TodyComm: Task-Oriented Dynamic Communication for Multi-Round LLM-based Multi-Agent System
This work addresses the need for adaptive communication in multi-agent systems for applications with dynamic adversaries or constraints, representing an incremental improvement over fixed-topology methods.
The paper tackled the problem of fixed communication topologies in multi-round LLM-based multi-agent systems, which are inadequate for dynamic scenarios like changing roles or constraints, by proposing TodyComm, a task-oriented dynamic communication algorithm that adapts topologies per round; experiments on five benchmarks showed it achieves superior task effectiveness while maintaining token efficiency and scalability under dynamic conditions.
Multi-round LLM-based multi-agent systems rely on effective communication structures to support collaboration across rounds. However, most existing methods employ a fixed communication topology during inference, which falls short in many realistic applications where the agents' roles may change \textit{across rounds} due to dynamic adversary, task progression, or time-varying constraints such as communication bandwidth. In this paper, we propose addressing this issue through TodyComm, a \textbf{t}ask-\textbf{o}riented \textbf{dy}namic \textbf{comm}unication algorithm. It produces behavior-driven collaboration topologies that adapt to the dynamics at each round, optimizing the utility for the task through policy gradient. Experiments on five benchmarks demonstrate that under both dynamic adversary and communications budgets, TodyComm delivers superior task effectiveness while retaining token efficiency and scalability.