1 paper
Qiqi Liu, Thorsten Holz, Shilin Ye +1
Multi-agent systems extend large language models (LLMs) by decomposing tasks among specialized agents, but their distributed decision process creates new attack surfaces. We identi…