Research map / 3D vision / 3D generation
SceneOrchestra: Efficient Agentic 3D Scene Synthesis via Full Tool-Call Trajectory Generation
Yun He, Kelin Yu, Matthias Zwicker
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 DPO training after SFT (ablation) | orchestrator · Training-strategy ablation, template prompts (average) | #Obj: 24.9 -> 31.1 | Improved |
| add stepwise training (ablation) | orchestrator · Training-strategy ablation, template prompts (average) | #Obj: 26.2 -> 31.1 | Improved |
| add discriminator (ablation) | orchestrator · Training-strategy ablation, template prompts (average) | #Obj: 27.6 -> 31.1 | Improved |
| add interleaved training (ablation) | orchestrator and discriminator · Training-strategy ablation, template prompts (average) | #Obj: 29.7 -> 31.1 | Improved |
| add interleaved training (ablation) | orchestrator and discriminator · Training-strategy ablation, template prompts (average) | Completeness: 7.8 -> 8.1 | Improved |
| add DPO training after SFT (ablation) | orchestrator · Training-strategy ablation, template prompts (average) | Realism: 8.0 -> 9.1 | Improved |
| use fine-tuned orchestrator instead of GPT-5.5 in-context (ablation) | orchestrator · Off-the-shelf orchestrator ICL, template prompts (average) | #Obj: 24.2 -> 31.1 | Improved |
| use fine-tuned orchestrator instead of GPT-5.5 in-context (ablation) | orchestrator · Off-the-shelf orchestrator ICL, template prompts (average) | time (min): 45.3 -> 39.7 | Improved |
| use fine-tuned orchestrator instead of untuned Qwen3-4B (ablation) | orchestrator · Off-the-shelf orchestrator ICL, template prompts (average) | #Obj: 18.2 -> 31.1 | Improved |
| use fine-tuned trajectory orchestrator instead of GPT-5.5 as SceneWeaver step-wise orchestrator | orchestrator · SceneWeaver orchestrator replacement | time (min): 82.6 -> 39.7 | Improved |
| replace SceneWeaver's step-by-step orchestration with full trajectory generation | SceneWeaver · 10 room types, template prompts (average) | time (min): 91.0 -> 39.7 | Improved |
| replace SceneWeaver with SceneOrchestra on complex instructions | SceneWeaver · Complex instructions (70, unseen rooms) | Functionality: 7.6 -> 8.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SceneWeaver — SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent (NeurIPS 2025) · better on 13 of 13 shared results
- Holodeck — Holodeck: Language Guided Generation of 3D Embodied AI Environments (CVPR 2024) · better on 2 of 3 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 0.92)
- SceneCraft: An LLM Agent for Synthesizing 3D Scenes as Blender Code (ICML 2024 · similarity 0.89)
- 3DCodeBench: Benchmarking Agentic Procedural 3D Modeling Via Code (NeurIPS 2026 · similarity 0.89)
- WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis (NeurIPS 2026 · similarity 0.87)
- Muses: 3D-Controllable Image Generation via Multi-Modal Agent Collaboration (AAAI 2025 · similarity 0.87)
- InfiniBench: Infinite Benchmarking for Visual Spatial Reasoning with Customizable Scene Complexity (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.
