Multi-scale Mixture of World Models for Embodied Agents in Evolving Environments
Quick Answer
MuSix is a new framework for embodied agents that enhances multi-scale reasoning and adaptation in evolving environments.
Quick Take
It introduces a two-stage routing mechanism and scale-dependent forgetting rates, outperforming state-of-the-art methods on benchmarks like EmbodiedBench and HAZARD.
Key Points
- MuSix addresses routing challenges in Mixture of Experts for embodied agents.
- Experiential distance informs scale selection in the two-stage routing mechanism.
- Scale-dependent forgetting rates enable rapid updates for low-scale knowledge.
- Gated inter-scale transfer ensures coherence across different knowledge scales.
- Experiments show significant improvements over existing methods on key benchmarks.
Paper Resources
📖 Reader Mode
~2 min readAbstract: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scales, and a uniform update policy cannot accommodate the different rates at which knowledge at each scale becomes outdated.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.
| Comments: | Accepted at ECCV 2026. 15 pages |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.00457 [cs.AI] |
| (or arXiv:2607.00457v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.00457 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jinwoo Jang [view email]
[v1]
Wed, 1 Jul 2026 05:23:56 UTC (9,191 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.