Research map / 3D vision / 3D generation
NaLA: A 3D Native LLM Layout Agent for High-quality 3D Scene Generation
Cheng Wan, Yongsen Mao, Wenzheng Wu, Yuxuan Xie, Chucheng Xiang, Runze Wang, Xiang Zhang, zhongyuan liu, Rushi Dai, Yuan 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 |
|---|---|---|---|
| encode asset and room point clouds natively into the LLM instead of text/boxes (ablation) | NaLA · layout generation | avg AI score: 2.82 -> 3.53 (text-only vs full, standard input) | Improved |
| add point cloud training, evaluated with text-only input at test (ablation) | NaLA · layout generation | avg AI score: 3.21 (text-only trained) vs 3.22 (full) | No clear effect |
| use coarse-to-fine anchoring + residual tokens instead of text coordinates (ablation) | NaLA · layout generation | avg AI score: 3.12 -> 3.53 | Improved |
| add pose anchoring tokens to residual-only output (ablation) | NaLA · layout generation | collision ratio (%): 19.19 -> 0.86 | Improved |
| add pose residual tokens to anchoring-only output (ablation) | NaLA · layout generation | avg AI score: 3.30 -> 3.53 | Improved |
| use grid B=20 instead of 5 (ablation) | NaLA · layout generation | avg AI score: 2.58 -> 3.53 | Improved |
| use grid B=20 instead of 100 (ablation) | NaLA · layout generation | avg AI score: 2.76 -> 3.53 | Improved |
| add input asset sequence shuffling (ablation) | NaLA · layout generation | avg AI score: 2.46 -> 3.53 | Improved |
| add asset replacement augmentation (ablation) | NaLA · layout generation | avg AI score: 2.60 -> 3.53 | Improved |
| add scene rotation augmentation (ablation) | NaLA · layout generation | avg AI score: 2.84 -> 3.53 | Improved |
| use two-stage training (3D-FRONT then Imaginarium) instead of 3D-FRONT only (ablation) | NaLA · layout generation | avg AI score: 2.90 -> 3.53 | Improved |
| use two-stage training instead of joint mixed training (ablation) | NaLA · layout generation | avg AI score: 3.31 -> 3.53 | Improved |
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 3 of 5 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
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.89)
- LayoutVLM: Differentiable Optimization of 3D Layout via Vision-Language Models (CVPR 2025 · similarity 0.88)
- Let Language Constrain Geometry: Vision–Language Models as Semantic and Spatial Critics for 3D Generation (ICML 2026 · similarity 0.87)
- Layout-your-3D: Controllable and Precise 3D Generation with 2D Blueprint (ICLR 2025 · similarity 0.87)
- Disentangled 3D Scene Generation with Layout Learning (ICML 2024 · similarity 0.87)
- Hierarchically-Structured Open-Vocabulary Indoor Scene Synthesis with Pre-trained Large Language Model (AAAI 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.
