Search papers, labs, and topics across Lattice.
1
0
4
3
VLA models struggle not with *what* they see, but with *how* they see it: FocusVLA reveals that directing attention to task-relevant visual regions unlocks significant performance gains in robotic manipulation.
