# WorldBench: Evaluating LLMs on Three.js Voxel World Generation

arXiv: https://arxiv.org/abs/2610.10622 · HTML: https://mlsanity.com/paper/2610.10622/

## Key facts

- No citing papers in the map yet (data through 7 Oct 2026).
- Builds on 11 of its 12 references that are in the map.
- First posted on arXiv 7 Oct 2026.

## Builds on (most cited first, 11 of 11)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Teaching Large Language Models to Self-Debug](https://mlsanity.com/paper/2304.05128/) (2023)
- [LLM Evaluators Recognize and Favor Their Own Generations](https://mlsanity.com/paper/2404.13076/) (2024)
- [SceneCraft: An LLM Agent for Synthesizing 3D Scene as Blender Code](https://mlsanity.com/paper/2403.01248/) (2024)
- [T3Bench: Benchmarking Current Progress in Text-to-3D Generation](https://mlsanity.com/paper/2310.02977/) (2023)
- [SceneEval: Evaluating Semantic Coherence in Text-Conditioned 3D Indoor Scene Synthesis](https://mlsanity.com/paper/2503.14756/) (2025)
- [GT23D-Bench: A Comprehensive General Text-to-3D Generation Benchmark](https://mlsanity.com/paper/2412.09997/) (2024)
- [Text-to-Scene with Large Reasoning Models](https://mlsanity.com/paper/2509.26091/) (2025)
- [WorldCoder-Bench: Benchmarking Physically Grounded 3D World Synthesis](https://mlsanity.com/paper/2606.01869/) (2026)
- [SceneOrchestra: Efficient Agentic 3D Scene Synthesis via Full Tool-Call Trajectory Generation](https://mlsanity.com/paper/2604.19907/) (2026)
- [SceneX: Procedural Controllable Large-scale Scene Generation](https://mlsanity.com/paper/2403.15698/) (2024)

## Abstract (arXiv)

Large language models can now write complete, interactive 3D worlds as code, but grading those worlds automatically is unreliable. Existing judges take one view of the output: a vision-language model scores a few rendered snapshots, or a language model reads the source. 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. We present WorldBench, a benchmark and judge for open-ended, LLM-generated Three.js worlds. 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. Neither channel is trusted on its own: a code quote counts only if it is text the source contains, and visual claims are checked against measured pixels where the property is measurable. A mutation test, in which we remove features by construction, shows that code-only judging gives full credit to four of five removed features, because their code remains in the file. Our judge cuts the points kept on removed features by a third (5.44 to 3.55 of 7.11), and what it still credits is mostly code that exists but never runs. We evaluate five frontier models: Claude Fable 5.1, GPT-6 Astra, Kimi K3, Grok 4.7 and Gemini 3.1 Pro. Code, prompt, tests and judge configuration are available at https://github.com/KrishBakshi/worldbench