@inproceedings{yeroushalmi-etal-2026-beyond,
title = "Beyond Accuracy: Improving {LLM}s' Science Communication Skills",
author = "Yeroushalmi, Mattan and
Bragilovski, Maxim and
Grinberg, Nir",
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.309/",
pages = "6538--6559",
ISBN = "979-8-89176-462-0",
abstract = "Science as a body of knowledge is important for human learning, yet beyond factual accuracy, little research has focused on the capacity of Large Language Models (LLMs) to communicate science effectively to the public. Here, we develop and validate a computational framework for measuring science communication skills of LLMs. Using this framework, we evaluate different alignment approaches: fine-tuning using top-ranked human explanations, distilling synthetic explanations from a teacher model, and preference optimization. Our results show that annotators overwhelmingly prefer LLM-generated science explanations over top-ranked organic answers, with this preference primarily attributed to better structure and readability. We find that distillation with synthetic answers achieves the most significant and stable improvement, allowing a smaller open-source model to achieve statistical parity with the proprietary teacher model in blind side-by-side comparisons."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yeroushalmi-etal-2026-beyond">
<titleInfo>
<title>Beyond Accuracy: Improving LLMs’ Science Communication Skills</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mattan</namePart>
<namePart type="family">Yeroushalmi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maxim</namePart>
<namePart type="family">Bragilovski</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nir</namePart>
<namePart type="family">Grinberg</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>Science as a body of knowledge is important for human learning, yet beyond factual accuracy, little research has focused on the capacity of Large Language Models (LLMs) to communicate science effectively to the public. Here, we develop and validate a computational framework for measuring science communication skills of LLMs. Using this framework, we evaluate different alignment approaches: fine-tuning using top-ranked human explanations, distilling synthetic explanations from a teacher model, and preference optimization. Our results show that annotators overwhelmingly prefer LLM-generated science explanations over top-ranked organic answers, with this preference primarily attributed to better structure and readability. We find that distillation with synthetic answers achieves the most significant and stable improvement, allowing a smaller open-source model to achieve statistical parity with the proprietary teacher model in blind side-by-side comparisons.</abstract>
<identifier type="citekey">yeroushalmi-etal-2026-beyond</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.309/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>6538</start>
<end>6559</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Beyond Accuracy: Improving LLMs’ Science Communication Skills
%A Yeroushalmi, Mattan
%A Bragilovski, Maxim
%A Grinberg, Nir
%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 yeroushalmi-etal-2026-beyond
%X Science as a body of knowledge is important for human learning, yet beyond factual accuracy, little research has focused on the capacity of Large Language Models (LLMs) to communicate science effectively to the public. Here, we develop and validate a computational framework for measuring science communication skills of LLMs. Using this framework, we evaluate different alignment approaches: fine-tuning using top-ranked human explanations, distilling synthetic explanations from a teacher model, and preference optimization. Our results show that annotators overwhelmingly prefer LLM-generated science explanations over top-ranked organic answers, with this preference primarily attributed to better structure and readability. We find that distillation with synthetic answers achieves the most significant and stable improvement, allowing a smaller open-source model to achieve statistical parity with the proprietary teacher model in blind side-by-side comparisons.
%U https://aclanthology.org/2026.findings-emnlp.309/
%P 6538-6559
Markdown (Informal)
[Beyond Accuracy: Improving LLMs’ Science Communication Skills](https://aclanthology.org/2026.findings-emnlp.309/) (Yeroushalmi et al., Findings 2026)
ACL