Dataset Card for Ubuntu-ToM: African Social Cognition Benchmark
Dataset Summary
Ubuntu-ToM is a highly specialized, programmatic evaluation benchmark consisting of 1,200 unique, culturally grounded scenarios. It is specifically designed to evaluate Large Language Models (LLMs) on non-Western social cognition, focusing on African communal cultures and the philosophical principles of Ubuntu ("I am because we are").
Traditional Theory of Mind (ToM) benchmarks heavily… See the full description on the dataset page: https://huggingface.co/datasets/gimmy256/ubuntu_tom_social_cognition.
