Bayes-Sufficient Compression Is Not Enough: How Does Communication Help Multi-Agent Systems?

2026-09-29

AI Summary

This study investigates when communicative messages outperform raw context in multi-agent LLM systems and how sender capability influences executor decision-making. It proposes a "receiver-relative bounded coordination" framework that formalizes message utility as the difference between informational gain and protocol tax, revealing the failure mechanism of Bayesian sufficient compression under finite executors and establishing a three-stage decomposition model comprising externalization, absorption, and action closure. Methodologically, this work integrates single-crossing condition theory, fixed-message replay, and inference-time selector algorithms. Experiments across six benchmarks demonstrate that the proposed approach significantly improves the Pareto frontier of the accuracy–communication cost trade-off.

Abstract

Multi-agent LLM systems pair a sender with broad context and an executor with a limited local view. We study when a short message improves the executor's next decision, when raw context is preferable, and when a stronger sender helps. Our framework, \emph{receiver-relative bounded coordination}, expresses message utility as receiver gain minus protocol tax. Compression beats raw context when tax savings exceed losses from omitted information and decoder mismatch. Even \emph{Bayes-sufficient} compression can fail when a bounded executor cannot use its surface form. A three-stage decomposition separates externalization, absorption, and \emph{action closure}, explaining how errors remain after the correct content reaches the receiver. Under a single-crossing condition, sender upgrades help above a receiver-burden threshold. Across six benchmarks, the same Qwen protocol raises ContextBench joint accuracy from $0.633$ to $0.775$ but lowers ToolSandbox from $0.889$ to $0.653$. Fixed-message replay reveals closure failures despite correct artifact recovery. These results guide an inference-time selector that improves the accuracy-cost frontier on the evaluated communication regimes.

Problem

• multi-agent LLM systems
• communication
• Bayes-sufficient compression
• bounded executor
• action closure

Innovation

• bounded coordination
• Bayes-sufficient compression
• action closure
• protocol tax
• inference-time selector