Search papers, labs, and topics across Lattice.
University of Maryland, University of Southern California
1
0
2
Pretrained VLA models can be transformed into safer navigators, reducing near-collisions by over 86% through better alignment of their internal representations with action outputs.
