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

2026-05-31

AI Summary

This work addresses the vulnerability of centralized coordinators in multi-agent systems by proposing a mean-field entropy dynamics framework that models coordination among large language models from an entropic perspective. It reveals a competitive mechanism between task-solving efficacy and contextual accumulation, and introduces a high-complexity, verifiable benchmark constructed via inverse workflow generation. By analyzing contextual entropy trajectories, the study quantifies and physically interprets the system uncertainty induced by coordinators for the first time. Empirical results demonstrate that reasoning-capable models, when acting as coordinators, are prone to failure due to context compression, offering critical theoretical insights and practical guidance for designing robust multi-agent architectures.

Abstract

The transition from single-turn models to Multi-Agent Systems (MAS) promises enhanced problem-solving capabilities, yet the centralized orchestration topology remains a critical point of fragility. To analyze this, we propose a Mean-Field Entropy Dynamics framework, modeling the orchestration process as a system governed by the competing forces of task resolution and cumulative context loading. To facilitate validation, we introduce Inverse Workflow Generation (IWG), a multi-agent pipeline that synthesizes process-verifiable, high-complexity benchmarks with dense intermediate checkpoints. We demonstrate that our entropy dynamics model fits empirical trajectories, providing physically interpretable parameters that quantify system stability and performance collapse. Crucially, our analysis uncovers a ``Reasoning Trap": while reasoning-heavy models excel in isolated tasks, they frequently fail as orchestrators due to context squeezing. Elucidating the physical mechanisms underlying the Orchestrator and quantifying systemic uncertainty offers insights for the MASs' architectural design.

Problem

• Multi-Agent Systems
• Orchestrator
• Entropy Dynamics
• Context Loading
• Reasoning Trap

Innovation

• Mean-Field Entropy Dynamics
• Multi-Agent Systems
• Orchestrator
• Inverse Workflow Generation
• Reasoning Trap