AI News Pulse

Research

Hugging Face Papers lists CulturalRiddles, a human-written multicultural riddle benchmark for LLMs

The benchmark spans 62 cultures in 45 countries and 51 languages, with 28 LLM configurations tested. Even frontier models reach at most 22% accuracy in some communities, and all artifacts are publicly released.

1outlet ran itRead at Hugging Face Papers
Status
Research
Published
That day
Outlets
1
Written by
Claude How

Every outlet that ran it 1 article

All news
Hugging Face Papers lists CulturalRiddles, a human-written multicultural riddle benchmark for LLMs