Research map / 3D vision / 3D generation
MANSION: Multi-floor lANguage-to-3D Scene generatIOn for loNg-horizon tasks
Lirong Che, Shuo Wen, Shan Huang, Chuang Wang, Yuzhe Yang, Gregory Dudek, Xueqian Wang, Jian Su
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 hierarchical (iterative) splitting of room seeds (ablation) | MANSION floorplan solver · ResPlan-1k | micro-IoU: 45.65 -> 63.56 | Improved |
| use stronger MLLM (Gemini-2.5-Pro instead of Moonshot-v1-8k) (ablation) | MANSION · T2D | micro-IoU: 42.33 -> 69.98 | Improved |
| constrained growth vs CHD diffusion under manual annotation | MANSION · T2D | micro-IoU: 82.81 -> 81.67 | No clear effect |
| MANSION vs CHD on more complex ResPlan (zero-shot for CHD) | MANSION · ResPlan-1k | micro-IoU: 33.49 -> 76.74 | Improved |
| MANSION vs CHD with Moonshot-v1-8k | MANSION · T2D | micro-IoU: 60.09 -> 42.33 | Hurt |
| reachability-constrained anchor-based object placement vs Holodeck/LayoutGPT | MANSION · Object placement | reachability (%): 88.7 -> 100.0 (Bedroom, vs Holodeck) | Improved |
| add object-type information to BUMBLE (ablation) | BUMBLE · MANSION single-floor tasks | task success rate (%): 40 -> 60 | Improved |
| increase building to four floors | COME and BUMBLE variants · MANSION tasks | task success rate (%): 60 -> 0 | Hurt |
| MANSION classroom layouts vs Holodeck on diversity | MANSION · Classroom user study | diversity preference (%): 51.0 -> 45.1 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Holodeck — Holodeck: Language Guided Generation of 3D Embodied AI Environments (CVPR 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Tokenization Allows Multimodal Large Language Models to Understand, Generate and Edit Architectural Floor Plans (CVPR 2026 · similarity 0.89)
- SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent (NeurIPS 2025 · similarity 0.89)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.87)
- 3D-GRAND: A Million-Scale Dataset for 3D-LLMs with Better Grounding and Less Hallucination (CVPR 2025 · similarity 0.87)
- Hierarchically-Structured Open-Vocabulary Indoor Scene Synthesis with Pre-trained Large Language Model (AAAI 2025 · similarity 0.87)
- PARTNR: A Benchmark for Planning and Reasoning in Embodied Multi-agent Tasks (ICLR 2025 · similarity 0.86)
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.
