Paper
Multilingual Large Language Models do not comprehend all natural languages to equal degrees
Authors
Natalia Moskvina, Raquel Montero, Masaya Yoshida, Ferdy Hubers, Paolo Morosi, Walid Irhaymi, Jin Yan, Tamara Serrano, Elena Pagliarini, Fritz Günther, Evelina Leivada
Abstract
Large Language Models (LLMs) play a critical role in how humans access information. While their core use relies on comprehending written requests, our understanding of this ability is currently limited, because most benchmarks evaluate LLMs in high-resource languages predominantly spoken by Western, Educated, Industrialised, Rich, and Democratic (WEIRD) communities. The default assumption is that English is the best-performing language for LLMs, while smaller, low-resource languages are linked to less reliable outputs, even in multilingual, state-of-the-art models. To track variation in the comprehension abilities of LLMs, we prompt 3 popular models on a language comprehension task across 12 languages, representing the Indo-European, Afro-Asiatic, Turkic, Sino-Tibetan, and Japonic language families. Our results suggest that the models exhibit remarkable linguistic accuracy across typologically diverse languages, yet they fall behind human baselines in all of them, albeit to different degrees. Contrary to what was expected, English is not the best-performing language, as it was systematically outperformed by several Romance languages, even lower-resource ones. We frame the results by discussing the role of several factors that drive LLM performance, such as tokenization, language distance from Spanish and English, size of training data, and data origin in high- vs. low-resource languages and WEIRD vs. non-WEIRD communities.
Metadata
Related papers
Vibe Coding XR: Accelerating AI + XR Prototyping with XR Blocks and Gemini
Ruofei Du, Benjamin Hersh, David Li, Nels Numan, Xun Qian, Yanhe Chen, Zhongy... • 2026-03-25
Comparing Developer and LLM Biases in Code Evaluation
Aditya Mittal, Ryan Shar, Zichu Wu, Shyam Agarwal, Tongshuang Wu, Chris Donah... • 2026-03-25
The Stochastic Gap: A Markovian Framework for Pre-Deployment Reliability and Oversight-Cost Auditing in Agentic Artificial Intelligence
Biplab Pal, Santanu Bhattacharya • 2026-03-25
Retrieval Improvements Do Not Guarantee Better Answers: A Study of RAG for AI Policy QA
Saahil Mathur, Ryan David Rittner, Vedant Ajit Thakur, Daniel Stuart Schiff, ... • 2026-03-25
MARCH: Multi-Agent Reinforced Self-Check for LLM Hallucination
Zhuo Li, Yupeng Zhang, Pengyu Cheng, Jiajun Song, Mengyu Zhou, Hao Li, Shujie... • 2026-03-25
Raw Data (Debug)
{
"raw_xml": "<entry>\n <id>http://arxiv.org/abs/2602.20065v1</id>\n <title>Multilingual Large Language Models do not comprehend all natural languages to equal degrees</title>\n <updated>2026-02-23T17:22:46Z</updated>\n <link href='https://arxiv.org/abs/2602.20065v1' rel='alternate' type='text/html'/>\n <link href='https://arxiv.org/pdf/2602.20065v1' rel='related' title='pdf' type='application/pdf'/>\n <summary>Large Language Models (LLMs) play a critical role in how humans access information. While their core use relies on comprehending written requests, our understanding of this ability is currently limited, because most benchmarks evaluate LLMs in high-resource languages predominantly spoken by Western, Educated, Industrialised, Rich, and Democratic (WEIRD) communities. The default assumption is that English is the best-performing language for LLMs, while smaller, low-resource languages are linked to less reliable outputs, even in multilingual, state-of-the-art models. To track variation in the comprehension abilities of LLMs, we prompt 3 popular models on a language comprehension task across 12 languages, representing the Indo-European, Afro-Asiatic, Turkic, Sino-Tibetan, and Japonic language families. Our results suggest that the models exhibit remarkable linguistic accuracy across typologically diverse languages, yet they fall behind human baselines in all of them, albeit to different degrees. Contrary to what was expected, English is not the best-performing language, as it was systematically outperformed by several Romance languages, even lower-resource ones. We frame the results by discussing the role of several factors that drive LLM performance, such as tokenization, language distance from Spanish and English, size of training data, and data origin in high- vs. low-resource languages and WEIRD vs. non-WEIRD communities.</summary>\n <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n <published>2026-02-23T17:22:46Z</published>\n <arxiv:comment>36 pages, 3 figures, 2 tables, 4 supplementary tables</arxiv:comment>\n <arxiv:primary_category term='cs.CL'/>\n <author>\n <name>Natalia Moskvina</name>\n </author>\n <author>\n <name>Raquel Montero</name>\n </author>\n <author>\n <name>Masaya Yoshida</name>\n </author>\n <author>\n <name>Ferdy Hubers</name>\n </author>\n <author>\n <name>Paolo Morosi</name>\n </author>\n <author>\n <name>Walid Irhaymi</name>\n </author>\n <author>\n <name>Jin Yan</name>\n </author>\n <author>\n <name>Tamara Serrano</name>\n </author>\n <author>\n <name>Elena Pagliarini</name>\n </author>\n <author>\n <name>Fritz Günther</name>\n </author>\n <author>\n <name>Evelina Leivada</name>\n </author>\n </entry>"
}