Hugging Face Introduces DOI Support for Models and Datasets

TL;DR

Hugging Face announced native DOI generation for any model or dataset on the Hub, enabling permanent, version‑aware citations that improve reproducibility and scholarly credit.

What a DOI Is and Why It Matters

A Digital Object Identifier (DOI) is a globally unique string that references a digital object such as a research article, dataset, or machine‑learning model. The DOI resolves to metadata—including the object's URL, version, creation date, and description—making it a stable, permanent link. In academic practice, DOIs function like an ISBN for books: they provide a trusted citation mechanism that does not change over time.

How Hugging Face Assigns DOIs

Hugging Face partnered with DataCite, a leading DOI registration agency, to issue DOIs directly from the Hub. The workflow is:

  1. A registered user opens the repo settings of a model or dataset.
  2. The user fills out required metadata (title, authors, description, etc.).
  3. Hugging Face registers the DOI with DataCite and returns the identifier to the user.
  4. The model or dataset page now displays a "Cite this model/dataset" button that reveals the DOI and citation format.

If a new version of the object is released, the DOI can be updated to point to the latest version while the previous DOI remains resolvable, indicating that it is outdated. This version‑aware behavior ensures that citations can target a specific release without breaking when the object evolves.

Technical Implications

  • Persistent Linking: The DOI URL never expires, guaranteeing that references remain valid even if the underlying Hub URL changes.
  • Version Control: Updating a DOI creates a new resolution target, preserving the citation chain for historical versions.
  • Deletion Policy: Objects with DOIs are intended to persist indefinitely; removal is only possible after a formal request to Hugging Face support, aligning with scholarly archiving standards.
  • Community‑Driven Feature: The DOI support originated from community feedback on the hub-docs GitHub issue tracker, illustrating Hugging Face’s open‑source development model.

How to Use the Feature

  1. Navigate to a model or dataset repository on the Hub.
  2. Click Settings → Generate DOI.
  3. Complete the metadata form and submit.
  4. Copy the generated DOI from the citation modal that appears.
  5. Include the DOI in papers, reports, or any scholarly work to provide a permanent reference.

"If ever there’s a new version of a model or dataset, the DOI can easily be updated, and the previous version of the DOI gets outdated. This makes it easy to refer to a specific version of an object, even if it has changed." – Hugging Face blog

Community and Future Directions

Hugging Face invites users to suggest further improvements via Twitter @HuggingFace or by opening issues in the huggingface/hub-docs repository. The partnership with DataCite and the contributions of team members such as Alix Leroy, Bram Vanroy, Daniel van Strien, and Yoshitomo Matsubara were highlighted as key to bringing the feature to production.

This post summarizes the official Hugging Face announcement dated 7 October 2022. All details are taken directly from the source blog entry.

Sources