1 paper
Ran Dan, Si-Tong Wei, Pengfei Xiong +3
Existing 3D large language models (LLMs) compromise on two fronts: they compress shapes into latent codebook indices or coordinate text, which removes spatial structure from what t…