WorldBench: Evaluating LLMs on Three.js Voxel World Generation
One prompt, one self-contained Three.js file per model. These are the five worlds graded in the paper, scored out of 280. Drag to orbit, scroll to zoom.
One prompt, one self-contained Three.js file per model. These are the five worlds graded in the paper, scored out of 280. Drag to orbit, scroll to zoom.