Search papers, labs, and topics across Lattice.
1
0
2
By refining VLM-derived reward signals with structural priors, SAFT transforms noisy feedback into a reliable guide for faster and more aligned policy learning.
