arXiv · AI· Keegan Wang, Anantika Mannby·· 4 天前精选AI 评分63
Only as Safe as the Weakest Agent: A Break That Appears Only When the System Is Multi-Agent
Only as Safe as the Weakest Agent: A Break That Appears Only When the System Is Multi-Agent
AI 导读
arXiv论文《Only as Safe as the Weakest Agent》提出多智能体系统安全取决于最弱子智能体,强编排模型在委托子代理时会在特定攻击下失败,Claude Code等系统在18个模型测试中表现出该现象。
推荐理由
论文指出多智能体系统的安全边界取决于最弱子智能体,强编排模型委托时会在特定攻击下失败,提供了多智能体安全评估的新视角。
来源:arXiv · AI · arxiv.org