Search papers, labs, and topics across Lattice.
Affiliation:
1
0
Predicting compact, learned future representations for world-action modeling slashes training costs by 65x and outperforms native visual targets, even with smaller models.
