Open Source Community Launches New Home for Papers with Code

A Collaborative Effort for AI Research

Following the retirement of the original Papers with Code platform by Meta in 2025, the open-source community, led by Hugging Face, has introduced a successor project to ensure the continued accessibility of AI research artifacts. The new platform, available at paperswithcode.co, serves as a repository for research papers, implementation code, models, datasets, and performance leaderboards.

Key Features of the Platform

The project maintains the core functionality of the original site while introducing modernized infrastructure. Key components include:

• Comprehensive Catalog: Provides access to papers linked with official code, project pages, and Hugging Face assets.
• Automated Curation: Employs agent workflows and scheduled jobs to import daily research, link tasks and methods, and track citations.
• Leaderboards: Tracks paper-native evaluation results on benchmarks through automated synchronization.
• Advanced Search: Features a hybrid search engine combining lexical and semantic retrieval using pgvector.
• Developer Support: Includes a REST API compliant with the original version, OpenAPI documentation, and an optional research-focused chat assistant.

Developer and Contributor Access

The platform is built with a focus on open maintenance and accessibility. The backend utilizes a PostgreSQL database, with architecture managed through environment variables for security. Developers can interact with the system using the `pwc` CLI, which allows for catalog maintenance and data indexing.

Those interested in the underlying architecture or contributing to the project can review the extensive documentation hosted in the repository, covering database schemas, data workflows, and deployment operations. The source code is released under the Apache License 2.0. Users are encouraged to verify project updates and documentation directly via the official GitHub repository or the live site.