Evolving Diagnostic Agents in a Virtual Clinical Environment

印为红字的 计算机科学 医学诊断 提交 强化学习 水准点(测量) 机器学习 诊断准确性 人工智能 虚拟病人 诊断试验 多样性(控制论) 医学物理学 体格检查 临床诊断 临床实习 临床决策支持系统 医学影像学 训练集 最佳实践 客观结构化临床检查 病史
作者
Qiu Pengcheng,Wu Chaoyi,Liu Junwei,Zheng Qiao-yu,Liao Yu-Sheng,Wang Haowen,Yue Yun,Fan, Qianrui,Zhen Shuai,Wang Jian,Gu, Jinjie,Wang YanFeng,Zhang, Ya,Xie, Weidi
出处
期刊:Cornell University - arXiv [Cornell University]
标识
DOI:10.48550/arxiv.2510.24654
摘要

In this paper, we present a framework for training large language models (LLMs) as diagnostic agents with reinforcement learning, enabling them to manage multi-turn diagnostic processes, adaptively select examinations, and commit to final diagnoses. Unlike instruction-tuned models trained on static case summaries, our method acquires diagnostic strategies through interactive exploration and outcome-based feedback. Our contributions are fourfold: (i) We present DiagGym, a diagnostics world model trained with electronic health records that emits examination outcomes conditioned on patient history and recommended examination, serving as a virtual clinical environment for realistic diagnosis training and evaluation; (ii) We train DiagAgent via end-to-end, multi-turn reinforcement learning to learn diagnostic policies that optimize both information yield and diagnostic accuracy; (iii) We introduce DiagBench, a diagnostic benchmark comprising 750 cases with physician-validated examination recommendations and 99 cases annotated with 973 physician-written rubrics on diagnosis process; (iv) we demonstrate superior performance across diverse diagnostic settings. DiagAgent significantly outperforms 10 state-of-the-art LLMs, including DeepSeek-v3 and GPT-4o, as well as two prompt-engineered agents. In single-turn settings, DiagAgent achieves 9.34% higher diagnostic accuracy and 44.03% improvement in examination recommendation hit ratio. In end-to-end settings, it delivers 15.12% increase in diagnostic accuracy and 23.09% boost in examination recommendation F1 score. In rubric-based evaluation, it surpasses the next-best model, Claude-sonnet-4, by 7.1% in weighted rubric score. These findings indicate that learning policies in interactive clinical environments confers dynamic and clinically meaningful diagnostic management abilities unattainable through passive training alone.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
我能私信骂你吗应助wt采纳,获得10
刚刚
1秒前
NexusExplorer应助donald采纳,获得10
1秒前
辞忧完成签到,获得积分10
2秒前
愉快敏发布了新的文献求助10
3秒前
Polly完成签到,获得积分10
3秒前
3秒前
Acer完成签到 ,获得积分10
4秒前
天天开心完成签到,获得积分10
5秒前
煎饼完成签到,获得积分10
5秒前
雪婷发布了新的文献求助10
6秒前
6秒前
henry先森完成签到,获得积分10
7秒前
lily发布了新的文献求助10
7秒前
7秒前
7秒前
郭淳完成签到,获得积分10
8秒前
8秒前
zzz完成签到 ,获得积分10
8秒前
liliqiong完成签到,获得积分10
9秒前
往哪李跑完成签到,获得积分10
9秒前
9秒前
科研通AI6.4应助儒雅老太采纳,获得10
9秒前
9秒前
单亚楠发布了新的文献求助10
9秒前
110011完成签到,获得积分10
10秒前
10秒前
英俊的铭应助天天开心采纳,获得10
11秒前
yy发布了新的文献求助10
11秒前
超帅的笑蓝完成签到,获得积分0
12秒前
LI关注了科研通微信公众号
13秒前
110011发布了新的文献求助10
13秒前
13秒前
搜集达人应助淡淡乾采纳,获得10
13秒前
今后应助愉快敏采纳,获得10
14秒前
大个应助yy采纳,获得10
14秒前
15秒前
忧虑的青旋关注了科研通微信公众号
15秒前
dyd发布了新的文献求助10
16秒前
16秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
HYDROLYSE ACIDE DE QUELQUES DIOXASPIROCYCLANES 1314
Essentials of Carbohydrate Chemistry and Biochemistry, 4th Edition 800
Navigating Normative Orders. Interdisciplinary Perspectives 800
1 Peter and Christ's Descent to the Dead in Its Early Christian Reception 700
Organizational Behavior 510
Management and the Arts 510
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 工程类 有机化学 化学工程 生物化学 计算机科学 内科学 物理 复合材料 催化作用 细胞生物学 无机化学 光电子学 物理化学 电极 基因
热门帖子
关注 科研通微信公众号,转发送积分 7746549
求助须知:如何正确求助?哪些是违规求助? 9294420
关于积分的说明 20224582
捐赠科研通 7326479
什么是DOI,文献DOI怎么找? 3308115
关于科研通互助平台的介绍 2460125
邀请新用户注册赠送积分活动 2319789