datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
xsum
Dataset Card for "xsum"
Dataset Summary
Extreme Summarization (XSum) Dataset.
There are three features:
document: Input news article.
summary: One sentence summary of the article.
id: BBC ID of the article.
Supported Tasks and Leaderboards
More Information Needed
Languages
More Information Needed
Dataset Structure
Data Instances
default
Size of downloaded dataset files: 257.30 MB
Size of the generated dataset:… See the full description on the dataset page: https://huggingface.co/datasets/EdinburghNLP/xsum.xsumThis is the XSUM subset of the GEM benchmark.xsum_validation_t5IGB_XSumxsum_train_t5xsum_factualityNeural abstractive summarization models are highly prone to hallucinate content that is unfaithful to the input
document. The popular metric such as ROUGE fails to show the severity of the problem. The dataset consists of
faithfulness and factuality annotations of abstractive summaries for the XSum dataset. We have crowdsourced 3 judgements
for each of 500 x 5 document-system pairs. This will be a valuable resource to the abstractive summarization community.xsumxsum
Dataset Card for SAMSum Corpus
Dataset Description
Links
Homepage: https://arxiv.org/abs/1808.08745
Repository: https://arxiv.org/abs/1808.08745
Paper: https://arxiv.org/abs/1808.08745
Point of Contact: https://huggingface.co/knkarthick
Dataset Summary
This repository contains data and code for our EMNLP 2018 paper "Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization".… See the full description on the dataset page: https://huggingface.co/datasets/knkarthick/xsum.xsum_swe
Dataset Card for Swedish Xsum Dataset
The Swedish xsum dataset has only been machine-translated to improve downstream fine-tuning on Swedish summarization tasks.
Dataset Summary
Read about the full details at original English version: https://huggingface.co/datasets/xsum
Data Fields
id: a string containing the heximal formated SHA1 hash of the url where the story was retrieved from
document: a string containing the body of the news article
summary: a string… See the full description on the dataset page: https://huggingface.co/datasets/Gabriel/xsum_swe.xsum_deepseek-moe-16b-chat_token_patternsstacked-xsum
xsum-stacked
The current version (corresponding to the stacked-booksum release): v0.3. See the Stacked Summaries org page for what this is and why it exists.
The maximum input length is 16384 tokens, and the maximum output length is 1024 tokens (measured with the Long-T5 tokenizer).
stats
[2023-01-09 19:36:25] INFO:root:INPUTS - basic stats - train
[2023-01-09 19:36:26] INFO:root:{'num_columns': 5,
'num_rows': 204045,
'num_unique_target': 203107,
'num_unique_text':… See the full description on the dataset page: https://huggingface.co/datasets/stacked-summaries/stacked-xsum.preprocessed_xsum
Dataset Card for "preprocessed_xsum"
More Information needed
xsum_trxsum-MIA-Benchmarkstacked-xsum-1024
stacked-xsum-1024
a "stacked" version of xsum
Original Dataset: copy of the base dataset
Stacked Rows: The original dataset is processed by stacking rows based on certain criteria:
Maximum Input Length: The maximum length for input sequences is 1024 tokens in the longt5 model tokenizer.
Maximum Output Length: The maximum length for output sequences is also 1024 tokens in the longt5 model tokenizer.
Special Token: The dataset utilizes the [NEXT_CONCEPT] token to indicate a new… See the full description on the dataset page: https://huggingface.co/datasets/stacked-summaries/stacked-xsum-1024.xsum_Qwen3-30B-A3B_moe_patternsxSum-processedxsum-kk3Extreme Summarization (XSum) Dataset.
There are three features:
- document: Input news article.
- summary: One sentence summary of the article.
- id: BBC ID of the article.xsum_deepseek-moe-16b-chat_moe_patternsonlystacked-xsum-1024
stacked-summaries/onlystacked-xsum-1024
Same thing as stacked-summaries/stacked-xsum-1024 but filtered such that is_stacked=True. Please refer to the original dataset for info and to raise issues if needed.
Basic info on train split:
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 116994 entries, 0 to 116993
Data columns (total 6 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 document 116994 non-null string
1… See the full description on the dataset page: https://huggingface.co/datasets/stacked-summaries/onlystacked-xsum-1024.xsum_Qwen3-30B-A3B_token_patternsnorwegian-xsumBOE-XSUM
BOE-XSUM Balanced Dataset - Reviewed and Cleaned
Description
The BOE 2025 Dataset is a collection of BOE articles with extreme summaries of them. This dataset has been carefully balanced and cleaned to ensure its quality and usefulness in natural language processing (NLP) tasks, primarily for evaluating generative models.
Read more in https://arxiv.org/abs/2509.24908
Dataset Content
The dataset is composed of the following subsets (splits):
train: Training… See the full description on the dataset page: https://huggingface.co/datasets/bertin-project/BOE-XSUM.mini_xsumxsum_nl
Dataset Card for XSum NL
Dataset Summary
This dataset is a machine translated dataset. It's the XSum dataset translated with this model from English to Dutch.
See the Hugginface page of the original dataset for more information on the format of this dataset.
Use with:
from datasets import load_dataset
load_dataset("csv", "ml6team/xsum_nl")
Languages
Dutch
Dataset Structure
Data Instances
[More Information Needed]
Data… See the full description on the dataset page: https://huggingface.co/datasets/ml6team/xsum_nl.xsum
Dataset Card for xsum
This dataset is a collection of pairs of news articles and their summaries. See xsum for additional information.
This dataset can be used directly with Sentence Transformers to train embedding models.
Dataset Subsets
pair subset
Columns: "article", "summary"
Column types: str, str
Examples:{
'article': 'Denny Solomona crossed for Castleford, but Wigan led at half-time against the run of play through Lewis Tierney\'s try and Matty… See the full description on the dataset page: https://huggingface.co/datasets/sentence-transformers/xsum.LID-XSUM
LID-XSUM: Stress-testing Machine Generated Text Detection: Shifting Language Models Writing Style to Fool Detectors
Datasets generated by aligning LLMs using Direct Preference Optimization to shift the machine-generated texts' (MGT) style toward human-written text (HWT).
This dataset is intended to be used to augment the training set of documents to train more robust MGT detectors.
Dataset Details
The adversarial generations obtained in the paper "Stress-testing… See the full description on the dataset page: https://huggingface.co/datasets/andreapdr/LID-XSUM.xsum
Dataset Card for xsum
This is a preprocessed version of xsum dataset for benchmarks in LM-Polygraph.
Dataset Details
Dataset Description
Curated by: https://huggingface.co/LM-Polygraph
License: https://github.com/IINemo/lm-polygraph/blob/main/LICENSE.md
Dataset Sources [optional]
Repository: https://github.com/IINemo/lm-polygraph
Uses
Direct Use
This dataset should be used for performing benchmarks on LM-polygraph.… See the full description on the dataset page: https://huggingface.co/datasets/LM-Polygraph/xsum.xsum_reducedscraped_xsum1