@inproceedings{finkelstein-etal-2025-google,
title = "{G}oogle {T}ranslate{'}s Research Submission to {WMT}2025",
author = "Finkelstein, Mara and
Kovacs, Geza and
Caswell, Isaac and
Domhan, Tobias and
Peter, Jan-Thorsten and
Juraska, Juraj and
Freitag, Markus and
Vilar, David",
editor = "Haddow, Barry and
Kocmi, Tom and
Koehn, Philipp and
Monz, Christof",
booktitle = "Proceedings of the Tenth Conference on Machine Translation",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.wmt-1.48/",
pages = "723--731",
ISBN = "979-8-89176-341-8",
abstract = "Large Language Models have shown impressive multilingual capabilities, where translation is one among many tasks. Google Translate{'}s submission to the 2025 WMT evaluation tries to research how these models behave when pushing their translation performance to the limit. Starting with the strong Gemma 3 model, we carry out supervised fine tuning on high quality, synthetically generated parallel data. Afterwards we perform an additional reinforcement learning step, with reward models based on translation metrics to push the translation capabilities even further. Controlling the combination of reward models, including reference-based and quality estimation metrics, we found that the behaviour of the model could be tailored towards a more literal or more creative translation style. Our two submissions correspond to those two models. We chose the more creative system as our primary submission, targetting a human preference for better sounding, more naturally flowing text, although at the risk of losing on the accuracy of the translation. It is an open question to find the sweet spot between these two dimensions, which certainly will depend on the specific domain to handle and user preferences."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="finkelstein-etal-2025-google">
<titleInfo>
<title>Google Translate’s Research Submission to WMT2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mara</namePart>
<namePart type="family">Finkelstein</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Geza</namePart>
<namePart type="family">Kovacs</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Isaac</namePart>
<namePart type="family">Caswell</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tobias</namePart>
<namePart type="family">Domhan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jan-Thorsten</namePart>
<namePart type="family">Peter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Juraj</namePart>
<namePart type="family">Juraska</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Markus</namePart>
<namePart type="family">Freitag</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Vilar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Tenth Conference on Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Barry</namePart>
<namePart type="family">Haddow</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tom</namePart>
<namePart type="family">Kocmi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Philipp</namePart>
<namePart type="family">Koehn</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christof</namePart>
<namePart type="family">Monz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-341-8</identifier>
</relatedItem>
<abstract>Large Language Models have shown impressive multilingual capabilities, where translation is one among many tasks. Google Translate’s submission to the 2025 WMT evaluation tries to research how these models behave when pushing their translation performance to the limit. Starting with the strong Gemma 3 model, we carry out supervised fine tuning on high quality, synthetically generated parallel data. Afterwards we perform an additional reinforcement learning step, with reward models based on translation metrics to push the translation capabilities even further. Controlling the combination of reward models, including reference-based and quality estimation metrics, we found that the behaviour of the model could be tailored towards a more literal or more creative translation style. Our two submissions correspond to those two models. We chose the more creative system as our primary submission, targetting a human preference for better sounding, more naturally flowing text, although at the risk of losing on the accuracy of the translation. It is an open question to find the sweet spot between these two dimensions, which certainly will depend on the specific domain to handle and user preferences.</abstract>
<identifier type="citekey">finkelstein-etal-2025-google</identifier>
<location>
<url>https://aclanthology.org/2025.wmt-1.48/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>723</start>
<end>731</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Google Translate’s Research Submission to WMT2025
%A Finkelstein, Mara
%A Kovacs, Geza
%A Caswell, Isaac
%A Domhan, Tobias
%A Peter, Jan-Thorsten
%A Juraska, Juraj
%A Freitag, Markus
%A Vilar, David
%Y Haddow, Barry
%Y Kocmi, Tom
%Y Koehn, Philipp
%Y Monz, Christof
%S Proceedings of the Tenth Conference on Machine Translation
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-341-8
%F finkelstein-etal-2025-google
%X Large Language Models have shown impressive multilingual capabilities, where translation is one among many tasks. Google Translate’s submission to the 2025 WMT evaluation tries to research how these models behave when pushing their translation performance to the limit. Starting with the strong Gemma 3 model, we carry out supervised fine tuning on high quality, synthetically generated parallel data. Afterwards we perform an additional reinforcement learning step, with reward models based on translation metrics to push the translation capabilities even further. Controlling the combination of reward models, including reference-based and quality estimation metrics, we found that the behaviour of the model could be tailored towards a more literal or more creative translation style. Our two submissions correspond to those two models. We chose the more creative system as our primary submission, targetting a human preference for better sounding, more naturally flowing text, although at the risk of losing on the accuracy of the translation. It is an open question to find the sweet spot between these two dimensions, which certainly will depend on the specific domain to handle and user preferences.
%U https://aclanthology.org/2025.wmt-1.48/
%P 723-731
Markdown (Informal)
[Google Translate’s Research Submission to WMT2025](https://aclanthology.org/2025.wmt-1.48/) (Finkelstein et al., WMT 2025)
ACL
- Mara Finkelstein, Geza Kovacs, Isaac Caswell, Tobias Domhan, Jan-Thorsten Peter, Juraj Juraska, Markus Freitag, and David Vilar. 2025. Google Translate’s Research Submission to WMT2025. In Proceedings of the Tenth Conference on Machine Translation, pages 723–731, Suzhou, China. Association for Computational Linguistics.