[Paper]
Macaron is a controlled, human-written benchmark for multilingual and multicultural reasoning created with a template-first approach.Each example is scenario-aligned across English and a local language, enabling controlled comparison of reasoning under culturally grounded premises.
True-False
3,954
Bilingual verification… See the full description on the dataset page:
https://huggingface.co/datasets/AlaaAhmed2444/Macaron.