Multi-scale Mixture of World Models for Embodied Agents in Evolving Environments
2026-07-01 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors address how robots or agents that act in the real world can better understand and adapt to different situations that happen at different scales or levels. They point out two problems with current methods that pick which part of the agent's knowledge to use: these methods don't clearly recognize scale and treat all updates the same way, even though some knowledge changes faster than others. Their new method, MuSix, uses a two-step process to choose knowledge based on how new or different a situation feels, and it updates knowledge at different speeds depending on its scale. Tests showed that MuSix helps agents think and adapt better over time compared to other approaches.
Embodied agentsMulti-scale reasoningMixture of Experts (MoE)Routing mechanismConstrual Level TheoryWorld modelScale-aware adaptationForgetting ratesHierarchical knowledgeDynamic adaptation
Authors
Jinwoo Jang, Daniel J. Rho, Sihyung Yoon, Hyunsuk Cho, Honguk Woo
Abstract
Embodied agents operating in the real world require multi-scale reasoning and knowledge adaptation as conditions change. We identify two challenges in applying Mixture of Experts (MoE) to this setting: routing lacks an explicit notion of scale, preventing targeted updates at specific scales, and a uniform update policy cannot accommodate the different rates at which knowledge at each scale becomes outdated. We present MuSix, a framework that addresses both challenges through scale-aware world model mixture and evolution. A two-stage routing mechanism grounds scale selection in experiential distance, a measure of situational novelty inspired by Construal Level Theory: a meta-router first maps this quantity to a weight over continuous scale space, then per-scale base routers select world models within the identified scale. For adaptation, scale-dependent forgetting rates allow low-scale knowledge to refresh rapidly while high-scale abstractions persist, and gated inter-scale transfer maintains coherence across the hierarchy. Experiments on EmbodiedBench and HAZARD show that MuSix improves over state-of-the-art baselines on multi-scale reasoning and dynamic adaptation.