@inproceedings{zhang-etal-2026-agentic-reward,
title = "Agentic Reward Modeling for Medical Reasoning",
author = "Zhang, Hang and
Wang, Ruheng and
Ji, Yuelyu and
Kwak, Min Gu and
Wu, Xizhi and
Li, Chenyu and
Zhang, Li and
Shi, Wenqi and
Peng, Yifan and
Wang, Yanshan",
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.766/",
pages = "16152--16168",
ISBN = "979-8-89176-462-0",
abstract = "Large language models have achieved strong performance on medical reasoning benchmarks, yet their deployment in clinical settings demands rigorous verification to ensure factual accuracy. While reward models offer a scalable approach for reasoning trace verification, existing methods face two limitations: they produce only scalar reward values without explicit justification, and they rely on single-pass retrieval that precludes adaptive knowledge access as verification unfolds. We introduce Med-TIV, an agentic framework that addresses these limitations by training medical reward models to iteratively query external medical evidence during evaluation. Our approach combines tool-augmented verification with an iterative reinforcement learning paradigm that requires only trace-level supervision, alongside an adaptive curriculum mechanism that dynamically adjusts training data distribution. Across four medical reasoning benchmarks, Med-TIV achieves substantial gains over existing methods, improving MedQA accuracy by 23.5{\%} and MedXpertQA by 32.0{\%} relative to the base generator. Crucially, Med-TIV demonstrates an $8\times$ reduction in sampling budget compared to prior reward model baselines. These findings establish that agentic medical reward modeling offers a principled path toward more reliable medical reasoning systems."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-agentic-reward">
<titleInfo>
<title>Agentic Reward Modeling for Medical Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hang</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruheng</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuelyu</namePart>
<namePart type="family">Ji</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Min</namePart>
<namePart type="given">Gu</namePart>
<namePart type="family">Kwak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xizhi</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chenyu</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Li</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenqi</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yifan</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanshan</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>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>Large language models have achieved strong performance on medical reasoning benchmarks, yet their deployment in clinical settings demands rigorous verification to ensure factual accuracy. While reward models offer a scalable approach for reasoning trace verification, existing methods face two limitations: they produce only scalar reward values without explicit justification, and they rely on single-pass retrieval that precludes adaptive knowledge access as verification unfolds. We introduce Med-TIV, an agentic framework that addresses these limitations by training medical reward models to iteratively query external medical evidence during evaluation. Our approach combines tool-augmented verification with an iterative reinforcement learning paradigm that requires only trace-level supervision, alongside an adaptive curriculum mechanism that dynamically adjusts training data distribution. Across four medical reasoning benchmarks, Med-TIV achieves substantial gains over existing methods, improving MedQA accuracy by 23.5% and MedXpertQA by 32.0% relative to the base generator. Crucially, Med-TIV demonstrates an 8\times reduction in sampling budget compared to prior reward model baselines. These findings establish that agentic medical reward modeling offers a principled path toward more reliable medical reasoning systems.</abstract>
<identifier type="citekey">zhang-etal-2026-agentic-reward</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.766/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>16152</start>
<end>16168</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Agentic Reward Modeling for Medical Reasoning
%A Zhang, Hang
%A Wang, Ruheng
%A Ji, Yuelyu
%A Kwak, Min Gu
%A Wu, Xizhi
%A Li, Chenyu
%A Zhang, Li
%A Shi, Wenqi
%A Peng, Yifan
%A Wang, Yanshan
%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 zhang-etal-2026-agentic-reward
%X Large language models have achieved strong performance on medical reasoning benchmarks, yet their deployment in clinical settings demands rigorous verification to ensure factual accuracy. While reward models offer a scalable approach for reasoning trace verification, existing methods face two limitations: they produce only scalar reward values without explicit justification, and they rely on single-pass retrieval that precludes adaptive knowledge access as verification unfolds. We introduce Med-TIV, an agentic framework that addresses these limitations by training medical reward models to iteratively query external medical evidence during evaluation. Our approach combines tool-augmented verification with an iterative reinforcement learning paradigm that requires only trace-level supervision, alongside an adaptive curriculum mechanism that dynamically adjusts training data distribution. Across four medical reasoning benchmarks, Med-TIV achieves substantial gains over existing methods, improving MedQA accuracy by 23.5% and MedXpertQA by 32.0% relative to the base generator. Crucially, Med-TIV demonstrates an 8\times reduction in sampling budget compared to prior reward model baselines. These findings establish that agentic medical reward modeling offers a principled path toward more reliable medical reasoning systems.
%U https://aclanthology.org/2026.findings-emnlp.766/
%P 16152-16168
Markdown (Informal)
[Agentic Reward Modeling for Medical Reasoning](https://aclanthology.org/2026.findings-emnlp.766/) (Zhang et al., Findings 2026)
ACL
- Hang Zhang, Ruheng Wang, Yuelyu Ji, Min Gu Kwak, Xizhi Wu, Chenyu Li, Li Zhang, Wenqi Shi, Yifan Peng, and Yanshan Wang. 2026. Agentic Reward Modeling for Medical Reasoning. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 16152–16168, Budapest, Hungary. Association for Computational Linguistics.