多智能体LLM流水线的协调税与异构红利：盈亏平衡条件
原标题：The Coordination Tax and the Heterogeneity Dividend: Break-Even Conditions for Multi-Agent LLM Pipelines
AI 摘要
该论文提出一个闭式模型，将多智能体LLM流水线的协调税（分解开销加每次交接的协议成本）与异构红利（把工作路由给更便宜模型带来的节省）进行对比，给出盈亏平衡阈值和可行性边界α<1/r。作者在135个HumanEval和ClassEval任务上、用Claude Opus 5.5和Haiku 5.5运行六种架构共2430次，测得OFP协议每次交接仅增加7.2毫秒（占流水线总时长0.11%）。结果显示规划者-编码者-审查者流水线成本是单次强模型调用的3.22倍、耗时3.24倍且未多解出任务；将规划者和审查者换成廉价模型虽把成本降至0.39倍，但pass@1下降6.7个百分点，仍比单次调用更贵更慢更不准。真正的节省来自按任务选择模型：在两个单次调用间做oracle选择，以强调用一半的每解出任务成本覆盖94.3%的运行。
正文节选
The Coordination Tax and the Heterogeneity Dividend: Break-Even Conditions for Multi-Agent LLM Pipelines Abstract Multi-agent LLM pipelines built on interoperability protocols such as the Open Floor Protocol (OFP) or A2A are usually justified on capability grounds, with their runtime cost treated as an overhead of unknown size. Comparing such a pipeline with one model call is underspecified, because a pipeline can give each role a different model. We give a closed-form model that sets the coordi