@inproceedings{buess-etal-2026-polar,
title = "{POLAR}: Polarity-Aware and Anatomy-Conditioned Radiology Report Evaluation",
author = "Buess, Lukas and
Nguyen, Tri-Thien and
Wang, Xiaoliang and
Perez-Toro, Paula Andrea and
Arias-Vergara, Tomas and
Kainz, Bernhard and
Maier, Andreas",
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.2675/",
pages = "60854--60872",
ISBN = "979-8-89176-459-0",
abstract = "Automated radiology report generation is expanding across imaging modalities and anatomical regions, yet its evaluation remains misaligned with diagnostic priorities. Existing metrics either rely on surface-level text similarity or require slow, costly generative inference. This limits their ability to detect clinically significant errors, such as hallucinated or omitted abnormal findings, and prevents their use as efficient reward signals during training. We introduce POLAR (POlarity-Aware Long-context Anatomy-conditioned Regression), an encoder-based evaluation metric built on ModernBERT. POLAR combines long-context inputs for multi-organ reports, anatomy conditioning for targeted body-region evaluation, and polarity-aware regression that explicitly distinguishes normal from abnormal findings, yielding a more clinically meaningful assessment than prior metrics. Across real-world, radiologist-annotated CXR benchmarks and expert preference datasets, POLAR achieves competitive alignment with radiologist judgments while running 11x faster than generative evaluators, making it practical for large-scale benchmarking and as a training-time reward signal. We publicly release POLAR and the POLAR-Synth dataset, along with a fully open-source, privacy-preserving pipeline for data generation, training, and inference, enabling the metric to be extended to new modalities and anatomies."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="buess-etal-2026-polar">
<titleInfo>
<title>POLAR: Polarity-Aware and Anatomy-Conditioned Radiology Report Evaluation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lukas</namePart>
<namePart type="family">Buess</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tri-Thien</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaoliang</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paula</namePart>
<namePart type="given">Andrea</namePart>
<namePart type="family">Perez-Toro</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tomas</namePart>
<namePart type="family">Arias-Vergara</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bernhard</namePart>
<namePart type="family">Kainz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andreas</namePart>
<namePart type="family">Maier</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>Automated radiology report generation is expanding across imaging modalities and anatomical regions, yet its evaluation remains misaligned with diagnostic priorities. Existing metrics either rely on surface-level text similarity or require slow, costly generative inference. This limits their ability to detect clinically significant errors, such as hallucinated or omitted abnormal findings, and prevents their use as efficient reward signals during training. We introduce POLAR (POlarity-Aware Long-context Anatomy-conditioned Regression), an encoder-based evaluation metric built on ModernBERT. POLAR combines long-context inputs for multi-organ reports, anatomy conditioning for targeted body-region evaluation, and polarity-aware regression that explicitly distinguishes normal from abnormal findings, yielding a more clinically meaningful assessment than prior metrics. Across real-world, radiologist-annotated CXR benchmarks and expert preference datasets, POLAR achieves competitive alignment with radiologist judgments while running 11x faster than generative evaluators, making it practical for large-scale benchmarking and as a training-time reward signal. We publicly release POLAR and the POLAR-Synth dataset, along with a fully open-source, privacy-preserving pipeline for data generation, training, and inference, enabling the metric to be extended to new modalities and anatomies.</abstract>
<identifier type="citekey">buess-etal-2026-polar</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2675/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>60854</start>
<end>60872</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T POLAR: Polarity-Aware and Anatomy-Conditioned Radiology Report Evaluation
%A Buess, Lukas
%A Nguyen, Tri-Thien
%A Wang, Xiaoliang
%A Perez-Toro, Paula Andrea
%A Arias-Vergara, Tomas
%A Kainz, Bernhard
%A Maier, Andreas
%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 buess-etal-2026-polar
%X Automated radiology report generation is expanding across imaging modalities and anatomical regions, yet its evaluation remains misaligned with diagnostic priorities. Existing metrics either rely on surface-level text similarity or require slow, costly generative inference. This limits their ability to detect clinically significant errors, such as hallucinated or omitted abnormal findings, and prevents their use as efficient reward signals during training. We introduce POLAR (POlarity-Aware Long-context Anatomy-conditioned Regression), an encoder-based evaluation metric built on ModernBERT. POLAR combines long-context inputs for multi-organ reports, anatomy conditioning for targeted body-region evaluation, and polarity-aware regression that explicitly distinguishes normal from abnormal findings, yielding a more clinically meaningful assessment than prior metrics. Across real-world, radiologist-annotated CXR benchmarks and expert preference datasets, POLAR achieves competitive alignment with radiologist judgments while running 11x faster than generative evaluators, making it practical for large-scale benchmarking and as a training-time reward signal. We publicly release POLAR and the POLAR-Synth dataset, along with a fully open-source, privacy-preserving pipeline for data generation, training, and inference, enabling the metric to be extended to new modalities and anatomies.
%U https://aclanthology.org/2026.emnlp-main.2675/
%P 60854-60872
Markdown (Informal)
[POLAR: Polarity-Aware and Anatomy-Conditioned Radiology Report Evaluation](https://aclanthology.org/2026.emnlp-main.2675/) (Buess et al., EMNLP 2026)
ACL
- Lukas Buess, Tri-Thien Nguyen, Xiaoliang Wang, Paula Andrea Perez-Toro, Tomas Arias-Vergara, Bernhard Kainz, and Andreas Maier. 2026. POLAR: Polarity-Aware and Anatomy-Conditioned Radiology Report Evaluation. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 60854–60872, Budapest, Hungary. Association for Computational Linguistics.