@inproceedings{sharma-etal-2026-nlp,
title = "{NLP}-{IIT} Patna at {WMT} 2026: Corpus-Driven Adaptation of Multilingual {MT} Models for Low-Resource {A}rabic{--}{A}sian Machine Translation",
author = "Sharma, Pushkar and
Ahtasam, Mo and
Singh, Kshetrimayum Boynao and
Kumar, Deepak and
Ekbal, Asif",
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.175/",
pages = "2329--2337",
ISBN = "979-8-89176-474-3",
abstract = "We investigate low-resource Arabic-Asian machine translation across six Arabic-centric translation directions: Arabic{\ensuremath{\leftrightarrow}}English, Arabic{\ensuremath{\leftrightarrow}}Hindi, and Arabic{\ensuremath{\leftrightarrow}}Urdu. Fine-tuned MADLAD-400 ranks first in four of the six directions on the official primary leaderboard and places within the top three in the remaining two. We fine-tune three pretrained multilingual models, MADLAD-400, NLLB-200, and GemmaX2, and compare them with their zero-shot counterparts using BLEU, chrF2++, COMET-22, and TER. Across all language pairs, supervised fine-tuning consistently outperforms zero-shot inference. Among the evaluated models, fine-tuned MADLAD-400 achieves the strongest overall performance, NLLB-200 delivers competitive results, and GemmaX2-28-9B exhibits the largest relative improvement over its zero-shot baseline. These findings demonstrate the effectiveness of supervised adaptation for low-resource Arabic-Asian machine translation and highlight MADLAD-400 as the most robust model across the evaluated translation directions."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sharma-etal-2026-nlp">
<titleInfo>
<title>NLP-IIT Patna at WMT 2026: Corpus-Driven Adaptation of Multilingual MT Models for Low-Resource Arabic–Asian Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pushkar</namePart>
<namePart type="family">Sharma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mo</namePart>
<namePart type="family">Ahtasam</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kshetrimayum</namePart>
<namePart type="given">Boynao</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Deepak</namePart>
<namePart type="family">Kumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Asif</namePart>
<namePart type="family">Ekbal</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 investigate low-resource Arabic-Asian machine translation across six Arabic-centric translation directions: Arabic\ensuremathłeftrightarrowEnglish, Arabic\ensuremathłeftrightarrowHindi, and Arabic\ensuremathłeftrightarrowUrdu. Fine-tuned MADLAD-400 ranks first in four of the six directions on the official primary leaderboard and places within the top three in the remaining two. We fine-tune three pretrained multilingual models, MADLAD-400, NLLB-200, and GemmaX2, and compare them with their zero-shot counterparts using BLEU, chrF2++, COMET-22, and TER. Across all language pairs, supervised fine-tuning consistently outperforms zero-shot inference. Among the evaluated models, fine-tuned MADLAD-400 achieves the strongest overall performance, NLLB-200 delivers competitive results, and GemmaX2-28-9B exhibits the largest relative improvement over its zero-shot baseline. These findings demonstrate the effectiveness of supervised adaptation for low-resource Arabic-Asian machine translation and highlight MADLAD-400 as the most robust model across the evaluated translation directions.</abstract>
<identifier type="citekey">sharma-etal-2026-nlp</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.175/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2329</start>
<end>2337</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T NLP-IIT Patna at WMT 2026: Corpus-Driven Adaptation of Multilingual MT Models for Low-Resource Arabic–Asian Machine Translation
%A Sharma, Pushkar
%A Ahtasam, Mo
%A Singh, Kshetrimayum Boynao
%A Kumar, Deepak
%A Ekbal, Asif
%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 sharma-etal-2026-nlp
%X We investigate low-resource Arabic-Asian machine translation across six Arabic-centric translation directions: Arabic\ensuremathłeftrightarrowEnglish, Arabic\ensuremathłeftrightarrowHindi, and Arabic\ensuremathłeftrightarrowUrdu. Fine-tuned MADLAD-400 ranks first in four of the six directions on the official primary leaderboard and places within the top three in the remaining two. We fine-tune three pretrained multilingual models, MADLAD-400, NLLB-200, and GemmaX2, and compare them with their zero-shot counterparts using BLEU, chrF2++, COMET-22, and TER. Across all language pairs, supervised fine-tuning consistently outperforms zero-shot inference. Among the evaluated models, fine-tuned MADLAD-400 achieves the strongest overall performance, NLLB-200 delivers competitive results, and GemmaX2-28-9B exhibits the largest relative improvement over its zero-shot baseline. These findings demonstrate the effectiveness of supervised adaptation for low-resource Arabic-Asian machine translation and highlight MADLAD-400 as the most robust model across the evaluated translation directions.
%U https://aclanthology.org/2026.wmt-main.175/
%P 2329-2337
Markdown (Informal)
[NLP-IIT Patna at WMT 2026: Corpus-Driven Adaptation of Multilingual MT Models for Low-Resource Arabic–Asian Machine Translation](https://aclanthology.org/2026.wmt-main.175/) (Sharma et al., WMT 2026)
ACL