---
pretty_name: "XSum"
license: "unknown"
language:
  - en
tags:
  - summarization
  - news
  - text
task_categories:
  - summarization
---

# XSum

This repository hosts a copy of the **XSum (Extreme Summarization)** dataset,
a benchmark for single-sentence abstractive summarization of news articles.

Each example typically contains a news article and a one-sentence summary.

## Contents
Typical splits:

- `train`
- `validation`
- `test`

A typical entry contains:
```json
{
  "document": "...",
  "summary": "...",
  "id": "..."
}
```

## Usage
```python
from datasets import load_dataset
ds = load_dataset("EdinburghNLP/xsum")
```

## Source
This dataset is taken from the Hugging Face dataset repository:
https://huggingface.co/datasets/EdinburghNLP/xsum

## License
The Hugging Face dataset card lists the license as **unknown**.
Please review the dataset’s original sources and terms of use before redistribution or commercial use.