副本还是来源？衡量LLM聚合器在多智能体系统中如何计数重述证据

中文摘要 AI 辅助

基于大型语言模型（LLM）的多智能体系统理所当然地会重述观察结果：中继转发它们，共享板重复它们，讨论轮次也回响它们。聚合此类消息的聚合器应计数来源，而非陈述。我们将报告的概率转换为独立阅读的单位，这赋予每次重述一个副本权重，对于计数来源的聚合器为0，对于计数每条陈述的聚合器为1，并在任何成本结构下产生隐含的决策。三个测试平台保持证据固定，并改变其重述方式：带有精确贝叶斯预言机的消息日志、附加副本的网页文档，以及由LLM智能体团队在四种通信协议下编写的日志。在来自三个提供商的四种模型中，转发副本的计数相当于新阅读的0.06至0.42，主要是因为某些回复计数每条陈述。在5%至40%的日志中，陈述同一阅读三次，报告的信信念暗示了预言机从未做出的早期承诺。在受控日志上计数副本最少和最多的模型，在网页副本和智能体编写的日志上也同样如此。一段关于副本贡献的声明将受控日志上的副本权重降至0.08或更低。一条规则让智能体引用阅读而非重述它们，将信念隐含的早期承诺从11.2%降至1.1%，并保留真正的确证。

英文摘要

Multi-agent systems built on large language models (LLMs) restate observations as a matter of course: relays forward them, shared boards repeat them and discussion rounds echo them. An aggregator that pools such messages should count sources, not statements. We convert a reported probability into units of independent readings, which assigns every restatement a copy weight, 0 for an aggregator that counts sources and 1 for one that counts every statement, and yields the implied decision under any cost structure. Three testbeds hold the evidence fixed and vary how it is restated: message logs with an exact Bayesian oracle, web documents with appended copies, and logs written by LLM agent teams under four communication protocols. Across four models from three providers, a forwarded copy counts for 0.06 to 0.42 of a new reading, mostly because some replies count every statement. On 5% to 40% of logs that state one reading three times, the reported belief implies an early commitment that the oracle never makes. The models that count copies least and most on controlled logs do so on web copies and agent-written logs as well. A one-paragraph declaration of what a copy contributes brings the copy weight on controlled logs to 0.08 or less. A rule that has agents refer to readings instead of restating them cuts belief-implied early commitment from 11.2% to 1.1% and preserves genuine corroboration.

发表机构

• China Agricultural University（中国农业大学）
• Jilin University（吉林大学）
• Tianjin University of Finance and Economics（天津财经大学）
• Tianjin University of Science and Technology（天津科技大学）

机构由 AI 辅助整理，请以论文原文为准。