AMI presented H-JEPA as the first end-to-end learned hierarchical world model for this application. The architecture is built atop SIGReg/LeWM and employs a stable training recipe that allows semantic abstractions to emerge naturally across its levels.
Tuesday, October 20, 2026
Story
AMI Launches First Hierarchical World Model for Long Horizon Visual Planning
AI15 days ago
The H-JEPA model improved success on the Visual AntMaze benchmark from 18% to 73% while decreasing the amount of planning compute required. The system learns world models at various levels of abstraction and performs planning by moving from coarse goals down to finer actions to handle long horizon visual tasks.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.