Ultra-FineWeb-L1 is a large-scale English web corpus built from Common Crawl snapshots. Within UltraData's L0-L4 tiered data management framework, it serves as the L1 filtered layer for general web data and provides the foundation for subsequent L2 selection and L3 refinement. Building on the FineWeb processing pipeline, we perform⦠See the full description on the dataset page:
https://huggingface.co/datasets/openbmb/Ultra-FineWeb-L1.