H-JEPA: End-to-End Learning of Hierarchical World Models for Visual Planning

2026-10-05

AI Summary

This study addresses the challenge of reasoning across temporal scales and abstraction levels in long-horizon visual planning. To this end, we propose an end-to-end hierarchical action-conditioned Joint Embedding Predictive Architecture (JEPA). By temporally decomposing fast and slow dynamics, our method constructs a hierarchical latent space model wherein a high-level predictor forecasts distant future states and generates subgoals for top-down planning, while a low-level module leverages inverse dynamics supervision to optimize multi-scale execution efficiency. Experimental results demonstrate that the proposed architecture significantly improves task success rates from 18% to 73% in simulated environments, effectively reduces computational overhead, and enhances planning fidelity when validated on real-world robot videos.

Abstract

Long-horizon planning with latent world models requires reasoning across timescales and levels of abstraction. Existing task-agnostic JEPA world models predict and plan at a single timescale or with multiple horizons in one shared latent space. We introduce H-JEPA, an end-to-end recipe for training a hierarchy of action-conditioned JEPAs in which each level predicts farther ahead in its own learned latent space. Planning proceeds top-down: the top level optimizes progress toward the goal, and each level's predictions become subgoals for the planner below it. When factors in the data evolve at separated timescales, higher levels discard fast, unpredictable detail and retain slower task-relevant state. Across four simulated navigation and manipulation environments, hierarchical planning improves over a flat JEPA; on Visual AntMaze, a three-level hierarchy raises success from 18% to 73% using less planner compute. Ablations attribute these gains to both temporal decomposition and higher-level goal representations. With inverse-dynamics supervision, the approach extends to diverse real-robot videos from DROID, where hierarchy improves offline planning fidelity at lower planner compute.

Problem

• long-horizon planning
• latent world models
• hierarchical abstraction
• JEPA
• visual planning

Innovation

• Hierarchical World Models
• JEPA
• Visual Planning
• Temporal Abstraction
• End-to-End Learning