LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics (arxiv.org)

Seems like this signals Yann Lecun's direction now that he's leaving Meta

The EMA teacher model still seems like black magic to me (present in DINO and JEPA series but gone now)

A breath of fresh air from the over-concentration on LLM research. Something different for once.