@inproceedings{popov-etal-2026-yandex,
title = "{Y}andex at {WMT}26: Translation Adaptation at 235{B} and Hard-Label Distillation to 8{B}",
author = "Popov, Dmitrii and
Kozlova, Elizaveta and
Taracheva, Ekaterina and
Makhotina, Elizaveta and
Mekhraliev, Artem and
Baratelia, Miron and
Enikeeva, Ekaterina and
Karpachev, Nikolay",
editor = "Haddow, Barry and
Kocmi, Tom and
Koehn, Philipp and
Monz, Christof",
booktitle = "Proceedings of the Eleventh Conference on Machine Translation",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.wmt-main.77/",
pages = "1397--1404",
ISBN = "979-8-89176-474-3",
abstract = "We describe Yandex and Yandex-8B, our unconstrained and constrained submissions to the WMT26 General Machine Translation task. Our previous WMT25 submission, a 7B translation specialist, shows strong Fluency but comparatively weaker Accuracy. To combine its translation behavior with the capacity of a much larger model without replaying the full 7B training pipeline at 235B scale, we add two forms of supervision to the instruction-tuning data of an in-house 235B pretrained model: direct translation targets generated by the 7B specialist and in-house translation-related tasks. The resulting Yandex model improves both Accuracy and Fluency and reduces the LLM-based error score relative to the same 235B model tuned only on general instructions. Relative to the 7B specialist, it improves Accuracy and reduces the error score, while their Fluency point estimates remain similar. Yandex then provides hard targets for Yandex-8B, which is trained for English-to-Russian, English-to-Belarusian, English-to-Kazakh, and English-to-Armenian using direct and instruction-conditioned translation data. Across four directions, Yandex-8B obtains the highest paragraph-level ChrF++ and the highest Accuracy and Fluency under our LLM-based evaluation among five public baselines."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="popov-etal-2026-yandex">
<titleInfo>
<title>Yandex at WMT26: Translation Adaptation at 235B and Hard-Label Distillation to 8B</title>
</titleInfo>
<name type="personal">
<namePart type="given">Dmitrii</namePart>
<namePart type="family">Popov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elizaveta</namePart>
<namePart type="family">Kozlova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Taracheva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elizaveta</namePart>
<namePart type="family">Makhotina</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Artem</namePart>
<namePart type="family">Mekhraliev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Miron</namePart>
<namePart type="family">Baratelia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Enikeeva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nikolay</namePart>
<namePart type="family">Karpachev</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 Eleventh 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">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-474-3</identifier>
</relatedItem>
<abstract>We describe Yandex and Yandex-8B, our unconstrained and constrained submissions to the WMT26 General Machine Translation task. Our previous WMT25 submission, a 7B translation specialist, shows strong Fluency but comparatively weaker Accuracy. To combine its translation behavior with the capacity of a much larger model without replaying the full 7B training pipeline at 235B scale, we add two forms of supervision to the instruction-tuning data of an in-house 235B pretrained model: direct translation targets generated by the 7B specialist and in-house translation-related tasks. The resulting Yandex model improves both Accuracy and Fluency and reduces the LLM-based error score relative to the same 235B model tuned only on general instructions. Relative to the 7B specialist, it improves Accuracy and reduces the error score, while their Fluency point estimates remain similar. Yandex then provides hard targets for Yandex-8B, which is trained for English-to-Russian, English-to-Belarusian, English-to-Kazakh, and English-to-Armenian using direct and instruction-conditioned translation data. Across four directions, Yandex-8B obtains the highest paragraph-level ChrF++ and the highest Accuracy and Fluency under our LLM-based evaluation among five public baselines.</abstract>
<identifier type="citekey">popov-etal-2026-yandex</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.77/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>1397</start>
<end>1404</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Yandex at WMT26: Translation Adaptation at 235B and Hard-Label Distillation to 8B
%A Popov, Dmitrii
%A Kozlova, Elizaveta
%A Taracheva, Ekaterina
%A Makhotina, Elizaveta
%A Mekhraliev, Artem
%A Baratelia, Miron
%A Enikeeva, Ekaterina
%A Karpachev, Nikolay
%Y Haddow, Barry
%Y Kocmi, Tom
%Y Koehn, Philipp
%Y Monz, Christof
%S Proceedings of the Eleventh Conference on Machine Translation
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-474-3
%F popov-etal-2026-yandex
%X We describe Yandex and Yandex-8B, our unconstrained and constrained submissions to the WMT26 General Machine Translation task. Our previous WMT25 submission, a 7B translation specialist, shows strong Fluency but comparatively weaker Accuracy. To combine its translation behavior with the capacity of a much larger model without replaying the full 7B training pipeline at 235B scale, we add two forms of supervision to the instruction-tuning data of an in-house 235B pretrained model: direct translation targets generated by the 7B specialist and in-house translation-related tasks. The resulting Yandex model improves both Accuracy and Fluency and reduces the LLM-based error score relative to the same 235B model tuned only on general instructions. Relative to the 7B specialist, it improves Accuracy and reduces the error score, while their Fluency point estimates remain similar. Yandex then provides hard targets for Yandex-8B, which is trained for English-to-Russian, English-to-Belarusian, English-to-Kazakh, and English-to-Armenian using direct and instruction-conditioned translation data. Across four directions, Yandex-8B obtains the highest paragraph-level ChrF++ and the highest Accuracy and Fluency under our LLM-based evaluation among five public baselines.
%U https://aclanthology.org/2026.wmt-main.77/
%P 1397-1404
Markdown (Informal)
[Yandex at WMT26: Translation Adaptation at 235B and Hard-Label Distillation to 8B](https://aclanthology.org/2026.wmt-main.77/) (Popov et al., WMT 2026)
ACL
- Dmitrii Popov, Elizaveta Kozlova, Ekaterina Taracheva, Elizaveta Makhotina, Artem Mekhraliev, Miron Baratelia, Ekaterina Enikeeva, and Nikolay Karpachev. 2026. Yandex at WMT26: Translation Adaptation at 235B and Hard-Label Distillation to 8B. In Proceedings of the Eleventh Conference on Machine Translation, pages 1397–1404, Budapest, Hungary. Association for Computational Linguistics.