m-a-d-i/

wori-wolof-instructions

WORI (Wolof Reverse Instruction) is a linguistically validated instruction-tuning dataset for Wolof, a low-resource language. The dataset provides 3,724 unique instruction-output pairs in Wolof, with parallel French translations. It was constructed via a reverse instruction pipeline and validated through a combination of automated language identification and manual review. For full methodological details, please refer to the… See the full description on the dataset page: https://huggingface.co/datasets/m-a-d-i/wori-wolof-instructions.