Datasets:

m-a-p
/

MSQA

license: cc-by-4.0
pretty_name: MSQA
task_categories:
  - question-answering
language:
  - en
  - zh
  - pt
  - th
  - ru
  - ko
  - fr
  - ja
  - ms
  - id
  - es
size_categories:
  - 1K<n<10K

MSQA — Coming Soon 🚧

MSQA (Multilingual and Multicultural SimpleQA) is a benchmark of 1,064 natively sourced questions spanning 11 language groups and 5 cultural dimensions.

The dataset is not yet publicly available — it will be released here soon.

📄 Paper: https://arxiv.org/abs/2607.00724
💻 Code: https://github.com/huayuankou333/MSQA

Citation

@misc{chen2026msqanativelysourcedmultilingual,
      title={MSQA: A Natively Sourced Multilingual and Multicultural SimpleQA Benchmark}, 
      author={Xianru Chen and Yukai Huang and Mingxiang Chen and Xinping Lei and Fangbing Deng and Jin Chen and Ge Zhang and Wenhao Huang and Jiaheng Liu},
      year={2026},
      eprint={2607.00724},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2607.00724}, 
}