@inproceedings{feng-etal-2026-diagbench,
title = "{D}iag{B}ench: Evaluating Diagnostic Revision in Large Language Models with Longitudinal Inpatient Records",
author = "Feng, Huawei and
Shi, Dongsheng and
Wang, Linlin",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.193/",
pages = "3821--3832",
ISBN = "979-8-89176-444-6",
abstract = "Reliable clinical decision support requires tracking evidence over a hospital stay and revising a disease diagnosis when new findings challenge the initial assessment. Strong performance on medical question answering does not establish that large language models (LLMs) can make these longitudinal decisions. We introduce DiagBench, a benchmark of 2,865 Chinese inpatient admissions that evaluates diagnostic revision as clinical evidence accumulates. It pairs admission-only prediction with staged evidence disclosure, distinguishing an unchanged diagnosis, refinement of an existing hypothesis, and replacement by a different diagnosis. Development and test cases have human-annotated reference labels. The best LLM configuration in the main evaluation achieves only 52.6{\%} macro recall; 80.3-87.7{\%} of errors across five DeepSeek-Chat (DeepSeek-V3.2, non-thinking mode) prompting strategies occur between adjacent revision categories. Strategies that improve final-diagnosis prediction do not consistently improve revision decisions. We further use DiagAgent to trace diagnoses and evidence across rounds, revealing failures to replace an initial hypothesis despite recognizing contradictory findings. On DeepSeek-Chat, a single agent with the same state outperforms the full system on macro recall."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="feng-etal-2026-diagbench">
<titleInfo>
<title>DiagBench: Evaluating Diagnostic Revision in Large Language Models with Longitudinal Inpatient Records</title>
</titleInfo>
<name type="personal">
<namePart type="given">Huawei</namePart>
<namePart type="family">Feng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dongsheng</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Linlin</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>Reliable clinical decision support requires tracking evidence over a hospital stay and revising a disease diagnosis when new findings challenge the initial assessment. Strong performance on medical question answering does not establish that large language models (LLMs) can make these longitudinal decisions. We introduce DiagBench, a benchmark of 2,865 Chinese inpatient admissions that evaluates diagnostic revision as clinical evidence accumulates. It pairs admission-only prediction with staged evidence disclosure, distinguishing an unchanged diagnosis, refinement of an existing hypothesis, and replacement by a different diagnosis. Development and test cases have human-annotated reference labels. The best LLM configuration in the main evaluation achieves only 52.6% macro recall; 80.3-87.7% of errors across five DeepSeek-Chat (DeepSeek-V3.2, non-thinking mode) prompting strategies occur between adjacent revision categories. Strategies that improve final-diagnosis prediction do not consistently improve revision decisions. We further use DiagAgent to trace diagnoses and evidence across rounds, revealing failures to replace an initial hypothesis despite recognizing contradictory findings. On DeepSeek-Chat, a single agent with the same state outperforms the full system on macro recall.</abstract>
<identifier type="citekey">feng-etal-2026-diagbench</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.193/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>3821</start>
<end>3832</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T DiagBench: Evaluating Diagnostic Revision in Large Language Models with Longitudinal Inpatient Records
%A Feng, Huawei
%A Shi, Dongsheng
%A Wang, Linlin
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F feng-etal-2026-diagbench
%X Reliable clinical decision support requires tracking evidence over a hospital stay and revising a disease diagnosis when new findings challenge the initial assessment. Strong performance on medical question answering does not establish that large language models (LLMs) can make these longitudinal decisions. We introduce DiagBench, a benchmark of 2,865 Chinese inpatient admissions that evaluates diagnostic revision as clinical evidence accumulates. It pairs admission-only prediction with staged evidence disclosure, distinguishing an unchanged diagnosis, refinement of an existing hypothesis, and replacement by a different diagnosis. Development and test cases have human-annotated reference labels. The best LLM configuration in the main evaluation achieves only 52.6% macro recall; 80.3-87.7% of errors across five DeepSeek-Chat (DeepSeek-V3.2, non-thinking mode) prompting strategies occur between adjacent revision categories. Strategies that improve final-diagnosis prediction do not consistently improve revision decisions. We further use DiagAgent to trace diagnoses and evidence across rounds, revealing failures to replace an initial hypothesis despite recognizing contradictory findings. On DeepSeek-Chat, a single agent with the same state outperforms the full system on macro recall.
%U https://aclanthology.org/2026.findings-aacl.193/
%P 3821-3832
Markdown (Informal)
[DiagBench: Evaluating Diagnostic Revision in Large Language Models with Longitudinal Inpatient Records](https://aclanthology.org/2026.findings-aacl.193/) (Feng et al., Findings 2026)
ACL