Hierarchical World Models with Implicit Dynamics
arXiv cs.AI · 31 Oct 2026, 03:59 AEST
The introduction of ID-WM (Implicit Dynamics World Models) offers a new approach to planning with world models, enhancing generalization for embodied agents. This hierarchical model improves performance in long-horizon tasks by decoupling state-space reachability from local control.
Key facts
- Planning with world models enables test-time adaptation for embodied agents, offering superior generalization over direct policy learning.
- Existing hierarchical methods often use rigid macro-actions or latent skills, leading to brittle representations and limited fine-grained control.
- Our framework utilizes a high-level model to sample feasible transitions across the state manifold and a low-level world model as an inverse dynamics solver to ground these transitions into precise actions.
- By prioritizing local physical feasibility over global behavioral patterns, ID-WM synthesizes novel long-horizon plans from short segments, transcending the specific behavioral modes seen during training.
- We demonstrate its effectiveness across three navigation and manipulation suites, achieving a 38.5% improvement over state-of-the-art policy and diffusion-based planners.
Read aloud
The introduction of ID-WM Implicit Dynamics World Models offers a new approach to planning with world models, enhancing generalization for embodied agents. This hierarchical model improves performance in long-horizon tasks by decoupling state-space reachability from local control. Planning with world models enables test-time adaptation for embodied agents, offering superior generalization over direct policy learning. Existing hierarchical methods often use rigid macro-actions or latent skills, leading to brittle representations and limited fine-grained control. Our framework utilizes a high-level model to sample feasible transitions across the state manifold and a low-level world model as an inverse dynamics solver to ground these transitions into precise actions. By prioritizing local physical feasibility over global behavioral patterns, ID-WM synthesizes novel long-horizon plans from short segments, transcending the specific behavioral modes seen during training. We demonstrate its effectiveness across three navigation and manipulation suites, achieving a thirty eight point five percent improvement over state-of-the-art policy and diffusion-based planners.
Sources in this cluster
- Hierarchical World Models with Implicit Dynamics
arXiv cs.AI · 31 Oct 2026, 03:59 AEST