Announcement
H-JEPA introduced as a hierarchical world model for long-horizon visual planning
The announcement says H-JEPA builds on SIGReg/LeWM and describes semantic abstractions emerging across levels during training.
TLDR
A post introducing H-JEPA calls it the first end-to-end learned hierarchical world model for long-horizon visual planning. It says H-JEPA builds on SIGReg/LeWM and offers a stable training recipe in which semantic abstractions naturally emerge across levels. The post links to a paper and code.
Combined views
18.2K
2 Sources, first seen ago
