Fast-JEPA
Fast-JEPAPaper availableOpening announced; not available yet
Participating organizationsHarbin Institute of Technology
Participating organizationsZhongguancun Academy
Participating organizationsEast China Normal University
Participating organizationsHuazhong University of Science and Technology
Participating organizationsWuhan University
Participating organizationsNanjing University
Participating organizationsZhongguancun Institute of Artificial Intelligence
Participating organizationsDeepCyboVersion scope and roles for these associations await verification.
Fast-JEPA is a latent world-model planner for visual goals. It builds a reference action plan from the current observation and goal image, then compares complete candidates by predicting their endpoints directly. The authors report 96.50% mean success across four simulation tasks.
Paper available
Author order: Original order
Original Fast-JEPA research implementation and evaluation
DeepCybo-PhysAI · Project page
Publication date unknown · Accessed: 2026-11-22
Fast-JEPA authors · Project page
Publication date unknown · Accessed: 2026-11-22