Taming "Zombie'' Agents: A Markov State-Aware Framework for Resilient Multi-Agent Evolution

TL;DR

This paper introduces AgentRevive, a Markov state-aware framework that enhances multi-agent system resilience by managing agent states and interactions to recover 'zombie' agents and improve efficiency.

Contribution

The paper presents a novel state-aware policy and edge optimization approach for resilient multi-agent evolution, addressing the issue of premature agent pruning.

Findings

Outperforms strong baselines across various tasks.

Reduces token consumption through agent scheduling.

Effectively manages agent states to recover 'zombie' agents.

Abstract

Recent advancements in LLM-based multi-agent systems have demonstrated remarkable collaborative capabilities across complex tasks. To improve overall efficiency, existing methods often rely on aggressive graph evolution among agents (e.g., node or edge pruning), which risks prematurely discarding valuable agents due to transient issues such as hallucinations or temporary knowledge gaps. However, such hard pruning overlooks the potential for ``zombie'' agents to recover and contribute in subsequent discussion rounds. In this paper, we propose AgentRevive, a Markov state-aware framework for resilient multi-agent evolution. Our approach dynamically manages agent collaboration through soft state transitions, implemented via two key components: (1) State-Aware Policy Learning: Agent states are divided into ``Active'', ``Standby'', and ``Terminated'' states, selectively propagating messages based on agent memory. The policy employs a risk estimator to optimize agent state transitions by assessing hallucination risk, minimizing the influence of unreliable nodes while safeguarding valuable ones. (2) State-Aware Edge Optimization: Subgraph edges are pruned according to states learned from the policy, permanently removing ``Terminated'' nodes and retaining ``Standby'' nodes for subsequent rounds to assess their potential future contributions. Extensive experiments on general reasoning, domain-specific, and hallucination challenge tasks show that our method consistently outperforms strong baselines and significantly reduces token consumption through state-aware agent scheduling.