<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="betala-etal-2025-picture">
    <titleInfo>
        <title>A Picture is Worth a Thousand (Correct) Captions: A Vision-Guided Judge-Corrector System for Multimodal Machine Translation</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Siddharth</namePart>
        <namePart type="family">Betala</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Kushan</namePart>
        <namePart type="family">Raj</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Vipul</namePart>
        <namePart type="family">Betala</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Rohan</namePart>
        <namePart type="family">Saswade</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-12</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the Twelfth Workshop on Asian Translation (WAT 2025)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Toshiaki</namePart>
            <namePart type="family">Nakazawa</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Isao</namePart>
            <namePart type="family">Goto</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Mumbai, India</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-309-8</identifier>
    </relatedItem>
    <abstract>In this paper, we describe our system under the team name BLEU Monday for the English-to-Indic Multimodal Translation Task at WAT 2025. We participate in the text-only translation tasks for English-Hindi, English-Bengali, English-Malayalam, and English-Odia language pairs. We present a two-stage approach that addresses quality issues in the training data through automated error detection and correction, followed by parameter-efficient model fine-tuning.Our methodology introduces a vision-augmented judge-corrector pipeline that leverages multimodal language models to systematically identify and correct translation errors in the training data. The judge component classifies translations into three categories: correct, visually ambiguous (requiring image context), or mistranslated (poor translation quality). Identified errors are routed to specialized correctors: GPT-4o-mini regenerates captions requiring visual disambiguation, while IndicTrans2 retranslates cases with pure translation quality issues. This automated pipeline processes 28,928 training examples across four languages, correcting an average of 17.1% of captions per language.We then apply Low-Rank Adaptation (LoRA) to fine-tune the IndicTrans2 en-indic 200M distilled model on both original and corrected datasets. Training on corrected data yields consistent improvements, with BLEU score gains of +1.30 for English-Bengali on the evaluation set (42.00 → 43.30) and +0.70 on the challenge set (44.90 → 45.60), +0.60 for English-Odia on the evaluation set (41.00 → 41.60), and +0.10 for English-Hindi on the challenge set (53.90 → 54.00).</abstract>
    <identifier type="citekey">betala-etal-2025-picture</identifier>
    <location>
        <url>https://aclanthology.org/2025.wat-1.13/</url>
    </location>
    <part>
        <date>2025-12</date>
        <extent unit="page">
            <start>124</start>
            <end>137</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T A Picture is Worth a Thousand (Correct) Captions: A Vision-Guided Judge-Corrector System for Multimodal Machine Translation
%A Betala, Siddharth
%A Raj, Kushan
%A Betala, Vipul
%A Saswade, Rohan
%Y Nakazawa, Toshiaki
%Y Goto, Isao
%S Proceedings of the Twelfth Workshop on Asian Translation (WAT 2025)
%D 2025
%8 December
%I Association for Computational Linguistics
%C Mumbai, India
%@ 979-8-89176-309-8
%F betala-etal-2025-picture
%X In this paper, we describe our system under the team name BLEU Monday for the English-to-Indic Multimodal Translation Task at WAT 2025. We participate in the text-only translation tasks for English-Hindi, English-Bengali, English-Malayalam, and English-Odia language pairs. We present a two-stage approach that addresses quality issues in the training data through automated error detection and correction, followed by parameter-efficient model fine-tuning.Our methodology introduces a vision-augmented judge-corrector pipeline that leverages multimodal language models to systematically identify and correct translation errors in the training data. The judge component classifies translations into three categories: correct, visually ambiguous (requiring image context), or mistranslated (poor translation quality). Identified errors are routed to specialized correctors: GPT-4o-mini regenerates captions requiring visual disambiguation, while IndicTrans2 retranslates cases with pure translation quality issues. This automated pipeline processes 28,928 training examples across four languages, correcting an average of 17.1% of captions per language.We then apply Low-Rank Adaptation (LoRA) to fine-tune the IndicTrans2 en-indic 200M distilled model on both original and corrected datasets. Training on corrected data yields consistent improvements, with BLEU score gains of +1.30 for English-Bengali on the evaluation set (42.00 → 43.30) and +0.70 on the challenge set (44.90 → 45.60), +0.60 for English-Odia on the evaluation set (41.00 → 41.60), and +0.10 for English-Hindi on the challenge set (53.90 → 54.00).
%U https://aclanthology.org/2025.wat-1.13/
%P 124-137

Markdown (Informal)

[A Picture is Worth a Thousand (Correct) Captions: A Vision-Guided Judge-Corrector System for Multimodal Machine Translation](https://aclanthology.org/2025.wat-1.13/) (Betala et al., WAT 2025)

• A Picture is Worth a Thousand (Correct) Captions: A Vision-Guided Judge-Corrector System for Multimodal Machine Translation (Betala et al., WAT 2025)

ACL

• Siddharth Betala, Kushan Raj, Vipul Betala, and Rohan Saswade. 2025. A Picture is Worth a Thousand (Correct) Captions: A Vision-Guided Judge-Corrector System for Multimodal Machine Translation. In Proceedings of the Twelfth Workshop on Asian Translation (WAT 2025), pages 124–137, Mumbai, India. Association for Computational Linguistics.