Part of the AEGIS-1 collection: a large-scale effort to translate key medical AI datasets into underrepresented languages to advance global health data equity.
This dataset is a multilingual translation of HealthBench, OpenAI's medical AI evaluation benchmark, covering 32 languages. The original HealthBench consists of 5,000 realistic multi-turn clinical conversations and more than 48,000 physician-authored rubric… See the full description on the dataset page:
https://huggingface.co/datasets/projetogabi/healthbench-multilingual.
