Tag
The paper introduces FineWeb-CLaR, a large-scale annotated dataset that labels web documents with culture, language, and region metadata to enable auditing of cultural representation in language model pretraining data and benchmarks.