MasrawiQA is a culturally rooted, manually curated evaluation benchmark containing 129 question-answer pairs designed exclusively for the Cairene and Lower Egyptian Arabic dialect (arz_arab). Introduced as part of the MRL 2026 Shared Task, this dataset evaluates language models on authentic street slang, everyday idioms, and dense regional knowledge spanning Egyptian sports history, local cuisine, and urban pop culture.
The dataset… See the full description on the dataset page:
https://huggingface.co/datasets/JohnnyGeorge/MasrawiQA.
