accuracy: 22.96 -> 36.96 | Improved |
Closest papers by meaning
- Think, Plan, Paint: Layout-Aware Reasoning for Controllable Image Generation in Unified Models (NeurIPS 2026 · similarity 0.90)
- Rethinking The Training And Evaluation of Rich-Context Layout-to-Image Generation (NeurIPS 2024 · similarity 0.89)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.89)
- Seeing is Improving: Visual Feedback for Iterative Text Layout Refinement (CVPR 2026 · similarity 0.88)
- Layout-your-3D: Controllable and Precise 3D Generation with 2D Blueprint (ICLR 2025 · similarity 0.88)
- SceneCraft: Layout-Guided 3D Scene Generation (NeurIPS 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.
