Research map / Large language models / Agents, code and math
HouseTune: Two-Stage Floorplan Generation with LLM Assistance
Ziyang Zong, Guanying Chen, Zhaohuan Zhan, Fengcheng Yu, Guang Tan
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 diffusion refinement stage after LLM layout (two-stage vs one-stage) (ablation) | HouseTune · RPLAN 6 rooms | Diversity (FID): 109.5 -> 7.5 | Improved |
| add diffusion refinement stage (ablation) | HouseTune · RPLAN 6 rooms | Compatibility: 10.6 -> 0.25 | Improved |
| replace task-only prompt with predefined CoT prompt (ablation) | HouseTune Layout-Init · RPLAN 6 rooms | IoU: 12.36% -> 21.41% | Improved |
| add 'think step by step' to prompt vs common sense prompt (ablation) | HouseTune Layout-Init · RPLAN 6 rooms | IoU: 14.57% -> 20.69% | Improved |
| predefined CoT vs basic CoT prompt (ablation) | HouseTune Layout-Init · RPLAN 6 rooms | IoU: 20.69% -> 21.41% | Improved |
| swap Layout-Init LLM among DeepSeek-R1, DouBao-1.5Pro, GPT-4o (ablation) | HouseTune · RPLAN 6 rooms | IoU: 34.3% / 30.2% / 33.8% | No clear effect |
| use DNPP-Diffusion instead of conditional diffusion (ablation) | HouseTune · RPLAN 6 rooms | Mic IoU: 21.84% -> 34.15% | Improved |
| use DNPP-Diffusion instead of conditional diffusion (ablation) | HouseTune · RPLAN 6 rooms | Mac IoU: 17.75% -> 32.83% | Improved |
| strong-to-weak prior preserving vs fixed (ablation) | HouseTune · RPLAN 6 rooms | Mic IoU: 28.03 -> 34.15 | Improved |
| strong-to-weak prior preserving vs weak-to-strong (ablation) | HouseTune · RPLAN 6 rooms | Mic IoU: 15.94 -> 34.15 | Improved |
| condition in both forward and reverse processes vs single (ablation) | HouseTune · RPLAN 6 rooms | Diversity (FID): 14.17 -> 7.47 | Improved |
| increase conditional participation ratio from 1e-3 to 1e-1 (ablation) | HouseTune · RPLAN 6 rooms | Mac IoU: 28.67% -> 31.37% | Mixed |
Closest papers by meaning
- Unified Vector Floorplan Generation via Markup Representation (CVPR 2026 · similarity 0.91)
- Tokenization Allows Multimodal Large Language Models to Understand, Generate and Edit Architectural Floor Plans (CVPR 2026 · similarity 0.88)
- GSDiff: Synthesizing Vector Floorplans via Geometry-enhanced Structural Graph Generation (AAAI 2025 · similarity 0.87)
- OptiScene: LLM-driven Indoor Scene Layout Generation via Scaled Human-aligned Data Synthesis and Multi-Stage Preference Optimization (NeurIPS 2025 · similarity 0.87)
- MaskPLAN: Masked Generative Layout Planning from Partial Input (CVPR 2024 · similarity 0.86)
- Constrained Layout Generation with Factor Graphs (CVPR 2024 · similarity 0.85)
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.
