SplitJEPA Separates Invariant and Variable Factors Without Reconstruction

SplitJEPA, a new paper by Helios Hua et al., combines JEPA-style predictive learning with paired observations differing only in variable factors, resolving block ambiguity in latent representations. It identifies invariant-variable structure without requiring reconstruction of observations.

2026-10-11 ~ 2026-10-11 · 2 related posts

• SplitJEPA: Disentangling Invariant-Variant Structure Without Reconstruction — udmrzn · 2026-10-11
• SplitJEPA paper separates invariant and variant factors in JEPA latent states — mayfer · 2026-10-11