arXiv:2508.19096cs.AI2025-08

提出新评估指标与可信医疗智能体,降低大模型误判风险。

Trustworthy Agents for Electronic Health Records through Confidence Estimation

  • 设计分步置信度估计机制,动态判断回答可靠性
  • 在MIMIC-III和eICU上,70%置信阈值下准确率提升44.23%和25.34%
  • 适合需高可靠性医疗AI部署的临床场景

大语言模型在电子健康记录(EHR)信息提取和临床决策中展现出潜力,但幻觉风险限制了其在临床环境中的应用。本文提出一种新指标HCAcc@k%,用于量化不同置信度阈值下的准确率-可靠性权衡。我们构建了具备置信度感知能力的TrustEHRAgent,通过分步置信度估计实现临床问答。在MIMIC-III和eICU数据集上的实验表明,该方法在严格可靠性约束下优于基线模型:在HCAcc@70%条件下,准确率分别提升44.23个百分点和25.34个百分点,而基线方法在此阈值下无法通过。结果揭示传统准确率指标在医疗AI评估中的局限性。本工作推动可信赖临床智能体的发展,使其在低置信度时能准确输出或明确表达不确定性。

原文摘要 · Abstract (English)

Large language models (LLMs) show promise for extracting information from Electronic Health Records (EHR) and supporting clinical decisions. However, deployment in clinical settings faces challenges due to hallucination risks. We propose Hallucination Controlled Accuracy at k% (HCAcc@k%), a novel metric quantifying the accuracy-reliability trade-off at varying confidence thresholds. We introduce TrustEHRAgent, a confidence-aware agent incorporating stepwise confidence estimation for clinical question answering. Experiments on MIMIC-III and eICU datasets show TrustEHRAgent outperforms baselines under strict reliability constraints, achieving improvements of 44.23%p and 25.34%p at HCAcc@70% while baseline methods fail at these thresholds. These results highlight limitations of traditional accuracy metrics in evaluating healthcare AI agents. Our work contributes to developing trustworthy clinical agents that deliver accurate information or transparently express uncertainty when confidence is low.

医疗AI大模型置信度估计可信计算

Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。