@inproceedings{alotaibi-2026-openbracket,
title = "{O}pen{B}racket at {WMT} 2026: {COMET}-{MBR} System Combination over {A}rabic-Native and Multilingual Models for Low-Resource {A}rabic-{E}nglish Translation",
author = "Alotaibi, Abrar",
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.167/",
pages = "2266--2272",
ISBN = "979-8-89176-474-3",
abstract = "OpenBracket submits constrained systems to both English-paired directions of the WMT 2026 Low-Resource Arabic-Asian Language Translation task, trained on the official data only (21,000 pairs) from publicly available checkpoints. Per direction we submit a primary COMET-22 MBR system combination over five fine-tuned models (NLLB-3.3B, two X-ALMA variants, ALLaM-7B, OPUS-MT) plus two contrastive systems: an ALLaM-7B fine-tune and sampling-based MBR over NLLB-3.3B. On the official test set, our primary ranks first among all teams in Arabic-to-English (BLEU 33.83, chrF 59.72, TER 57.00, COMET 82.49), and in English-to-Arabic it attains the highest COMET on the primary leaderboard (BLEU 21.81, chrF 54.99, TER 69.29, COMET 85.23)."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="alotaibi-2026-openbracket">
<titleInfo>
<title>OpenBracket at WMT 2026: COMET-MBR System Combination over Arabic-Native and Multilingual Models for Low-Resource Arabic-English Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Abrar</namePart>
<namePart type="family">Alotaibi</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>OpenBracket submits constrained systems to both English-paired directions of the WMT 2026 Low-Resource Arabic-Asian Language Translation task, trained on the official data only (21,000 pairs) from publicly available checkpoints. Per direction we submit a primary COMET-22 MBR system combination over five fine-tuned models (NLLB-3.3B, two X-ALMA variants, ALLaM-7B, OPUS-MT) plus two contrastive systems: an ALLaM-7B fine-tune and sampling-based MBR over NLLB-3.3B. On the official test set, our primary ranks first among all teams in Arabic-to-English (BLEU 33.83, chrF 59.72, TER 57.00, COMET 82.49), and in English-to-Arabic it attains the highest COMET on the primary leaderboard (BLEU 21.81, chrF 54.99, TER 69.29, COMET 85.23).</abstract>
<identifier type="citekey">alotaibi-2026-openbracket</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.167/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2266</start>
<end>2272</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T OpenBracket at WMT 2026: COMET-MBR System Combination over Arabic-Native and Multilingual Models for Low-Resource Arabic-English Translation
%A Alotaibi, Abrar
%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 alotaibi-2026-openbracket
%X OpenBracket submits constrained systems to both English-paired directions of the WMT 2026 Low-Resource Arabic-Asian Language Translation task, trained on the official data only (21,000 pairs) from publicly available checkpoints. Per direction we submit a primary COMET-22 MBR system combination over five fine-tuned models (NLLB-3.3B, two X-ALMA variants, ALLaM-7B, OPUS-MT) plus two contrastive systems: an ALLaM-7B fine-tune and sampling-based MBR over NLLB-3.3B. On the official test set, our primary ranks first among all teams in Arabic-to-English (BLEU 33.83, chrF 59.72, TER 57.00, COMET 82.49), and in English-to-Arabic it attains the highest COMET on the primary leaderboard (BLEU 21.81, chrF 54.99, TER 69.29, COMET 85.23).
%U https://aclanthology.org/2026.wmt-main.167/
%P 2266-2272
Markdown (Informal)
[OpenBracket at WMT 2026: COMET-MBR System Combination over Arabic-Native and Multilingual Models for Low-Resource Arabic-English Translation](https://aclanthology.org/2026.wmt-main.167/) (Alotaibi, WMT 2026)
ACL