MedCaseAgent:从多模态临床证据到罕见病病例报告
摘要
该论文介绍了 MedCaseAgent,这是一种智能体框架,通过将分层患者数据挖掘与迭代式工具集成推理相结合,将非结构化临床笔记转化为达到发表标准的罕见病病例报告。论文还发布了 MedCase-150K 语料库、MedCase-Bench 评估套件,以及一个基于 Qwen3.5-9B 进行后训练的开源模型,其表现优于现有的医学大语言模型和智能体。
arXiv:2610.32258v1 Announce Type: new
Abstract: Although Large Language Models (LLMs) have demonstrated potential in clinical documentation, generating the rigorous, evidence-based reasoning required for unprecedented, atypical, and rare disease case reports remains a significant challenge. We present MedCaseAgent, an end-to-end agentic reasoning paradigm that bridges this gap by synthesizing publication-quality reports from unstructured clinical notes. MedCaseAgent leverages a hierarchical structure, integrating patient-level data mining with iterative, tool-integrated reasoning. By dynamically querying medical literature and knowledge bases, the agent maintains high clinical fidelity and assesses case novelty. The framework is inherently modular, supporting various backbone LLMs in a plug-and-play fashion. To facilitate this research, we release MedCase-150K, a large-scale multimodal clinical case-report corpus, and MedCase-Bench, a robust temporally held-out evaluation suite. We also introduce an open-source model post-trained from Qwen3.5-9B and optimized specifically for this task. Under both LLM-judged and rule-based evaluation, MedCaseAgent outperforms existing general-purpose and specialized medical LLMs and agents. Blinded expert reviews and ablations further support the system's clinical relevance and evidence grounding.
相似文章
Inkling 2 Mini 在 CyberBench 上与 Opus 5.5 并列(@ValsAI)
根据 ValsAI 的结果,Inkling 2 Mini 在 CyberBench 基准测试中的表现与 Opus 5.5 相当。
饥饿管家效应:大语言模型中隐藏的商业激励
介绍了“饥饿管家效应”,即大语言模型在输出中嵌入隐蔽商业动机的现象,并探讨了隐藏的说服意图和逐利目标如何在大语言模型中产生。
问答:让多模态基础模型具备更类人的交互和理解能力
MIT-IBM 计算研究实验室与 Jim Glass 的一场问答探讨了多模态基础模型失效模式的研究、新型基准测试,以及减少对标注训练数据依赖的数据高效利用技术。
@ClementDelangue:美国开源AI还没死呢 @thinkymachines 今天将以Apache 2.0发布Inkling 2 Mini,并在开源智能体模型中达到帕累托前沿……
Thinking Machines Lab 将发布 Inkling 2 Mini,这是一款紧凑的开放权重多模态模型(文本+图像转文本),采用 Apache 2.0 协议,面向智能体、编码和 RAG 应用,检查点大小约 20GB,可在消费级硬件上运行。
@rohanpaul_ai:Google 与 Johns Hopkins 的新论文表明,AI 研究智能体在追踪性能瓶颈时表现更好,而不仅仅是……
Google 与 Johns Hopkins 的一篇论文提出了 Pro2 Agent,它通过单变量实验来追踪疑似瓶颈并更新信念,而非只保留得分最高的代码,在 48 小时内使用 Gemini 3.5 Flash 完成了全部 75 个 MLE-Bench 任务,达到了创纪录的 76.0% 奖牌率。