@inproceedings{tamo-etal-2026-provena,
title = "{PROVENA}-{MED}: Benchmarking Provenance Faithfulness in Multimodal Clinical Reasoning",
author = "Tamo, J. Ben and
Yuan, Yining and
Marteau, Benoit Louis and
Nnamdi, Micky C. and
Lu, Yuxing and
Wang, May Dongmei",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2393/",
pages = "54405--54421",
ISBN = "979-8-89176-459-0",
abstract = "Clinical language and vision-language models increasingly generate diagnostic and treatment rationales, but those claims remain difficult to audit. Existing benchmarks usually score final answers over fully revealed records, without testing whether each claim is grounded in valid patient evidence or whether models can seek missing information efficiently. We introduce \textsc{Provena-Med}, a benchmark for evaluating provenance-faithful multimodal clinical reasoning over linked EHR tables, clinical notes, chest radiographs, and guideline rules. Models must produce a staged artifact comprising a problem list, differential, workup, therapy, and monitoring, in which each claim cites typed evidence units and may abstain when information is insufficient. \textsc{Provena-Med} supports both full-information evaluation and an interactive evidence-seeking track. We further propose MM-AIS, which measures citation validity, objective support under evidence-removal tests, salient-evidence recall, and, for open-weight models, sensitivity to cited evidence. Across ten open-weight models and five cohorts, the best attribution yield is only 0.42, citation hallucination is common, and diagnostic accuracy, safety, and provenance do not reliably co-vary. \textsc{Provena-Med} provides an auditable testbed for evaluating grounded clinical reasoning beyond final-answer accuracy."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tamo-etal-2026-provena">
<titleInfo>
<title>PROVENA-MED: Benchmarking Provenance Faithfulness in Multimodal Clinical Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">J</namePart>
<namePart type="given">Ben</namePart>
<namePart type="family">Tamo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yining</namePart>
<namePart type="family">Yuan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Benoit</namePart>
<namePart type="given">Louis</namePart>
<namePart type="family">Marteau</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Micky</namePart>
<namePart type="given">C</namePart>
<namePart type="family">Nnamdi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuxing</namePart>
<namePart type="family">Lu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">May</namePart>
<namePart type="given">Dongmei</namePart>
<namePart type="family">Wang</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Clinical language and vision-language models increasingly generate diagnostic and treatment rationales, but those claims remain difficult to audit. Existing benchmarks usually score final answers over fully revealed records, without testing whether each claim is grounded in valid patient evidence or whether models can seek missing information efficiently. We introduce Provena-Med, a benchmark for evaluating provenance-faithful multimodal clinical reasoning over linked EHR tables, clinical notes, chest radiographs, and guideline rules. Models must produce a staged artifact comprising a problem list, differential, workup, therapy, and monitoring, in which each claim cites typed evidence units and may abstain when information is insufficient. Provena-Med supports both full-information evaluation and an interactive evidence-seeking track. We further propose MM-AIS, which measures citation validity, objective support under evidence-removal tests, salient-evidence recall, and, for open-weight models, sensitivity to cited evidence. Across ten open-weight models and five cohorts, the best attribution yield is only 0.42, citation hallucination is common, and diagnostic accuracy, safety, and provenance do not reliably co-vary. Provena-Med provides an auditable testbed for evaluating grounded clinical reasoning beyond final-answer accuracy.</abstract>
<identifier type="citekey">tamo-etal-2026-provena</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2393/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>54405</start>
<end>54421</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T PROVENA-MED: Benchmarking Provenance Faithfulness in Multimodal Clinical Reasoning
%A Tamo, J. Ben
%A Yuan, Yining
%A Marteau, Benoit Louis
%A Nnamdi, Micky C.
%A Lu, Yuxing
%A Wang, May Dongmei
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F tamo-etal-2026-provena
%X Clinical language and vision-language models increasingly generate diagnostic and treatment rationales, but those claims remain difficult to audit. Existing benchmarks usually score final answers over fully revealed records, without testing whether each claim is grounded in valid patient evidence or whether models can seek missing information efficiently. We introduce Provena-Med, a benchmark for evaluating provenance-faithful multimodal clinical reasoning over linked EHR tables, clinical notes, chest radiographs, and guideline rules. Models must produce a staged artifact comprising a problem list, differential, workup, therapy, and monitoring, in which each claim cites typed evidence units and may abstain when information is insufficient. Provena-Med supports both full-information evaluation and an interactive evidence-seeking track. We further propose MM-AIS, which measures citation validity, objective support under evidence-removal tests, salient-evidence recall, and, for open-weight models, sensitivity to cited evidence. Across ten open-weight models and five cohorts, the best attribution yield is only 0.42, citation hallucination is common, and diagnostic accuracy, safety, and provenance do not reliably co-vary. Provena-Med provides an auditable testbed for evaluating grounded clinical reasoning beyond final-answer accuracy.
%U https://aclanthology.org/2026.emnlp-main.2393/
%P 54405-54421
Markdown (Informal)
[PROVENA-MED: Benchmarking Provenance Faithfulness in Multimodal Clinical Reasoning](https://aclanthology.org/2026.emnlp-main.2393/) (Tamo et al., EMNLP 2026)
ACL