35,837 TL;DR summaries of Hugging Face model and dataset cards, generated by Qwen/Qwen3.8-27B-FP8 on Hugging Face Jobs. Each row pairs a full card (untruncated) with a summary of a requested length — one, two, or three sentences — making this distillation data for a small card-summarisation model. Part of the hub-tldr collection.
This is raw teacher output. Cleanup (dedup, format screens) is deferred to the SFT build.
Source cards come from… See the full description on the dataset page:
https://huggingface.co/datasets/davanstrien/hub-tldr-v4.
