Announcement
3JSBench is introduced to evaluate LLMs on Three.js asset creation
A post introducing the benchmark says frontier models still struggle to build complex 3D objects.
TLDR
A post introducing 3JSBench describes it as the first open-source benchmark for evaluating LLMs’ ability to generate coherent Three.js assets. Its author says frontier models still struggle to build complex 3D objects and asks how to measure what they get wrong.
Combined views
4.9K
2 Sources, first seen ago
