发表机构
University of Texas at Dallas（德克萨斯大学达拉斯分校）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出Inherit-MAS，通过工作流与执行继承机制，在测试时利用反馈改进多智能体系统，提升任务完成率并显著降低令牌消耗。
AI 中文摘要
由大型语言模型构建的多智能体系统（MAS）协调专业智能体处理复杂任务，但有效的工作流难以预先设计。测试时演化利用执行反馈来改进工作流，然而大范围的修改可能干扰有用组件，而重新执行未更改的请求则可能造成冗余计算。受生物演化中继承与选择相互作用的启发，我们提出了Inherit-MAS，该方法在工作流和执行两个层面显式地引入继承机制。一个元模型首先综合出一个由工作者智能体组成的工作流，其中包含声明的角色、通信输入和工具权限，并由一个单独提示的评判者（judge）对每个执行候选进行评估并诊断其缺陷。在常规的改进轮次中，工作流继承从最近完成的候选开始，可能丢弃被判定为无用的可移除节点，并应用经过验证的编辑来解决已诊断的缺陷。当新候选执行时，执行继承仅在完整解析的请求和执行上下文匹配的情况下继承符合条件的已存储结果，从而避免冗余的模型和工具调用。使用GPT-4o-mini工作者，Inherit-MAS在WorkBench上达到55.4%的完成率，在HotpotQA FullWiki上达到49.7%的联合F1分数，优于EvoAgent、EvoMAS和TacoMAS。使用Qwen3-32B工作者时，它在两个基准上也超过了这些演化型MAS基线。与禁用执行继承的相同控制器重新运行相比，执行继承在WorkBench上将工作者令牌使用量减少了29.1%，在HotpotQA上减少了34.6%，总令牌使用量分别减少了5.3%和18.1%。
英文摘要
Multi-agent systems (MAS) built from large language models coordinate specialized agents to tackle complex tasks, but effective workflows are difficult to design in advance. Test-time evolution refines workflows using execution feedback, yet broad revisions can disturb useful components, while re-executing unchanged requests can incur redundant computation. Inspired by the interplay of inheritance and selection in biological evolution, we introduce Inherit-MAS, which makes inheritance explicit at the workflow and execution levels. A meta-model first synthesizes a workflow of worker agents with declared roles, communication inputs, and tool permissions, and a separately prompted judge scores each executed candidate and diagnoses its deficiencies. In ordinary refinement rounds, \emph{workflow inheritance} starts from the latest completed candidate, may discard removable nodes judged unhelpful, and applies a validated edit to address the diagnosed deficiency. When the new candidate executes, \emph{execution inheritance} inherits eligible stored results only if the complete resolved request and execution context match, avoiding redundant model and tool calls. With GPT-4o-mini workers, Inherit-MAS achieves 55.4\% completion on WorkBench and 49.7\% joint F1 on HotpotQA FullWiki, outperforming EvoAgent, EvoMAS, and TacoMAS. With Qwen3-32B workers, it also exceeds these evolving-MAS baselines on both benchmarks. Compared with rerunning the same controller with execution inheritance disabled, execution inheritance reduces worker-token usage by 29.1\% on WorkBench and 34.6\% on HotpotQA, and total token usage by 5.3\% and 18.1\%.
