AINeutralarXiv – CS AI · 6h ago6/10
🧠
NRITYAM: Language Models Meet Art and Heritage of Dance
Researchers have introduced NRITYAM, a comprehensive multilingual benchmark dataset containing 9,260 question-answer pairs across 12 languages designed to evaluate how well language models understand global dance traditions and cultural heritage. Developed in collaboration with native dance artists and speakers, the dataset addresses a critical gap in AI evaluation by testing cultural comprehension beyond Western-centric knowledge, establishing new standards for assessing AI systems' ability to reason about traditional performing arts.