Essentially *all* of my own research today is on non-sample-contrastive methods for JEPA.
That tells you what I think about them.

→ View original post on X — @ylecun