@inproceedings{ruan-etal-2026-llm-driven,
title = "An {LLM}-Driven Pipeline for Extremely Low-Resource {E}nglish-to-{I}ndic Translation {---} From Vocabulary Injection to Post-Training and Decoding",
author = "Ruan, Lu and
Jia, Shaoying and
Ning, Fan and
Wang, Wei and
Cai, Zhengzhen and
Hu, Fei and
Hu, Heng and
Wang, Chenzi",
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.179/",
pages = "2360--2367",
ISBN = "979-8-89176-474-3",
abstract = "This submission addresses WMT26 Category 2 for five extremely low-resource English-to-X translation directions: Bodo, Karbi, Kokborok, Nagamese, and Tagin. We develop an LLM-driven pipeline in which large language models are used both to construct bilingual lexical resources and as translation models. First, an English seed dictionary and qwen3-max are used to build word-level bilingual vocabularies, whose top candidates are injected into translation prompts as a ``Helpful Vocabulary'' block. On an 80/20 train-held-out split of the Tagin data, vocabulary injection is associated with a 7.09-point BLEU improvement and a 49.35-point TER reduction. Second, we compare LoRA adaptation of Qwen2.5-32B-Instruct with full-parameter fine-tuning of Hunyuan-MT-7B, with optional DPO. Hunyuan performs better in the reported Bodo and Kokborok comparisons, including an 11.87-point ChrF advantage on Bodo. Third, post-hoc repeated-ngram and length truncation, combined with per-sentence selection of the shorter of two decoding outputs, mitigates generation degeneration and improves Bodo BLEU from 19.27 to 27.18 without retraining. We submit two contrastive systems per direction: a take-shorter system with output truncation for Bodo, and augmented and baseline SFT systems for the other directions. Code and configurations are publicly available."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ruan-etal-2026-llm-driven">
<titleInfo>
<title>An LLM-Driven Pipeline for Extremely Low-Resource English-to-Indic Translation — From Vocabulary Injection to Post-Training and Decoding</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Ruan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shaoying</namePart>
<namePart type="family">Jia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fan</namePart>
<namePart type="family">Ning</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhengzhen</namePart>
<namePart type="family">Cai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fei</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Heng</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chenzi</namePart>
<namePart type="family">Wang</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 submission addresses WMT26 Category 2 for five extremely low-resource English-to-X translation directions: Bodo, Karbi, Kokborok, Nagamese, and Tagin. We develop an LLM-driven pipeline in which large language models are used both to construct bilingual lexical resources and as translation models. First, an English seed dictionary and qwen3-max are used to build word-level bilingual vocabularies, whose top candidates are injected into translation prompts as a “Helpful Vocabulary” block. On an 80/20 train-held-out split of the Tagin data, vocabulary injection is associated with a 7.09-point BLEU improvement and a 49.35-point TER reduction. Second, we compare LoRA adaptation of Qwen2.5-32B-Instruct with full-parameter fine-tuning of Hunyuan-MT-7B, with optional DPO. Hunyuan performs better in the reported Bodo and Kokborok comparisons, including an 11.87-point ChrF advantage on Bodo. Third, post-hoc repeated-ngram and length truncation, combined with per-sentence selection of the shorter of two decoding outputs, mitigates generation degeneration and improves Bodo BLEU from 19.27 to 27.18 without retraining. We submit two contrastive systems per direction: a take-shorter system with output truncation for Bodo, and augmented and baseline SFT systems for the other directions. Code and configurations are publicly available.</abstract>
<identifier type="citekey">ruan-etal-2026-llm-driven</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.179/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2360</start>
<end>2367</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T An LLM-Driven Pipeline for Extremely Low-Resource English-to-Indic Translation — From Vocabulary Injection to Post-Training and Decoding
%A Ruan, Lu
%A Jia, Shaoying
%A Ning, Fan
%A Wang, Wei
%A Cai, Zhengzhen
%A Hu, Fei
%A Hu, Heng
%A Wang, Chenzi
%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 ruan-etal-2026-llm-driven
%X This submission addresses WMT26 Category 2 for five extremely low-resource English-to-X translation directions: Bodo, Karbi, Kokborok, Nagamese, and Tagin. We develop an LLM-driven pipeline in which large language models are used both to construct bilingual lexical resources and as translation models. First, an English seed dictionary and qwen3-max are used to build word-level bilingual vocabularies, whose top candidates are injected into translation prompts as a “Helpful Vocabulary” block. On an 80/20 train-held-out split of the Tagin data, vocabulary injection is associated with a 7.09-point BLEU improvement and a 49.35-point TER reduction. Second, we compare LoRA adaptation of Qwen2.5-32B-Instruct with full-parameter fine-tuning of Hunyuan-MT-7B, with optional DPO. Hunyuan performs better in the reported Bodo and Kokborok comparisons, including an 11.87-point ChrF advantage on Bodo. Third, post-hoc repeated-ngram and length truncation, combined with per-sentence selection of the shorter of two decoding outputs, mitigates generation degeneration and improves Bodo BLEU from 19.27 to 27.18 without retraining. We submit two contrastive systems per direction: a take-shorter system with output truncation for Bodo, and augmented and baseline SFT systems for the other directions. Code and configurations are publicly available.
%U https://aclanthology.org/2026.wmt-main.179/
%P 2360-2367
Markdown (Informal)
[An LLM-Driven Pipeline for Extremely Low-Resource English-to-Indic Translation — From Vocabulary Injection to Post-Training and Decoding](https://aclanthology.org/2026.wmt-main.179/) (Ruan et al., WMT 2026)
ACL
- Lu Ruan, Shaoying Jia, Fan Ning, Wei Wang, Zhengzhen Cai, Fei Hu, Heng Hu, and Chenzi Wang. 2026. An LLM-Driven Pipeline for Extremely Low-Resource English-to-Indic Translation — From Vocabulary Injection to Post-Training and Decoding. In Proceedings of the Eleventh Conference on Machine Translation, pages 2360–2367, Budapest, Hungary. Association for Computational Linguistics.