Research map / Large language models / Agents, code and math
Code2Worlds: Empowering Coding LLMs for 4D World Generation
Yi Zhang, Yunshuang Wang, Zeyu Zhang, Hao Tang
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 retrieval of reference scripts (ablation) | Code2Worlds object stream · Code4D object generation | SGS: 23.5 -> 61.4 | Improved |
| add structured procedural parameters library Lparam (ablation) | Code2Worlds object stream · Code4D object generation | SGS: 48.8 -> 61.4 | Improved |
| add object-level VLM-Critic self-reflection (ablation) | Code2Worlds object stream · Code4D object generation | SGS: 58.6 -> 61.4 | Improved |
| add VLM-Motion critic closed-loop refinement (ablation) | Code2Worlds dynamics · Code4D dynamics | physics failure rate: 60% -> 10% | Improved |
| add VLM-Motion critic closed-loop refinement (ablation) | Code2Worlds dynamics · Code4D dynamics | HRS: 47 -> 55.4 | Improved |
| add Scene Stream environmental orchestration (ablation) | Code2Worlds · Code4D scene generation | Richness: 26.4 -> 62.3 | Improved |
| add Planner and Solver agents (ablation) | Code2Worlds scene stream · Code4D scene generation | S-CLIP: 0.2251 -> 0.2432 | Improved |
| add Planner and Solver agents (ablation) | Code2Worlds scene stream · Code4D scene generation | Richness: 50.9 -> 62.3 | Improved |
| code-based 3D rendering instead of video diffusion | Code2Worlds vs video models · Code4D dynamics vs video models | Temporal Flickering: 0.9899 -> 0.9949 | Improved |
| code-based 3D rendering instead of AnimateDiff | Code2Worlds vs AnimateDiff · Code4D dynamics vs video models | Subject Consistency: 0.9778 -> 0.9415 | Hurt |
| Code2Worlds vs strongest code-based baseline | Code2Worlds · Code4D object generation | SGS: 43.5 -> 61.4 | Improved |
Closest papers by meaning
- WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis (NeurIPS 2026 · similarity 0.89)
- Code2World: A GUI World Model via Renderable Code Generation (NeurIPS 2026 · similarity 0.87)
- 4DWorldBench: A Comprehensive Evaluation Framework for 3D/4D World Generation Models (CVPR 2026 · similarity 0.86)
- SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning (NeurIPS 2026 · similarity 0.86)
- SceneCraft: An LLM Agent for Synthesizing 3D Scenes as Blender Code (ICML 2024 · similarity 0.85)
- WorldMem: Long-term Consistent World Simulation with Memory (NeurIPS 2025 · similarity 0.85)
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.
