@inproceedings{vilar-etal-2026-benchmark,
title = "Benchmark Data Contamination in Underrepresented Languages: A Comprehensive Analysis Using {B}razilian Data",
author = "Vilar, Iriedson Souto Maior de Moraes and
Maia, David Candeia and
Brunet, Jo{\~a}o and
Morais, Fabio and
Balby Marinho, Leandro",
editor = "Piperidis, Stelios and
Bel, N{\'u}ria and
van den Heuvel, Henk and
Ide, Nancy and
Krek, Simon and
Toral, Antonio",
booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resource Association",
url = "https://aclanthology.org/2026.lrec-1.374/",
doi = "10.63317/39wbjvajnh7t",
pages = "4765--4777",
abstract = "Large Language Models (LLMs) are typically evaluated using standardized benchmarks to enable consistent performance measurement and model comparison. However, the reliability of these benchmarks can be undermined by data contamination, which occurs when evaluation items are inadvertently included in training corpora. While this issue has been investigated primarily in high-resource languages such as English and Chinese, its impact on underrepresented languages {---} such as Brazilian Portuguese {---} remains understudied. In this paper, we present one of the first systematic investigations of benchmark data contamination (BDC) in an underrepresented language setting, using Brazilian Portuguese as a case study. Using validated methodologies from the literature, we evaluate specialized and multilingual models across four benchmarks: BLUEX, ENEM Challenge, OAB Exams, and HealthQA-BR. Our approach applyes TS-Guessing to detect contamination via memorized knowledge, alongside a 50-character n-gram similarity strategy to identify benchmark items leaked into training data. Our results provide consistent evidence of contamination, revealing that models with stronger memorization and retrieval abilities tend to achieve artificially inflated benchmark scores. Our contributions include: (i) classifying models according to their contamination risk, (ii) identifying the benchmarks most affected by data leakage, and (iii) reporting contaminated training corpora."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="vilar-etal-2026-benchmark">
<titleInfo>
<title>Benchmark Data Contamination in Underrepresented Languages: A Comprehensive Analysis Using Brazilian Data</title>
</titleInfo>
<name type="personal">
<namePart type="given">Iriedson</namePart>
<namePart type="given">Souto</namePart>
<namePart type="given">Maior</namePart>
<namePart type="given">de</namePart>
<namePart type="given">Moraes</namePart>
<namePart type="family">Vilar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="given">Candeia</namePart>
<namePart type="family">Maia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">João</namePart>
<namePart type="family">Brunet</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fabio</namePart>
<namePart type="family">Morais</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Leandro</namePart>
<namePart type="family">Balby Marinho</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fifteenth Language Resources and Evaluation Conference</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stelios</namePart>
<namePart type="family">Piperidis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Núria</namePart>
<namePart type="family">Bel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Henk</namePart>
<namePart type="family">van den Heuvel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nancy</namePart>
<namePart type="family">Ide</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Krek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Toral</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resource Association</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Large Language Models (LLMs) are typically evaluated using standardized benchmarks to enable consistent performance measurement and model comparison. However, the reliability of these benchmarks can be undermined by data contamination, which occurs when evaluation items are inadvertently included in training corpora. While this issue has been investigated primarily in high-resource languages such as English and Chinese, its impact on underrepresented languages — such as Brazilian Portuguese — remains understudied. In this paper, we present one of the first systematic investigations of benchmark data contamination (BDC) in an underrepresented language setting, using Brazilian Portuguese as a case study. Using validated methodologies from the literature, we evaluate specialized and multilingual models across four benchmarks: BLUEX, ENEM Challenge, OAB Exams, and HealthQA-BR. Our approach applyes TS-Guessing to detect contamination via memorized knowledge, alongside a 50-character n-gram similarity strategy to identify benchmark items leaked into training data. Our results provide consistent evidence of contamination, revealing that models with stronger memorization and retrieval abilities tend to achieve artificially inflated benchmark scores. Our contributions include: (i) classifying models according to their contamination risk, (ii) identifying the benchmarks most affected by data leakage, and (iii) reporting contaminated training corpora.</abstract>
<identifier type="citekey">vilar-etal-2026-benchmark</identifier>
<identifier type="doi">10.63317/39wbjvajnh7t</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.374/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>4765</start>
<end>4777</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Benchmark Data Contamination in Underrepresented Languages: A Comprehensive Analysis Using Brazilian Data
%A Vilar, Iriedson Souto Maior de Moraes
%A Maia, David Candeia
%A Brunet, João
%A Morais, Fabio
%A Balby Marinho, Leandro
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F vilar-etal-2026-benchmark
%X Large Language Models (LLMs) are typically evaluated using standardized benchmarks to enable consistent performance measurement and model comparison. However, the reliability of these benchmarks can be undermined by data contamination, which occurs when evaluation items are inadvertently included in training corpora. While this issue has been investigated primarily in high-resource languages such as English and Chinese, its impact on underrepresented languages — such as Brazilian Portuguese — remains understudied. In this paper, we present one of the first systematic investigations of benchmark data contamination (BDC) in an underrepresented language setting, using Brazilian Portuguese as a case study. Using validated methodologies from the literature, we evaluate specialized and multilingual models across four benchmarks: BLUEX, ENEM Challenge, OAB Exams, and HealthQA-BR. Our approach applyes TS-Guessing to detect contamination via memorized knowledge, alongside a 50-character n-gram similarity strategy to identify benchmark items leaked into training data. Our results provide consistent evidence of contamination, revealing that models with stronger memorization and retrieval abilities tend to achieve artificially inflated benchmark scores. Our contributions include: (i) classifying models according to their contamination risk, (ii) identifying the benchmarks most affected by data leakage, and (iii) reporting contaminated training corpora.
%R 10.63317/39wbjvajnh7t
%U https://aclanthology.org/2026.lrec-1.374/
%U https://doi.org/10.63317/39wbjvajnh7t
%P 4765-4777
Markdown (Informal)
[Benchmark Data Contamination in Underrepresented Languages: A Comprehensive Analysis Using Brazilian Data](https://aclanthology.org/2026.lrec-1.374/) (Vilar et al., LREC 2026)
ACL