Global Exams
Collecting exams in local languages from all over the world and structuring as a standardized collection.
MMLU is a test covers 57 tasks including elementary mathematics, computer science, law, and more. It is made up of exams from US based systems. However, as an English based dataset it first requires translation which is known to introduce linguistic and cultural artifacts. Instead, in this initiative we would collect exams in local languages from all over the world and structure as a standardized collection which can be released at the end of the 6 weeks. Anyone can participate from different countries around the world to recommend exam collections which are accessible and can be released as a dataset. This will be the localized MMLU to evaluate current generations on state-of-art models.
My Contribution: Bengali Exams dataset (15k) from University Entrance exams and Public Service exams.
🤗 University Entrance Exams (10k)
🤗 Public Service Exams (4k)
