UrduBench: An Urdu Reasoning Benchmark using Contextually Ensembled Translations with Human-in-the-Loop

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Shafique, Muhammad Ali, Mehboob, Areej, Fiaz, Layba, Qadeer, Muhammad Usman, Farooq, Hamza
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915760477044736
author Shafique, Muhammad Ali
Mehboob, Areej
Fiaz, Layba
Qadeer, Muhammad Usman
Farooq, Hamza
author_facet Shafique, Muhammad Ali
Mehboob, Areej
Fiaz, Layba
Qadeer, Muhammad Usman
Farooq, Hamza
contents Recent advances in large language models (LLMs) have led to strong reasoning capabilities; however, evaluating such models in low-resource languages remains challenging due to the lack of standardized benchmarks. In particular, Urdu reasoning evaluation has been limited by the sensitivity of machine translation and an emphasis on general language tasks rather than reasoning benchmarks. In this paper, we propose a contextually ensembled translation framework with human-in-the-loop validation that leverages multiple translation systems to develop Urdu reasoning benchmarks while preserving contextual and structural integrity. Using this framework, we translate widely adopted reasoning and question-answering benchmarks, including MGSM, MATH-500, CommonSenseQA, and OpenBookQA, into Urdu, collectively referred to as UrduBench, and conduct a comprehensive evaluation of both reasoning-oriented and instruction-tuned LLMs across multiple prompting strategies. Our analysis reveals performance differences across (1) four datasets, (2) five task difficulty levels, (3) diverse model architectures, (4) multiple model scaling settings, and (5) language consistency tests. We find that multi-step and symbolic reasoning tasks pose significant challenges in Urdu, and that stable language alignment is a critical prerequisite for robust reasoning. Overall, our work establishes a scalable methodology for standardized reasoning evaluation in Urdu and provides empirical insights into multilingual reasoning failures. This experimental setup is also broadly applicable to other low-resource languages. The code and datasets will be publicly released.
format Preprint
id arxiv_https___arxiv_org_abs_2601_21000
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle UrduBench: An Urdu Reasoning Benchmark using Contextually Ensembled Translations with Human-in-the-Loop
Shafique, Muhammad Ali
Mehboob, Areej
Fiaz, Layba
Qadeer, Muhammad Usman
Farooq, Hamza
Computation and Language
Artificial Intelligence
I.2.7; I.2.6
Recent advances in large language models (LLMs) have led to strong reasoning capabilities; however, evaluating such models in low-resource languages remains challenging due to the lack of standardized benchmarks. In particular, Urdu reasoning evaluation has been limited by the sensitivity of machine translation and an emphasis on general language tasks rather than reasoning benchmarks. In this paper, we propose a contextually ensembled translation framework with human-in-the-loop validation that leverages multiple translation systems to develop Urdu reasoning benchmarks while preserving contextual and structural integrity. Using this framework, we translate widely adopted reasoning and question-answering benchmarks, including MGSM, MATH-500, CommonSenseQA, and OpenBookQA, into Urdu, collectively referred to as UrduBench, and conduct a comprehensive evaluation of both reasoning-oriented and instruction-tuned LLMs across multiple prompting strategies. Our analysis reveals performance differences across (1) four datasets, (2) five task difficulty levels, (3) diverse model architectures, (4) multiple model scaling settings, and (5) language consistency tests. We find that multi-step and symbolic reasoning tasks pose significant challenges in Urdu, and that stable language alignment is a critical prerequisite for robust reasoning. Overall, our work establishes a scalable methodology for standardized reasoning evaluation in Urdu and provides empirical insights into multilingual reasoning failures. This experimental setup is also broadly applicable to other low-resource languages. The code and datasets will be publicly released.
title UrduBench: An Urdu Reasoning Benchmark using Contextually Ensembled Translations with Human-in-the-Loop
topic Computation and Language
Artificial Intelligence
I.2.7; I.2.6
url https://arxiv.org/abs/2601.21000