Hugging Face Open-Sources Its Papers With Code Rebuild
The research-discovery service’s rebuilt application is available under Apache-2.0, with data imports needed for a self-hosted catalog.
TL;DR
- The rebuilt Papers with Code service has made its application code available under Apache-2.0.
- The project connects research papers with code, datasets and benchmark results.
- A fresh installation starts with an empty database, so operators must populate and maintain their own catalog.
The developers of Papers with Code have made the rebuilt service’s source code available under Apache-2.0. Zeniteq and Another Coding Blog reported the release, and the project’s public repository contains code for the research index. [1] [2]
The project indexes research papers alongside implementations, datasets and leaderboards. Zeniteq reported that the repository includes the web application, import workflows and maintenance tools. [1]
The code release concerns a rebuild of Papers with Code, whose original service was retired by Meta in 2025. Zeniteq traced the revived website’s launch to May 2026; the release of its code is a separate development. [1]
The repository’s setup instructions say a new installation begins with an empty database, so operating a useful index requires importing records and maintaining updates. Publishing the application code does not automatically provide a copy of the hosted catalog. [1]
Why it matters
A reusable codebase lets research groups adapt the discovery system to their own fields and workflows. Running a comparable service still depends on data coverage, maintenance and ongoing curation.
Editor's note
The repository and independent coverage were checked. The rebuilt website predated the code release, and the software license does not itself imply that the hosted catalog is bundled with a new installation.