Recognize Your Orchestrator: An Entropy Dynamics Perspective for LLM Multi-Agent Systems

Junze Zhu, Weihao Chen, Xuanwang Zhang, Zhen Wu, Xinyu Dai · 2026

Read Recognize Your Orchestrator: An Entropy Dynamics Perspective for LLM Multi-Agent Systems by Junze Zhu, Weihao Chen, Xuanwang Zhang, Zhen Wu, Xinyu Dai (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-01351.txt.

Description

Zhu and co-authors model why multi-agent orchestrators fail. Multi-agent systems promise stronger problem solving than single models. Their central orchestration remains a fragile point, though. A mean-field entropy dynamics frame captures the process. Task resolution and context loading act as competing forces. To test it, they build Inverse Workflow Generation. That pipeline synthesizes hard benchmarks with dense checkpoints. Each checkpoint can be verified against the process.

The entropy model fits real trajectories well. Its parameters carry physical meaning about stability and collapse. A reasoning trap emerges from their analysis. Reasoning-heavy models shine on isolated tasks. Yet they often fail as orchestrators through context squeezing. The findings inform multi-agent architecture design.

Catalog record

CREATORJunze Zhu, Weihao Chen, Xuanwang Zhang, Zhen Wu, Xinyu Dai
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE72 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.01351

Browse the full library · About the Tower of Babel