@inproceedings{van-huy-etal-2026-medspa,
title = "{M}ed{SPA}: Self-planning Medical Agents",
author = "Van Huy, Luu and
Wang, Bowen and
Okura, Fumio and
Nakashima, Yuta",
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.1368/",
pages = "31200--31220",
ISBN = "979-8-89176-459-0",
abstract = "Reliable medical diagnosis requires a precise final prediction, which is strictly supported by observable evidence. Recent large vision-language models perform well on medical image understanding tasks, yet they often lack explicit control over which evidence to inspect before making a prediction. To facilitate flexible planning of evidence exploration, we propose MedSPA, a self-planning medical reasoning framework that separates diagnostic evidence acquisition from evidence-grounded prediction with two cooperative agents: a planning-and-reasoning agent that adaptively decides what to inspect and when to stop, and a summary agent that generates the final output from the accumulated evidence. MedSPA distills medical knowledge from strict diagnostic procedures but acquires flexibility through reinforcement learning to generate optimal outputs. Experiments on medical report generation and medical visual question answering show that MedSPA improves clinical performance, faithfulness, and controllability over existing reasoning baselines."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="van-huy-etal-2026-medspa">
<titleInfo>
<title>MedSPA: Self-planning Medical Agents</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luu</namePart>
<namePart type="family">Van Huy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bowen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fumio</namePart>
<namePart type="family">Okura</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuta</namePart>
<namePart type="family">Nakashima</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>Reliable medical diagnosis requires a precise final prediction, which is strictly supported by observable evidence. Recent large vision-language models perform well on medical image understanding tasks, yet they often lack explicit control over which evidence to inspect before making a prediction. To facilitate flexible planning of evidence exploration, we propose MedSPA, a self-planning medical reasoning framework that separates diagnostic evidence acquisition from evidence-grounded prediction with two cooperative agents: a planning-and-reasoning agent that adaptively decides what to inspect and when to stop, and a summary agent that generates the final output from the accumulated evidence. MedSPA distills medical knowledge from strict diagnostic procedures but acquires flexibility through reinforcement learning to generate optimal outputs. Experiments on medical report generation and medical visual question answering show that MedSPA improves clinical performance, faithfulness, and controllability over existing reasoning baselines.</abstract>
<identifier type="citekey">van-huy-etal-2026-medspa</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1368/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>31200</start>
<end>31220</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T MedSPA: Self-planning Medical Agents
%A Van Huy, Luu
%A Wang, Bowen
%A Okura, Fumio
%A Nakashima, Yuta
%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 van-huy-etal-2026-medspa
%X Reliable medical diagnosis requires a precise final prediction, which is strictly supported by observable evidence. Recent large vision-language models perform well on medical image understanding tasks, yet they often lack explicit control over which evidence to inspect before making a prediction. To facilitate flexible planning of evidence exploration, we propose MedSPA, a self-planning medical reasoning framework that separates diagnostic evidence acquisition from evidence-grounded prediction with two cooperative agents: a planning-and-reasoning agent that adaptively decides what to inspect and when to stop, and a summary agent that generates the final output from the accumulated evidence. MedSPA distills medical knowledge from strict diagnostic procedures but acquires flexibility through reinforcement learning to generate optimal outputs. Experiments on medical report generation and medical visual question answering show that MedSPA improves clinical performance, faithfulness, and controllability over existing reasoning baselines.
%U https://aclanthology.org/2026.emnlp-main.1368/
%P 31200-31220
Markdown (Informal)
[MedSPA: Self-planning Medical Agents](https://aclanthology.org/2026.emnlp-main.1368/) (Van Huy et al., EMNLP 2026)
ACL
- Luu Van Huy, Bowen Wang, Fumio Okura, and Yuta Nakashima. 2026. MedSPA: Self-planning Medical Agents. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 31200–31220, Budapest, Hungary. Association for Computational Linguistics.