enPurified/tulu-3-sft-mixture-enPurified-openai-messages

Dataset Card: enPurified Collection **This dataset was updated on January 13th, 2026 to strip out even more math/code. The pruning process reduced the dataset from 940,000 to 88,782 rows of high-quality English prose. (The script used for this process is uploaded in the files section) Dataset Summary The enPurified collection is a curated suite of datasets designed to isolate high-quality English prose from existing high-value open-source datasets. The primary… See the full description on the dataset page: https://huggingface.co/datasets/enPurified/tulu-3-sft-mixture-enPurified-openai-messages.