@inproceedings{tabassum-etal-2026-urdummlu,
title = "{U}rdu{MMLU}: A Massive Multitask Benchmark for {U}rdu Language Understanding",
author = "Tabassum, Ahmer and
Ahmad, Sarfraz and
Iqbal, Hasan and
Aijaz, Owais and
Ahsan, Momina and
Nakov, Preslav",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.450/",
pages = "10341--10370",
ISBN = "979-8-89176-459-0",
abstract = "Meaningful multilingual evaluation must test models in the target language and educational context. Urdu, spoken by more than 230 million people, lacks a broad MMLU-style benchmark built from native educational sources. We introduce UrduMMLU, a benchmark of 26,389 Urdu MCQs across 26 subjects and five domains, collected from native Urdu MCQ banks and public examination PDFs. Unlike translation-based benchmarks, UrduMMLU combines academic subjects with content specific to Urdu and regional education. We label the exam-derived portion through dual human annotation with strict consensus filtering. We evaluate 30 LLMs under English and Urdu prompts, yielding 60 zero-shot evaluations, and further evaluate four open-source LLMs under multiple few-shot settings across both prompt languages. Gemini-3.5-Flash performs best, reaching 90.23{\%} and 90.45{\%} accuracy, while no other model exceeds 85{\%}. The strongest open-source model trails by 7.78 and 9.12 points, and many models lose 25 to 40 points on Urdu-centered Humanities subjects compared with STEM. Few-shot prompting yields only modest gains. Results on UrduMMLU show that current LLMs have uneven Urdu knowledge, particularly for content grounded in the regional context."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tabassum-etal-2026-urdummlu">
<titleInfo>
<title>UrduMMLU: A Massive Multitask Benchmark for Urdu Language Understanding</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ahmer</namePart>
<namePart type="family">Tabassum</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sarfraz</namePart>
<namePart type="family">Ahmad</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hasan</namePart>
<namePart type="family">Iqbal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Owais</namePart>
<namePart type="family">Aijaz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Momina</namePart>
<namePart type="family">Ahsan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Preslav</namePart>
<namePart type="family">Nakov</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 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Meaningful multilingual evaluation must test models in the target language and educational context. Urdu, spoken by more than 230 million people, lacks a broad MMLU-style benchmark built from native educational sources. We introduce UrduMMLU, a benchmark of 26,389 Urdu MCQs across 26 subjects and five domains, collected from native Urdu MCQ banks and public examination PDFs. Unlike translation-based benchmarks, UrduMMLU combines academic subjects with content specific to Urdu and regional education. We label the exam-derived portion through dual human annotation with strict consensus filtering. We evaluate 30 LLMs under English and Urdu prompts, yielding 60 zero-shot evaluations, and further evaluate four open-source LLMs under multiple few-shot settings across both prompt languages. Gemini-3.5-Flash performs best, reaching 90.23% and 90.45% accuracy, while no other model exceeds 85%. The strongest open-source model trails by 7.78 and 9.12 points, and many models lose 25 to 40 points on Urdu-centered Humanities subjects compared with STEM. Few-shot prompting yields only modest gains. Results on UrduMMLU show that current LLMs have uneven Urdu knowledge, particularly for content grounded in the regional context.</abstract>
<identifier type="citekey">tabassum-etal-2026-urdummlu</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.450/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>10341</start>
<end>10370</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T UrduMMLU: A Massive Multitask Benchmark for Urdu Language Understanding
%A Tabassum, Ahmer
%A Ahmad, Sarfraz
%A Iqbal, Hasan
%A Aijaz, Owais
%A Ahsan, Momina
%A Nakov, Preslav
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F tabassum-etal-2026-urdummlu
%X Meaningful multilingual evaluation must test models in the target language and educational context. Urdu, spoken by more than 230 million people, lacks a broad MMLU-style benchmark built from native educational sources. We introduce UrduMMLU, a benchmark of 26,389 Urdu MCQs across 26 subjects and five domains, collected from native Urdu MCQ banks and public examination PDFs. Unlike translation-based benchmarks, UrduMMLU combines academic subjects with content specific to Urdu and regional education. We label the exam-derived portion through dual human annotation with strict consensus filtering. We evaluate 30 LLMs under English and Urdu prompts, yielding 60 zero-shot evaluations, and further evaluate four open-source LLMs under multiple few-shot settings across both prompt languages. Gemini-3.5-Flash performs best, reaching 90.23% and 90.45% accuracy, while no other model exceeds 85%. The strongest open-source model trails by 7.78 and 9.12 points, and many models lose 25 to 40 points on Urdu-centered Humanities subjects compared with STEM. Few-shot prompting yields only modest gains. Results on UrduMMLU show that current LLMs have uneven Urdu knowledge, particularly for content grounded in the regional context.
%U https://aclanthology.org/2026.emnlp-main.450/
%P 10341-10370
Markdown (Informal)
[UrduMMLU: A Massive Multitask Benchmark for Urdu Language Understanding](https://aclanthology.org/2026.emnlp-main.450/) (Tabassum et al., EMNLP 2026)
ACL