WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis
📰 ArXiv cs.AI
arXiv:2606.01869v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly asked not only to write static interfaces, but to construct executable interactive worlds from natural language. Browser-native 3D, commonly built with Three.js, is a natural next frontier: generated programs must integrate assets, obey spatial and physical constraints, and keep user-facing controls synchronized with hidden runtime state. Existing web-generation benchmarks and evaluators, however, large
Full Article
Title: WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis
Abstract:
arXiv:2606.01869v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly asked not only to write static interfaces, but to construct executable interactive worlds from natural language. Browser-native 3D, commonly built with Three.js, is a natural next frontier: generated programs must integrate assets, obey spatial and physical constraints, and keep user-facing controls synchronized with hidden runtime state. Existing web-generation benchmarks and evaluators, however, large
Abstract:
arXiv:2606.01869v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly asked not only to write static interfaces, but to construct executable interactive worlds from natural language. Browser-native 3D, commonly built with Three.js, is a natural next frontier: generated programs must integrate assets, obey spatial and physical constraints, and keep user-facing controls synchronized with hidden runtime state. Existing web-generation benchmarks and evaluators, however, large
DeepCamp AI