WorldBench: Evaluating LLMs on Three.js Voxel World Generation

Key points

• Large language models can now write complete, interactive 3D worlds as code, but grading those worlds automatically is unreliable.1
• On worlds written by five frontier models we find that the two views disagree on 32% of required items, mostly code that no frame shows, and that fixed views miss small close-up contents.1
• From one prompt describing a floating voxel island with ten biomes, physics, and day/night and seasonal cycles, the judge explores the running world, controlling its clock, orbiting it, and sending a navigator agent to frame each biome, and reads the code for what it sees.1
• We evaluate five frontier models: Claude Fable 5.1, GPT-6 Astra, Kimi K3, Grok 4.7 and Gemini 3.1 Pro.1

Sources (1)

• [1]WorldBench: Evaluating LLMs on Three.js Voxel World Generation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 08:52 AM

  We present WorldBench, a benchmark and judge for open-ended, LLM-generated Three.js worlds.

  Large language models can now write complete, interactive 3D worlds as code, but grading those worlds automatically is unreliable.

Extractive summary: sentences quoted from the sources.