Search papers, labs, and topics across Lattice.
1
0
3
Text-only foundation models can perform surprisingly well on complex 3D spatial reasoning tasks, rivaling multimodal models, when equipped with a structured spatial representation derived from 3D reconstruction.
