How Task Structure Limits Multi-Agent Success: An Information-Theoretic Analysis

2026-06-11

AI Summary

This work addresses the limited information-processing capacity of multi-agent systems in collaborative tasks, which often stems from task structure and communication constraints and can prevent them from outperforming single-agent solutions. The authors propose the “minimum cut cost,” denoted $C_{\text{min}}$, as a key metric quantifying how the connectivity of a task’s constraint graph influences the system’s success probability. They establish a unified theoretical bound applicable to both open and closed systems, leveraging information-theoretic analysis, graph partitioning theory, and constraint graph modeling. Empirical validation through synthetic experiments and real-world data from SWE-bench demonstrates that success probability decays exponentially with information bottlenecks. Crucially, tasks with high $C_{\text{min}}$ benefit more from restructuring the task graph than from merely adding agents or bandwidth, offering a new design paradigm for multi-agent systems.

Abstract

Multi-agent systems (MAS) were expected to overcome the limitation of single-agent systems (SAS) through collaboration. However, under typicality conditions on the task's constraint graph and bounded inter-agent communication, we prove that the success probability of a MAS is closely tied to the connectivity of task constraints, where each agent has limited information-processing capacity. Specifically, the success probability decays exponentially with an information bottleneck that emerges from partitioning the task's constraint graph among agents. We define this quantity as the \emph{minimum cut cost} $C_{\min}$ of the potential constraint graph of each task. This information-theoretic bound applies to both open systems with external feedback and closed systems without. We validate our theory on both synthetic experiments and real-world empirical data from SWE-bench submissions. From our framework, effective MAS design should incorporate task-inherent constraints alongside engineering optimization, and when $\Cmin$ is high, practitioners should restructure tasks rather than simply scaling agents or communication.

Problem

• multi-agent systems
• task structure
• information bottleneck
• constraint graph
• success probability

Innovation

• multi-agent systems
• information bottleneck
• constraint graph
• minimum cut cost
• task structure