@inproceedings{o-reilly-etal-2026-semantic,
title = "Semantic Averaging with {MU}ltiple Languages ({SAMUL})",
author = "O{'}Reilly, Cliff and
Jim{\'e}nez-Ruiz, Ernesto and
Weyde, Tillman",
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.1827/",
doi = "10.18653/2026.findings-emnlp.1827",
pages = "39013--39034",
ISBN = "979-8-89176-462-0",
abstract = "We propose Semantic Averaging with MUltiple Languages (SAMUL) as a method to isolate concept semantics of text by averaging concept activations derived via Sparse Autoencoders (SAEs) for multiple text versions in different languages. We combine autoencoder activations to derive a conceptual average using different approaches. In our experiments with SAMUL, we start with English text from benchmark datasets (Ontology Alignment Evaluation Initiative 2025: conference track, SemEval2014: cross-level semantic similarity, Sentences Involving Compositional Knowledge: relatedness) and automatically translate into French, Chinese, German and Japanese. We obtain SAE feature activations for each class and language version using autoencoders, specifically the open source Gemma Scope 1 and OpenAI{'}s GPT2 suites of Sparse Autoencoders and average them to SAMUL vectors. We measure performance in predicting semantic similarity of short texts using predefined ground truth or correlation to the ground truth mapping m between ontology classes in the case of the ontology alignment task. Our results show a consistent improvement over any single translated language, and improvement over English alone in 27 of 32 comparisons. Conceptual averaging therefore yields sparse representations that align more closely with human similarity judgements, suggesting a route to more reliable use of SAE features in semantic tasks such as ontology alignment. Our implementation and experimentation code is openly available."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="o-reilly-etal-2026-semantic">
<titleInfo>
<title>Semantic Averaging with MUltiple Languages (SAMUL)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Cliff</namePart>
<namePart type="family">O’Reilly</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ernesto</namePart>
<namePart type="family">Jiménez-Ruiz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tillman</namePart>
<namePart type="family">Weyde</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>We propose Semantic Averaging with MUltiple Languages (SAMUL) as a method to isolate concept semantics of text by averaging concept activations derived via Sparse Autoencoders (SAEs) for multiple text versions in different languages. We combine autoencoder activations to derive a conceptual average using different approaches. In our experiments with SAMUL, we start with English text from benchmark datasets (Ontology Alignment Evaluation Initiative 2025: conference track, SemEval2014: cross-level semantic similarity, Sentences Involving Compositional Knowledge: relatedness) and automatically translate into French, Chinese, German and Japanese. We obtain SAE feature activations for each class and language version using autoencoders, specifically the open source Gemma Scope 1 and OpenAI’s GPT2 suites of Sparse Autoencoders and average them to SAMUL vectors. We measure performance in predicting semantic similarity of short texts using predefined ground truth or correlation to the ground truth mapping m between ontology classes in the case of the ontology alignment task. Our results show a consistent improvement over any single translated language, and improvement over English alone in 27 of 32 comparisons. Conceptual averaging therefore yields sparse representations that align more closely with human similarity judgements, suggesting a route to more reliable use of SAE features in semantic tasks such as ontology alignment. Our implementation and experimentation code is openly available.</abstract>
<identifier type="citekey">o-reilly-etal-2026-semantic</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.1827</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1827/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>39013</start>
<end>39034</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Semantic Averaging with MUltiple Languages (SAMUL)
%A O’Reilly, Cliff
%A Jiménez-Ruiz, Ernesto
%A Weyde, Tillman
%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 o-reilly-etal-2026-semantic
%X We propose Semantic Averaging with MUltiple Languages (SAMUL) as a method to isolate concept semantics of text by averaging concept activations derived via Sparse Autoencoders (SAEs) for multiple text versions in different languages. We combine autoencoder activations to derive a conceptual average using different approaches. In our experiments with SAMUL, we start with English text from benchmark datasets (Ontology Alignment Evaluation Initiative 2025: conference track, SemEval2014: cross-level semantic similarity, Sentences Involving Compositional Knowledge: relatedness) and automatically translate into French, Chinese, German and Japanese. We obtain SAE feature activations for each class and language version using autoencoders, specifically the open source Gemma Scope 1 and OpenAI’s GPT2 suites of Sparse Autoencoders and average them to SAMUL vectors. We measure performance in predicting semantic similarity of short texts using predefined ground truth or correlation to the ground truth mapping m between ontology classes in the case of the ontology alignment task. Our results show a consistent improvement over any single translated language, and improvement over English alone in 27 of 32 comparisons. Conceptual averaging therefore yields sparse representations that align more closely with human similarity judgements, suggesting a route to more reliable use of SAE features in semantic tasks such as ontology alignment. Our implementation and experimentation code is openly available.
%R 10.18653/2026.findings-emnlp.1827
%U https://aclanthology.org/2026.findings-emnlp.1827/
%U https://doi.org/10.18653/2026.findings-emnlp.1827
%P 39013-39034
Markdown (Informal)
[Semantic Averaging with MUltiple Languages (SAMUL)](https://aclanthology.org/2026.findings-emnlp.1827/) (O’Reilly et al., Findings 2026)
ACL
- Cliff O’Reilly, Ernesto Jiménez-Ruiz, and Tillman Weyde. 2026. Semantic Averaging with MUltiple Languages (SAMUL). In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 39013–39034, Budapest, Hungary. Association for Computational Linguistics.