When Messages Go Wrong: Understanding Communication Faults in LLM-Based Multi-Agent Systems

Key points

• Because agents often exchange open-ended natural language, anomalies can alter task-relevant information while leaving messages processable, allowing semantic errors to propagate through subsequent interactions.1
• Systematic studies of faults in inter-agent communication remain limited.1
• This paper presents a large-scale empirical study of how communication faults manifest, propagate, and affect task outcomes in LLM-MAS.1
• These observations provide empirical insights into communication-fault propagation, workflow resilience, and protection design for LLM-MAS.1

Sources (1)

• [1]When Messages Go Wrong: Understanding Communication Faults in LLM-Based Multi-Agent Systems
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 08:59 AM

  LLM-based multi-agent systems (LLM-MAS) use specialized agents that exchange messages to accomplish complex tasks.

  Because agents often exchange open-ended natural language, anomalies can alter task-relevant information while leaving messages processable, allowing semantic errors to propagate through subsequent interactions.

Extractive summary: sentences quoted from the sources.