Randall Balestriero(@randall_balestr):LeWM applied SIGReg per time-step, so nothing prevents temporal collapse. If your data has such slow features, simply apply SIGReg on the (N*T, D) tensor. We showed this solves your problem here: https://t.co/FNFujpuFnO And @vlad_is_ai already spoked about JEPA+Slow features 4 years ago here

2026.10.01 18:30

LeWM applied SIGReg per time-step, so nothing prevents temporal collapse. If your data has such slow features, simply apply SIGReg on the (N*T, D) tensor. We showed this solves your problem here: And @vlad_is_ai already spoked about JEPA+Slow features 4 years ago here

Hugues Van Assel@hugues_va

2026.10.01 16:58

JEPA has a huge blind spot no one talks about: it’s starving for stable, slow features. Add a SINGLE irrelevant patch, keep it fixed within each episode, and planning falls apart. Make it unpredictable, and performance almost returns to normal. Wild! More in a blog post soon.

190

轉發到社區