DA-BIRD is a Danish text-to-SQL benchmark for evaluating large language models on natural language to SQL generation. All tasks are in Danish and use SQLite databases. The dataset is designed for use with the Harbor evaluation framework.
It combines two corpora — 363 tasks across 22 unique databases:
dst_*
213
11
Danish (original)
medium /… See the full description on the dataset page:
https://huggingface.co/datasets/oliverkinch/da-bird.
