Research map / 3D vision / 3D generation
SceneSmith: Agentic Generation of Simulation-Ready Indoor Scenes
Nicholas Pfaff, Thomas Cohn, Sergey Zakharov, Rick Cory, Russ Tedrake
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 generated text-to-3D assets instead of HSSD retrieval (ablation) | SceneSmith · User study | realism win rate: 63.8% | Improved |
| add asset validation (ablation) | SceneSmith · User study | realism win rate: 63.0% | Improved |
| add visual scene observation (ablation) | SceneSmith · User study | realism win rate: 61.5% | Improved |
| add specialized furniture and manipuland tools (ablation) | SceneSmith · User study | realism win rate: 54.8% | No clear effect |
| add agent session memory (ablation) | SceneSmith · User study | realism win rate: 53.4% | No clear effect |
| add critic agent (ablation) | SceneSmith · User study | realism win rate: 51.8% | No clear effect |
| add critic agent (ablation) | SceneSmith · object count | objects per room | Improved |
| SceneSmith vs strongest baseline HSM | SceneSmith · User study | realism win rate: 88.5 | Improved |
| SceneSmith vs baselines on physics | SceneSmith · SceneEval | collision rate: 1.2±0.6 vs 3.0±1.5 (I-Design) to 28.9±3.5 | Improved |
| SceneSmith vs baselines on stability | SceneSmith · SceneEval | STB: 95.6±1.7 vs 60.8±4.6 (I-Design) | Improved |
| higher object density | SceneSmith · SceneEval | ACC and NAV: ACC 83.4±1.5 vs HSM 88.9±1.7 | Hurt |
| use CoACD instead of V-HACD for collision geometry (ablation) | SceneSmith · qualitative | collision quality | Hurt |
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 7 of 7 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
- SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent (NeurIPS 2025 · similarity 0.90)
- InternScenes: A Large-scale Simulatable Indoor Scene Dataset with Realistic Layouts (NeurIPS 2025 · similarity 0.87)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (NeurIPS 2025 · similarity 0.87)
- PhyScene: Physically Interactable 3D Scene Synthesis for Embodied AI (CVPR 2024 · similarity 0.87)
- SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning (NeurIPS 2026 · similarity 0.87)
- Holodeck: Language Guided Generation of 3D Embodied AI Environments (CVPR 2024 · 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.
