Search papers, labs, and topics across Lattice.
1
0
3
2
LeVJEPA achieves up to 20.8x less pretraining compute while surpassing the performance of leading video representation methods, reshaping the landscape of video-based learning.