Search papers, labs, and topics across Lattice.
1
0
3
4
Verification at test time can be a surprisingly effective alternative to scaling policy learning for vision-language-action alignment, yielding substantial gains in both simulated and real-world robotic tasks.
