Crossref arXiv Citations: 6 Million Journal-Article Citations to 1 Million arXiv Preprints (COMET)

Provider
COMET (cometadata)
License
CC0-1.0 (source)
Size
6 files
Format
Parquet, JSON (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/cometadata/crossref-arxiv-citations)
Access
Free and public. No account with the data provider is needed.
Category
Open Data
Source
https://github.com/cometadata/crossref-arxiv-citation-extraction

About this dataset

A citation graph linking arXiv preprints to the Crossref-registered works (journal articles, proceedings, books) that cite them. The citations were extracted from the reference lists in the Crossref metadata snapshot of 2026-07 and validated against the DataCite data file of the same month, using the crossref-citation-extraction tool v2.2.0. The default configuration covers 1,042,011 cited arXiv works, 5,999,691 citing works and 6,028,756 reference matches. Each record gives the arXiv DOI and id, the reference and citation counts, and the list of citing DOIs, each with its provenance. The provenance values are publisher-asserted, Crossref-matched, or mined from reference text such as an arXiv id, DOI or URL, and the raw matched string and the Crossref reference metadata come with each match. The asserted and mined configurations separate high-confidence links from mined ones. Bibliometrics and science-of-science researchers use it to measure how preprints are cited in the published literature, which standard indexes undercount. The data is Parquet under data/; load it with Hugging Face datasets or pyarrow. Released by COMET under CC0 1.0.

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("cometadata/crossref-arxiv-citations")
print(ds)

Data structure

File types in sampled files: Parquet, JSON

Files

Example file: data/all.parquet

Analyze this dataset without downloading it

On Scigantic, Crossref arXiv Citations: 6 Million Journal-Article Citations to 1 Million arXiv Preprints (COMET) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Open Data datasets