Tian Xia, Tianrun Gao, Wenhao Deng, Long Wei, Xiaowei Qian, Chenglei Yu, Tailin Wu$^{†}$
ICML 2026
We introduce BuildArena that studies whether LLM agents can design and build functional machines in a physics simulator. Given natural language goals, LLM agents construct rockets, bridges, vehicles, and other 3D mechanisms from scratch.