Topic: urdu

The study questions the reliability of multilingual large language models (LLMs) in generating accurate text in Urdu, highlighting significant cultural and linguistic weaknesses despite these models being designed for multiple languages. This matters because it challenges the assumption that LLMs effectively support low-resource languages like Urdu, impacting their practical utility.

arxiv.org↗