Search papers, labs, and topics across Lattice.
This paper introduces MuSix, a framework designed to enhance embodied agents' performance in evolving environments by implementing a multi-scale mixture of world models. It addresses the limitations of traditional Mixture of Experts (MoE) by incorporating a two-stage routing mechanism that utilizes experiential distance for scale selection and employs scale-dependent forgetting rates for knowledge adaptation. Experimental results demonstrate that MuSix significantly outperforms existing state-of-the-art methods in both multi-scale reasoning and dynamic adaptation tasks on benchmarks like EmbodiedBench and HAZARD.
MuSix enables embodied agents to adaptively manage knowledge across scales, outperforming traditional models in dynamic environments.
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scales, and a uniform update policy cannot accommodate the different rates at which knowledge at each scale becomes outdated.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.
