基于Koopman谱分析认证多智能体系统集体推理
原标题：Certifying Collective Reasoning in Multi-Agent Systems via Koopman Spectral Analysis
AI 摘要
该研究提出一种基于Koopman算子理论的新框架，用于验证多智能体系统中大型语言模型（LLM）集体推理的收敛性和可解释性。通过将集体视为非线性动力系统，从交互轨迹中估计Koopman传递算子，其谱可提供收敛期限、派系识别和压缩消息基础等机器可检查的证书。实验表明，该框架在注意力共识模型上以高相关性预测收敛时间，并在多数配置中提供有效边界，且计算高效，可在CPU上数分钟内完成。
正文节选
Computer Science > Multiagent Systems Title:Certifying Collective Reasoning in Multi-Agent Systems via Koopman Spectral Analysis View PDF HTML (experimental) Abstract:Orchestrated collectives of large language model (LLM) agents that debate and vote are an emerging form of computational intelligence: the intelligent behaviour resides in the \emph{interaction}, not in any single agent. They improve task accuracy, yet remain black boxes at the system level: there is no principled test