@inproceedings{mansouri-etal-2026-verisim,
title = "{V}eri{S}im: A Configurable Framework for Stress-Testing Medical {AI} Under Patient Communication Noise",
author = "Mansouri, Sina and
Marvania, Mohit and
Shihorkar, Vibhavari and
Tran, Han Ngoc and
Shafiei, Kazhal and
Fazli, Mehrdad and
Li, Yikuan and
Zhu, Ziwei",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.2109/",
pages = "44933--44958",
ISBN = "979-8-89176-462-0",
abstract = "Medical large language models are typically evaluated on idealized patient cases that do not reflect how real patients communicate. We introduce VeriSim, a patient simulation framework that injects controllable noise along six clinically grounded communication dimensions while substantially preserving each patient{'}s medical record. Truth adherence is supported by a verifier that extracts atomic claims from each candidate utterance and judges them against a UMLS-grounded vector index built with BioLORD embeddings, using the retrieved atoms' structured clinical metadata (e.g., drug class, anatomical site, treats-condition relations) rather than surface-text similarity alone. Across seven open-weight LLMs, realistic noise reduces diagnostic accuracy by 15-25 percentage points and increases conversation length by 34-55{\%}; the 7-8B models degrade 1.4x more than 70B+ models. A board-certified physician and a licensed nurse rate VeriSim{'}s conversations highly on truth, realism, clinical utility, and noise fidelity (inter-annotator agreement {\ensuremath{>}}=0.80 across all dimensions), and an LLM-as-judge closely tracks their ratings, enabling scalable evaluation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mansouri-etal-2026-verisim">
<titleInfo>
<title>VeriSim: A Configurable Framework for Stress-Testing Medical AI Under Patient Communication Noise</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sina</namePart>
<namePart type="family">Mansouri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Marvania</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vibhavari</namePart>
<namePart type="family">Shihorkar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Han</namePart>
<namePart type="given">Ngoc</namePart>
<namePart type="family">Tran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kazhal</namePart>
<namePart type="family">Shafiei</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mehrdad</namePart>
<namePart type="family">Fazli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yikuan</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ziwei</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Medical large language models are typically evaluated on idealized patient cases that do not reflect how real patients communicate. We introduce VeriSim, a patient simulation framework that injects controllable noise along six clinically grounded communication dimensions while substantially preserving each patient’s medical record. Truth adherence is supported by a verifier that extracts atomic claims from each candidate utterance and judges them against a UMLS-grounded vector index built with BioLORD embeddings, using the retrieved atoms’ structured clinical metadata (e.g., drug class, anatomical site, treats-condition relations) rather than surface-text similarity alone. Across seven open-weight LLMs, realistic noise reduces diagnostic accuracy by 15-25 percentage points and increases conversation length by 34-55%; the 7-8B models degrade 1.4x more than 70B+ models. A board-certified physician and a licensed nurse rate VeriSim’s conversations highly on truth, realism, clinical utility, and noise fidelity (inter-annotator agreement \ensuremath>=0.80 across all dimensions), and an LLM-as-judge closely tracks their ratings, enabling scalable evaluation.</abstract>
<identifier type="citekey">mansouri-etal-2026-verisim</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2109/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>44933</start>
<end>44958</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T VeriSim: A Configurable Framework for Stress-Testing Medical AI Under Patient Communication Noise
%A Mansouri, Sina
%A Marvania, Mohit
%A Shihorkar, Vibhavari
%A Tran, Han Ngoc
%A Shafiei, Kazhal
%A Fazli, Mehrdad
%A Li, Yikuan
%A Zhu, Ziwei
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F mansouri-etal-2026-verisim
%X Medical large language models are typically evaluated on idealized patient cases that do not reflect how real patients communicate. We introduce VeriSim, a patient simulation framework that injects controllable noise along six clinically grounded communication dimensions while substantially preserving each patient’s medical record. Truth adherence is supported by a verifier that extracts atomic claims from each candidate utterance and judges them against a UMLS-grounded vector index built with BioLORD embeddings, using the retrieved atoms’ structured clinical metadata (e.g., drug class, anatomical site, treats-condition relations) rather than surface-text similarity alone. Across seven open-weight LLMs, realistic noise reduces diagnostic accuracy by 15-25 percentage points and increases conversation length by 34-55%; the 7-8B models degrade 1.4x more than 70B+ models. A board-certified physician and a licensed nurse rate VeriSim’s conversations highly on truth, realism, clinical utility, and noise fidelity (inter-annotator agreement \ensuremath>=0.80 across all dimensions), and an LLM-as-judge closely tracks their ratings, enabling scalable evaluation.
%U https://aclanthology.org/2026.findings-emnlp.2109/
%P 44933-44958
Markdown (Informal)
[VeriSim: A Configurable Framework for Stress-Testing Medical AI Under Patient Communication Noise](https://aclanthology.org/2026.findings-emnlp.2109/) (Mansouri et al., Findings 2026)
ACL
- Sina Mansouri, Mohit Marvania, Vibhavari Shihorkar, Han Ngoc Tran, Kazhal Shafiei, Mehrdad Fazli, Yikuan Li, and Ziwei Zhu. 2026. VeriSim: A Configurable Framework for Stress-Testing Medical AI Under Patient Communication Noise. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 44933–44958, Budapest, Hungary. Association for Computational Linguistics.