Skip to content

Dataset · EdinburghNLP

XSum: BBC news brief summaries

EdinburghNLP/xsum

BBC news articles with single-sentence summaries for extreme summarization training.

What's inside

XSum (Extreme Summarization) contains BBC news articles and their single-sentence summaries. Each example features document, summary, and BBC article ID fields. The English dataset size ranges from 100k to 1M examples; downloaded files take about 257.30 MB.

What you can do with it

  • Train abstractive summarization models on article-to-single-sentence pairs.
  • Fine-tune LoRA for BART/T5 on XSum.
  • Evaluate summarizers based on single-sentence summary quality.
Tasks
Summarization
Languages
English
Size
100K–1M rows
Downloads this month
19.6K
Likes
158

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("EdinburghNLP/xsum")