Research map / 3D vision / 3D generation
Text-to-Scene with Large Reasoning Models
Frédéric Berdoz, Luca A Lanzendörfer, Nick Tuninga, Roger Wattenhofer
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 |
|---|---|---|---|
| use Reason-3D instead of Holodeck for retrieval | object retrieval · Retrieval benchmark | top-1 accuracy: 7% -> 75% | Improved |
| use Reason-3D instead of Holodeck | scene synthesis · Human eval full scene | win-rate: 4.8 -> 95.2 | Improved |
| use Reason-3D instead of LayoutVLM at complexity 1 | object placement · Human eval placement | rating (1-5): 2.8 -> 4.4 | Improved |
| use Reason-3D instead of LayoutVLM at complexity 5 | object placement · Human eval placement | rating (1-5): 2.4 -> 4.3 | Improved |
| raise instruction complexity from 1 to 5 | LayoutVLM · Human eval placement | rating (1-5): 2.8 -> 2.4 | Hurt |
| raise instruction complexity from 1 to 5 | Reason-3D · Human eval placement | rating (1-5): 4.4 -> 4.3 | No clear effect |
| use Gemini 2.5 Pro instead of GPT-4.1 (no reasoning) | Reason-3D LRM · LRM benchmark | Elo: 1500 -> 2091 | Improved |
| use reasoning model GPT-o3 instead of GPT-4.1 | Reason-3D LRM · LRM benchmark | Elo: 1500 -> 1938 | Improved |
| use Sonnet 4 instead of GPT-4.1 | Reason-3D LRM · LRM benchmark | Elo: 1500 -> 1566 | Improved |
| use DeepSeek-R1 instead of Sonnet 4 | Reason-3D LRM · LRM benchmark | Elo: 1566 -> 1809 | 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 4 of 4 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
- 3D Space as a Scratchpad for Editable Text-to-Image Generation (CVPR 2026 · similarity 0.88)
- MLLM-For3D: Adapting Multimodal Large Language Model for 3D Reasoning Segmentation (NeurIPS 2025 · similarity 0.88)
- FirePlace: Geometric Refinements of LLM Common Sense Reasoning for 3D Object Placement (CVPR 2025 · similarity 0.88)
- Video Spatial Reasoning with Object-Centric 3D Rollout (AAAI 2026 · similarity 0.87)
- Hypo3D: Exploring Hypothetical Reasoning in 3D (ICML 2025 · similarity 0.87)
- ReasonMap: Towards Fine-Grained Visual Reasoning from Transit Maps (CVPR 2026 · similarity 0.87)
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.
