V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning
Original research
MODEL PROFILE · Enabling world model
V-JEPA 2
Learns predictive visual representations and supports action-conditioned robot planning.
Research organisationMeta AI / FAIR
Research approachRepresentation-only
Published year2025
BackboneJoint-embedding predictive architecture
Overview
V-JEPA 2 is a self-supervised video representation system. Its action-conditioned variant V-JEPA 2-AC supports visual-goal robot planning; the broader architecture is enabling predictive technology, not a generalised WAM robot policy.
Source record
Core research identity and broad method were source-checked as of 2026-11-03. Independent benchmarking, commercial availability, weights and licensing have not been assessed.