@inproceedings{sinha-etal-2026-jh,
title = "{JH}{\_}{NLP}{\_}{APAS}hreya2: {L}o{RA} Fine-Tuning of {NLLB}-200 for {E}nglish{--}{M}anipuri Machine Translation at {WMT} 2026",
author = "Sinha, Aparajita and
Agarwal, Monika and
Bhat, Shreya and
Bonal, Shreya and
Agrahari, Aryan",
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.163/",
pages = "2231--2238",
ISBN = "979-8-89176-474-3",
abstract = "This paper describes the submission of team JH{\_}NLP{\_}APAShreya2 to the English{--}Manipuri (Bengali-script) track of the WMT 2026 Low-Resource Indic Machine translation shared task. Our system adapts the multilingual NLLB-200-distilled-600M model using Low-Rank Adaptation (LoRA) on the official English-Manipuri parallel corpus. The corpus contains 23,687 sentence pairs. We used no external parallel or monolingual data, synthetic data, or back-translation. We conducted a progressive sequence of experiments with different data-pool sizes and training durations. The final model was trained for seven epochs and used to translate all 1,000 English source sentences in the official blind test set. The primary submission obtained 8.03 BLEU, 20.15 METEOR, 82.56 TER, 40.93 chrF++, 84.86 BERTScore, and 68.18 COMET. The system ranked within the top three primary systems according to five of the six reported evaluation metrics. These results indicate that parameter-efficient adaptation of multilingual models is a practical approach to low-resource English-Manipuri machine translation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sinha-etal-2026-jh">
<titleInfo>
<title>JH_NLP_APAShreya2: LoRA Fine-Tuning of NLLB-200 for English–Manipuri Machine Translation at WMT 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Aparajita</namePart>
<namePart type="family">Sinha</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Monika</namePart>
<namePart type="family">Agarwal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shreya</namePart>
<namePart type="family">Bhat</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shreya</namePart>
<namePart type="family">Bonal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aryan</namePart>
<namePart type="family">Agrahari</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>This paper describes the submission of team JH_NLP_APAShreya2 to the English–Manipuri (Bengali-script) track of the WMT 2026 Low-Resource Indic Machine translation shared task. Our system adapts the multilingual NLLB-200-distilled-600M model using Low-Rank Adaptation (LoRA) on the official English-Manipuri parallel corpus. The corpus contains 23,687 sentence pairs. We used no external parallel or monolingual data, synthetic data, or back-translation. We conducted a progressive sequence of experiments with different data-pool sizes and training durations. The final model was trained for seven epochs and used to translate all 1,000 English source sentences in the official blind test set. The primary submission obtained 8.03 BLEU, 20.15 METEOR, 82.56 TER, 40.93 chrF++, 84.86 BERTScore, and 68.18 COMET. The system ranked within the top three primary systems according to five of the six reported evaluation metrics. These results indicate that parameter-efficient adaptation of multilingual models is a practical approach to low-resource English-Manipuri machine translation.</abstract>
<identifier type="citekey">sinha-etal-2026-jh</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.163/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2231</start>
<end>2238</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T JH_NLP_APAShreya2: LoRA Fine-Tuning of NLLB-200 for English–Manipuri Machine Translation at WMT 2026
%A Sinha, Aparajita
%A Agarwal, Monika
%A Bhat, Shreya
%A Bonal, Shreya
%A Agrahari, Aryan
%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 sinha-etal-2026-jh
%X This paper describes the submission of team JH_NLP_APAShreya2 to the English–Manipuri (Bengali-script) track of the WMT 2026 Low-Resource Indic Machine translation shared task. Our system adapts the multilingual NLLB-200-distilled-600M model using Low-Rank Adaptation (LoRA) on the official English-Manipuri parallel corpus. The corpus contains 23,687 sentence pairs. We used no external parallel or monolingual data, synthetic data, or back-translation. We conducted a progressive sequence of experiments with different data-pool sizes and training durations. The final model was trained for seven epochs and used to translate all 1,000 English source sentences in the official blind test set. The primary submission obtained 8.03 BLEU, 20.15 METEOR, 82.56 TER, 40.93 chrF++, 84.86 BERTScore, and 68.18 COMET. The system ranked within the top three primary systems according to five of the six reported evaluation metrics. These results indicate that parameter-efficient adaptation of multilingual models is a practical approach to low-resource English-Manipuri machine translation.
%U https://aclanthology.org/2026.wmt-main.163/
%P 2231-2238
Markdown (Informal)
[JH_NLP_APAShreya2: LoRA Fine-Tuning of NLLB-200 for English–Manipuri Machine Translation at WMT 2026](https://aclanthology.org/2026.wmt-main.163/) (Sinha et al., WMT 2026)
ACL