Hugging Face Papers: Guide to AI Research on HF

Target audience and objective: This guide is for artificial intelligence researchers, computational linguists, and open-source developers seeking to track emerging preprints on Hugging Face Papers, index their publications, link model weights and datasets, and verify paper authorship.

What Hugging Face Papers Is

Hugging Face Papers (commonly accessed via huggingface.co/papers or hf.co/papers) is a scholarly curation and community indexing platform that connects machine learning preprints directly to open-source software artefacts. In modern machine learning research, academic manuscripts posted to preprint servers often describe novel architectures without providing an immediate mechanism to test code or inspect pre-trained checkpoints. Hugging Face Papers resolves this fragmentation by associating arXiv papers directly with runnable model weights, benchmark datasets, and interactive demonstration applications.

Originally inspired by the daily paper curation streams initiated by machine learning researcher Ahsen Khaliq (widely known as AK), the platform integrates social discovery mechanisms into scientific publishing. Rather than treating academic papers as static PDF documents, Hugging Face Papers operates as a dynamic nexus within the broader Hugging Face Hub ecosystem. When scholars evaluate models across machine learning and artificial intelligence, the platform enables them to explore community commentary, inspect replication spaces, and monitor trending research trajectories in real time.

Who Runs It and What It Covers

Hugging Face Papers is developed, maintained, and operated by Hugging Face, Inc., the open-source machine learning company founded by Clément Delangue, Julien Chaumond, and Thomas Wolf. In May 2023, Hugging Face formalized Daily Papers as an integrated Hub feature, bringing Ahsen Khaliq’s ongoing curation work directly into the primary platform architecture. Following the retirement of Papers with Code by Meta in July 2025, Hugging Face Papers became the primary destination for incoming web traffic seeking community-curated machine learning literature.

The platform’s scope primarily encompasses computer science, artificial intelligence, computational linguistics, and computer vision literature. While the overwhelming majority of indexed manuscripts originate from arXiv (accounting for approximately 95% of paper links across the Hub), the platform’s indexing infrastructure continues to expand to accommodate broader scientific venues. Coverage across Hugging Face Papers is structured around three core operational categories:

• Daily Trending Feed: A curated and algorithmically ranked stream of recently published papers, organized by publication date, community upvotes, and discussion velocity.
• Artifact Linking: Bidirectional mappings connecting specific publications to repository cards on the Hugging Face Hub, encompassing neural network weights, evaluation datasets, and interactive Gradio demonstrations hosted in Hugging Face Spaces.
• Verified Author Profiles: Academic attribution systems that enable researchers to claim their papers, link their institutional identities, and highlight their publications on their public developer profiles.

How to Search or Use It

Researchers interacting with Hugging Face Papers can search existing literature, index novel preprints, and link software artefacts using documented platform workflows:

1. Browsing Daily and Trending Literature: Navigate to https://huggingface.co/papers to inspect the daily feed. Use the calendar selector or navigation toggles to review past daily digests, or filter by trending velocity to identify high-impact publications.
2. Searching by ArXiv Identifier or Title: Enter a publication title or exact arXiv identifier (for example, 2305.18290) into the search bar. Direct URL navigation is supported via the pattern hf.co/papers/[arxiv-id].
3. Indexing a New Paper: If an arXiv paper has not yet been indexed on the Hub, navigate directly to hf.co/papers/[arxiv-id] or submit the identifier through the search interface. The platform automatically retrieves bibliographic metadata, abstracts, and author lists from the arXiv API.
4. Linking Models, Datasets, and Spaces: To associate a research artefact with a publication, include the paper URL or arXiv identifier in the repository card (README.md) metadata. The Hub parser automatically extracts the arXiv tag and establishes a bidirectional link between the model and the paper page.
5. Claiming Paper Authorship: On any indexed paper page, locate your name in the author roster, click on your name, and select the “claim authorship” option. The platform redirects to account verification settings where requests are processed by administrative moderators.
6. Formatting Software Attribution: When linking repositories, adhere to open research software standards such as Citation File Format (CITATION.cff) in your source repository to maintain unambiguous citation tracking across academic and developer venues.
7. Participating in Community Discussions: Use the integrated discussion tab on individual paper pages to submit technical queries, share replication notes, or link third-party code implementations.

Strengths and Limits

Evaluating the technical capabilities and structural constraints of Hugging Face Papers helps researchers integrate the platform effectively into their discovery and dissemination routines.

Key Strengths

• Seamless Ecosystem Integration: Unlike traditional bibliometric databases, Hugging Face Papers directly bridges the gap between theoretical publications and executable code, datasets, and pre-trained weights.
• Immediate Discovery Velocity: New arXiv submissions appear on the platform within hours of release, enabling rapid community evaluation and discussion long before formal conference proceedings.
• Verified Author Presence: Enabling researchers to claim authorship and showcase verified papers directly on their Hub profiles enhances personal academic visibility within the open-source engineering community.
• Interactive Demonstrations: Direct linking to Hugging Face Spaces allows reviewers and fellow scientists to test model capabilities in a browser without setting up local GPU environments.

Technical and Operational Limitations

• ArXiv Dependency: The platform is heavily coupled to the arXiv ecosystem, offering limited native support for papers exclusively published in closed commercial journals or venues outside computer science.
• Upvote and Popularity Bias: Algorithmic ranking driven by community upvotes can favor headline-grabbing or generative AI topics over foundational theoretical work, niche subdisciplines, or negative experimental results.
• Absence of Formal Peer Review: Daily Papers functions as an open discovery layer rather than an archival publisher or peer-reviewed journal; papers displayed on the feed have not undergone formal peer evaluation.
• Variability in Artifact Quality: Linked models and spaces are uploaded by independent community members, meaning that repository documentation, execution environments, and checkpoint validity can vary substantially.

Alternatives

Researchers seeking tools for scientific literature discovery, preprint monitoring, and artifact tracking can evaluate Hugging Face Papers alongside several established alternatives:

Platform	Primary Focus	Code and Model Linking	Community Features	Content Scope
Hugging Face Papers	Preprints paired with models, datasets, and Spaces	Native integration with Hugging Face Hub repositories	Upvotes, author claims, discussion threads	Predominantly computer science and ML preprints (arXiv)
Papers with Code (Archive)	Historical paper-to-code mapping and SOTA benchmark tables	Historical GitHub links and evaluation tables	Archived community submissions and leaderboards	Archival dataset dumps (CC-BY-SA 4.0)
arXiv Sanity Preserver	Semantic paper recommendations and keyword tracking	Limited (external community links)	Personalized library collections and similarity search	Full arXiv repository categories
AlphaSignal	Automated newsletter summarizing trending research	Curated links to GitHub repositories	Editorial summaries and algorithmically ranked digests	High-velocity machine learning preprints
Semantic Scholar	Academic citation graphs and influential citation analysis	External links to GitHub and institutional repositories	Reading feeds, author claiming, citation alerts	Broad multidisciplinary scientific literature

Frequently Asked Questions

How does a paper get featured on Hugging Face Daily Papers?

Papers are submitted by community members or authors via the Hugging Face search interface within 14 days of publication on arXiv. Submissions are curated and ranked based on community upvotes, discussion engagement, and editorial moderation.

Can an author index a paper that does not have an existing model or dataset on the Hub?

Yes. Any researcher can navigate to hf.co/papers/[arxiv-id] to create an indexed paper page. The Hub automatically retrieves the title, abstract, and author list from arXiv even if no accompanying model weights or dataset repositories have been uploaded.

How does authorship verification work on Hugging Face Papers?

Authors click their name on the paper page and select “claim authorship.” Hugging Face verifies the request using matching institutional email addresses associated with the user account or through manual review by platform moderators. Verified papers then display on the researcher’s public profile.

Is Hugging Face Papers a replacement for peer-reviewed publication?

No. Hugging Face Papers is a preprint discovery and artifact aggregation platform, not a peer-reviewed academic journal. Authors utilize the platform to share preprints, model weights, and demo code prior to or alongside formal peer review at conferences and scholarly journals.

Sources

• Hugging Face Documentation (read 2026-11-07): Hugging Face Hub Documentation: Paper Pages. Available at: https://huggingface.co/docs/hub/en/paper-pages
• Hugging Face (read 2026-11-07): Hugging Face Daily Papers and Trending Feed. Available at: https://huggingface.co/papers
• Hugging Face (read 2026-11-07): Trending Research and Artifact Explorer. Available at: https://huggingface.co/papers/trending