SentryLine：肿瘤护理中不断演变的文档上的证据支撑问答
SentryLine: Evidence-Grounded Question Answering over Evolving Documents in Oncology Care
- Arizona State University（亚利桑那州立大学）
- Mayo Clinic（梅奥诊所）
- Jawaharlal Nehru Medical College（贾瓦哈拉尔·尼赫鲁医学院）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
针对肿瘤护理中动态指南更新带来的问答挑战，提出SENTRYLINE系统，采用无向量分层RAG管道检索指南并生成带引用的角色特定答案，构建ASCOBENCH基准验证，在推理和角色特定问题上显著优于基线。
AI中文摘要：
肿瘤护理工作持续承受着吸收生物医学领域快速演变证据基础的压力。美国临床肿瘤学会（ASCO）通过制定动态指南来应对这一问题，但该格式引入了新的负担：任何建议都可能在任何时间点发生变化，且这些变化分散在多个版本化的文档中。我们提出了SENTRYLINE，一个感知动态指南的临床问答系统。SENTRYLINE通过无向量分层RAG管道检索指南段落，并返回具有角色特定性的答案，附带内联引用、事实和时间验证报告，以及漂移检测注释，用于提示指南何时已被更新。我们构建了ASCOBENCH基准，包含405个三轮对话，涵盖四个问题类别，并配有来自专家标注者（临床医生）的金标准答案，并使用测试集在LLM-as-judge框架下将SENTRYLINE与五个基线进行比较。在三个生成骨干上的实验表明，与四个检索基线和ASCO的指南助手相比，SENTRYLINE取得了持续改进，特别是在需要多跳综合和语域适应的推理和角色特定问题上表现尤为突出。
英文摘要：
Oncology care operates at constant pressure of absorbing rapidly evolving evidence base in biomedicine. The American Society of Clinical Oncology (ASCO) addresses this through living guidelines, but the format introduces a new burden: any recommendation can change at any point, across multiple versioned documents. We present SENTRYLINE, a living guideline-aware clinical question answering system. SENTRYLINE retrieves guideline passages through a vectorless hierarchical RAG pipeline and returns a role-specific answer with inline citations, factual and temporal verification reports, and drift detection notes that surface when a guideline has been updated. We construct ASCOBENCH, a benchmark of 405 three-turn conversations across four question categories with gold answers from expert annotators(clinicians), and use test set to evaluate SENTRYLINE against five baselines under an LLM-as-judge framework. Experiments across three generation backbones show consistent improvements over four retrieval baselines and ASCO's guideline assistant, with particularly strong gains on Reasoning and Role-Specific questions where multi-hop synthesis and register adaptation are required
