@inproceedings{vamvas-etal-2026-translation,
title = "Translation Asymmetry in {LLM}s as a Data Augmentation Factor: A Case Study for 6 {R}omansh Language Varieties",
author = "Vamvas, Jannis and
P{\'e}rez Prat, Ignacio and
Heldstab, Angela and
Fischer, Dominic P. and
Ahmadi, Sina and
Sennrich, Rico",
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.230/",
pages = "4837--4861",
ISBN = "979-8-89176-462-0",
abstract = "Recent strategies for low-resource machine translation rely on LLMs to generate synthetic data based on text in higher-resource languages. We revisit this idea for Romansh, a language with 6 distinct varieties. LLMs tend to confuse these varieties when translating into Romansh, but they are quite good at translating out of Romansh into a high-resource language such as German. Due to this asymmetry, the direction of data augmentation is a crucial choice. We find that contrary to recent strategies, creating synthetic translations into the higher-resource language is the superior approach, and only this approach allows us to surpass a Gemini 3 Pro baseline on German-Romansh translation (+23 BLEU over Gemini in the lowest-resource variety). A human evaluation confirms that our experiments yield the first model that generates fluent translations in the individual Romansh varieties."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="vamvas-etal-2026-translation">
<titleInfo>
<title>Translation Asymmetry in LLMs as a Data Augmentation Factor: A Case Study for 6 Romansh Language Varieties</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jannis</namePart>
<namePart type="family">Vamvas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ignacio</namePart>
<namePart type="family">Pérez Prat</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Angela</namePart>
<namePart type="family">Heldstab</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dominic</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Fischer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sina</namePart>
<namePart type="family">Ahmadi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rico</namePart>
<namePart type="family">Sennrich</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>Recent strategies for low-resource machine translation rely on LLMs to generate synthetic data based on text in higher-resource languages. We revisit this idea for Romansh, a language with 6 distinct varieties. LLMs tend to confuse these varieties when translating into Romansh, but they are quite good at translating out of Romansh into a high-resource language such as German. Due to this asymmetry, the direction of data augmentation is a crucial choice. We find that contrary to recent strategies, creating synthetic translations into the higher-resource language is the superior approach, and only this approach allows us to surpass a Gemini 3 Pro baseline on German-Romansh translation (+23 BLEU over Gemini in the lowest-resource variety). A human evaluation confirms that our experiments yield the first model that generates fluent translations in the individual Romansh varieties.</abstract>
<identifier type="citekey">vamvas-etal-2026-translation</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.230/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>4837</start>
<end>4861</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Translation Asymmetry in LLMs as a Data Augmentation Factor: A Case Study for 6 Romansh Language Varieties
%A Vamvas, Jannis
%A Pérez Prat, Ignacio
%A Heldstab, Angela
%A Fischer, Dominic P.
%A Ahmadi, Sina
%A Sennrich, Rico
%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 vamvas-etal-2026-translation
%X Recent strategies for low-resource machine translation rely on LLMs to generate synthetic data based on text in higher-resource languages. We revisit this idea for Romansh, a language with 6 distinct varieties. LLMs tend to confuse these varieties when translating into Romansh, but they are quite good at translating out of Romansh into a high-resource language such as German. Due to this asymmetry, the direction of data augmentation is a crucial choice. We find that contrary to recent strategies, creating synthetic translations into the higher-resource language is the superior approach, and only this approach allows us to surpass a Gemini 3 Pro baseline on German-Romansh translation (+23 BLEU over Gemini in the lowest-resource variety). A human evaluation confirms that our experiments yield the first model that generates fluent translations in the individual Romansh varieties.
%U https://aclanthology.org/2026.findings-emnlp.230/
%P 4837-4861
Markdown (Informal)
[Translation Asymmetry in LLMs as a Data Augmentation Factor: A Case Study for 6 Romansh Language Varieties](https://aclanthology.org/2026.findings-emnlp.230/) (Vamvas et al., Findings 2026)
ACL