README.md

metadata

configs:
  - config_name: humanities
    data_files:
      - split: test
        path: data/humanities.csv
  - config_name: language
    data_files:
      - split: test
        path: data/language.csv
  - config_name: misc
    data_files:
      - split: test
        path: data/misc.csv
  - config_name: social_science
    data_files:
      - split: test
        path: data/social_science.csv
  - config_name: stem
    data_files:
      - split: test
        path: data/stem.csv
license: cc-by-nc-4.0
task_categories:
  - question-answering
language:
  - ar
tags:
  - exams
  - MMLU
size_categories:
  - 10K<n<100K

Regrouped ArabicMMLU dataset. Groups are in accordance with the original dataset's categorization (see Table 1 in the original paper).

Original ungrouped dataset is available under MBZUAI