Search papers, labs, and topics across Lattice.
1
0
3
1
Injecting 3D scene understanding into VLAs without extra labels yields surprisingly large gains in robotic manipulation, suggesting a critical missing piece in current vision-language-action models.
