CohortFlow：支持局部语义修正的可操控临床队列构建智能体
原标题：CohortFlow: A Steerable Clinical Cohort Curation Agent with Localized Semantic Correction
AI 摘要
研究团队提出 CohortFlow，一个可操控的临床队列构建智能体，将自然语言修正表示为对结构化队列规范的局部更新，并与持久执行图协调，通过显式语义依赖选择性失效受影响的计算、保留有效结果。在 100 个源自临床试验入组标准的场景（126 轮修正）上，局部引导的需求级 micro-F1 为 77.0%，与全量重生成的 76.4% 无显著差异，但修正延迟在完成后降低 64.8%、执行中降低 44.3%，模型调用减少 38.1%。在 MIMIC 上的概念验证中，13 个队列状态均成功执行。
正文节选
CohortFlow: A Steerable Clinical Cohort Curation Agent with Localized Semantic Correction Abstract Language model agents for clinical cohort curation should accommodate evolving eligibility criteria without repeatedly processing unchanged requirements. We introduce CohortFlow, a steerable agent that represents natural language corrections as localized updates to a structured cohort specification and reconciles them with a persistent execution graph. Explicitly recorded semantic dependencies enab