更多智能体更可靠？运行剖面不确定下多智能体LLM的系统级可靠性
原标题：More Agents, More Reliable? System-Level Reliability of Multi-Agent LLMs under Operational-Profile Uncertainty
AI 摘要
该论文提出一种面向多智能体LLM系统的系统级可靠性评估框架，将HIP-LLM对单个LLM的分层不精确概率评估与编码系统失效逻辑的贝叶斯网络结合，并把智能体输出合并规则本身视为可能失效的环节。在提取受监管化学品危害声明的真实流水线案例中，对三个模型进行多数投票虽能抵御最弱模型，但并未明显超过最佳模型；若投票器在5%的需求上失效，其相对中间模型的优势将基本消失。作者据此指出，更多智能体带来的是更可辩护的可靠性声明，而非自动更可靠的系统。
正文节选
More Agents, More Reliable? System-Level Reliability of Multi-Agent LLMs under Operational-Profile Uncertainty Abstract Teams increasingly build software from cooperating large language model (LLM) agents, often assuming more agents mean more reliability. Yet the usual evidence, average benchmark accuracy, ignores the input mix a deployed system meets and the rule merging the agents’ answers. Building on HIP-LLM, a hierarchical imprecise-probability assessment of single LLMs, we propose a framew