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

2026-05-17

AI Summary

This work addresses the premature discarding of “zombie” agents—those with potential for recovery—in existing large language model–based multi-agent systems during graph evolution, which often results from aggressive pruning and leads to valuable information loss. To mitigate this issue, the authors propose AgentRevive, a framework featuring a three-state (active, standby, terminated) soft state transition mechanism. By integrating Markovian state awareness, hallucination risk assessment, and state-aware edge pruning, AgentRevive enables resilient agent evolution and dynamic collaboration management. The approach effectively preserves potentially useful agents while suppressing the influence of unreliable nodes, significantly outperforming strong baselines across general reasoning, domain-specific tasks, and high-hallucination scenarios. Furthermore, state-aware scheduling substantially reduces token consumption.

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.

Problem

• multi-agent systems
• agent pruning
• zombie agents
• hallucination
• graph evolution

Innovation

• Markov state-aware
• multi-agent evolution
• soft state transition
• hallucination risk estimation
• agent revival