Hugging Face Paper Publisher

Publish research papers on Hugging Face Hub, link them to models and datasets, and manage authorship.

Overview

Hugging Face Hub hosts Paper Pages -- dedicated pages for research papers that aggregate all linked models, datasets, and Spaces citing a paper. This skill guides users through the full lifecycle: indexing papers from arXiv, linking them to Hub repositories via YAML metadata, claiming authorship, and generating markdown research articles.

Use this skill when users want to:

• •Index a paper on Hugging Face from its arXiv ID
• •Add paper citations to model or dataset README metadata
• •Claim or verify authorship on published papers
• •Create research articles from markdown templates
• •Manage which papers appear on their HF profile

Prerequisites:

• •Hugging Face account with a write-access token (HF_TOKEN env variable)
• •Python 3.10+ with huggingface_hub, pyyaml, requests packages
• •arXiv paper IDs for papers to be published

Learning Path

Level 1: Paper Indexing

Papers are indexed on Hugging Face by visiting their Paper Page URL. The Hub automatically creates a page when any user first accesses https://huggingface.co/papers/{arxiv-id}.

Index a paper programmatically:

python3 scripts/paper_manager.py index --arxiv-id "2301.12345"

This checks whether the paper exists on HF. If not indexed yet, it provides the URL to visit for automatic indexing. The Hub pulls metadata (title, authors, abstract) directly from the arXiv API.

Check if a paper is already indexed:

python3 scripts/paper_manager.py check --arxiv-id "2301.12345"

Direct browser access: Visit https://huggingface.co/papers/2301.12345 to trigger indexing and view the Paper Page.

Level 2: Linking Papers to Repositories

When a model or dataset README contains an arXiv URL, the Hub automatically:

1. •Extracts the arXiv ID from the link
2. •Adds an arxiv:PAPER_ID tag to the repository
3. •Shows the repository on the Paper Page under linked artifacts
4. •Makes the paper discoverable through Hub search filters

Link a paper to a model:

python3 scripts/paper_manager.py link \
  --repo-id "username/model-name" \
  --repo-type "model" \
  --arxiv-id "2301.12345"

Link to a dataset:

python3 scripts/paper_manager.py link \
  --repo-id "username/dataset-name" \
  --repo-type "dataset" \
  --arxiv-id "2301.12345"

Link multiple papers at once:

python3 scripts/paper_manager.py link \
  --repo-id "username/model-name" \
  --repo-type "model" \
  --arxiv-ids "2301.12345,2302.67890,2303.11111"

The script modifies the repository README to include the arXiv URL and optionally adds a BibTeX citation block. See references/yaml-metadata.md for the YAML frontmatter format.

Level 3: Authorship and Visibility

Claim authorship on a paper:

1. •Navigate to https://huggingface.co/papers/{arxiv-id}
2. •Find your name in the author list
3. •Click your name and select "Claim authorship"
4. •The HF admin team verifies the claim (usually within 24-48 hours)

Programmatic claim:

python3 scripts/paper_manager.py claim \
  --arxiv-id "2301.12345" \
  --email "your.email@institution.edu"

Manage paper visibility on your profile:

python3 scripts/paper_manager.py list-my-papers
python3 scripts/paper_manager.py toggle-visibility \
  --arxiv-id "2301.12345" --show true

Verified papers appear on your HF profile under the Papers tab. Toggle visibility in account settings to control which papers are public.

Best Practices

Do

• •Index papers immediately after arXiv publication for early visibility
• •Link papers to all related repositories (models, datasets, Spaces)
• •Include BibTeX citations in model and dataset cards
• •Use institutional email for authorship claims to speed verification
• •Add arxiv:PAPER_ID tags to repository metadata for discoverability

Don't

• •Skip YAML frontmatter in model/dataset cards (Hub needs it for metadata)
• •Use bare arXiv IDs without the full URL (Hub extracts IDs from URLs only)
• •Claim authorship on papers where you are not a listed author
• •Forget HF_TOKEN before running link operations (write access required)

Common Questions

Q: How long does authorship verification take?

A: The HF admin team typically verifies claims within 24-48 hours. Use an institutional email matching the paper's author affiliations to expedite.

Q: Can I link non-arXiv papers?

A: Paper Pages currently support arXiv papers only. Conference proceedings or journal papers can be referenced in README text, but won't generate a dedicated Paper Page with aggregated artifact views.

Q: What happens when I link a paper to a model?

A: The Hub adds an arxiv:PAPER_ID tag to the model. The Paper Page at hf.co/papers/{id} shows your model under "Models citing this paper." Users browsing the paper discover your model directly.

Q: Do I need a paid account?

A: No. Paper indexing, linking, and authorship are free features. You only need a write-access token for modifying repository READMEs via the API.

Examples

Example 1: Publish New Research End-to-End

# 1. Index on Hugging Face after arXiv publication
python3 scripts/paper_manager.py index --arxiv-id "2401.15839"

# 2. Link to your trained model
python3 scripts/paper_manager.py link \
  --repo-id "jdoe/llama-finetuned-legal" \
  --repo-type "model" \
  --arxiv-id "2401.15839"

# 3. Link to your training dataset
python3 scripts/paper_manager.py link \
  --repo-id "jdoe/legal-qa-dataset" \
  --repo-type "dataset" \
  --arxiv-id "2401.15839"

# 4. Claim authorship
python3 scripts/paper_manager.py claim \
  --arxiv-id "2401.15839" \
  --email "jane.doe@stanford.edu"

Example 2: Add Paper Reference to Existing Model

# Check if the paper is already indexed
python3 scripts/paper_manager.py check --arxiv-id "2310.06825"

# Link paper with a custom BibTeX citation
python3 scripts/paper_manager.py link \
  --repo-id "myorg/mistral-7b-instruct" \
  --repo-type "model" \
  --arxiv-id "2310.06825" \
  --citation "@article{jiang2023mistral,
  title={Mistral 7B},
  author={Jiang, Albert Q and Sablayrolles, Alexandre and others},
  journal={arXiv preprint arXiv:2310.06825},
  year={2023}
}"

Example 3: Create a Research Article

# Generate article from template
python3 scripts/paper_manager.py create \
  --template "standard" \
  --title "Efficient Fine-Tuning with Adapter Layers" \
  --authors "Jane Doe, John Smith" \
  --output "paper.md"

# Available templates: standard, modern, arxiv, ml-report

Workflow Patterns

Pattern 1: New Paper Publication

Write paper -> Submit to arXiv -> Index on HF -> Link to repos -> Claim authorship

python3 scripts/paper_manager.py create --template modern --output paper.md
# (edit paper.md, submit to arXiv, obtain arXiv ID)
python3 scripts/paper_manager.py index --arxiv-id "2401.XXXXX"
python3 scripts/paper_manager.py link --repo-id "user/model" \
  --repo-type model --arxiv-id "2401.XXXXX"
python3 scripts/paper_manager.py claim --arxiv-id "2401.XXXXX" \
  --email "user@institution.edu"

Pattern 2: Link Existing Paper to Multiple Repos

for repo in "user/model-v1" "user/model-v2" "user/training-data"; do
  repo_type="model"
  [[ "$repo" == *data* ]] && repo_type="dataset"
  python3 scripts/paper_manager.py link \
    --repo-id "$repo" --repo-type "$repo_type" \
    --arxiv-id "2301.12345"
done

Pattern 3: Author Portfolio Management

python3 scripts/paper_manager.py list-my-papers
python3 scripts/paper_manager.py toggle-visibility \
  --arxiv-id "2301.12345" --show true
python3 scripts/paper_manager.py toggle-visibility \
  --arxiv-id "2205.99999" --show false

Error Handling

Resources

References

• •references/yaml-metadata.md -- Model and dataset card YAML format with examples
• •references/paper-pages-guide.md -- How Paper Pages work on HF Hub

Scripts

• •scripts/paper_manager.py -- CLI tool for all paper operations (index, link, claim, create)

External

• •Hugging Face Paper Pages
• •Model Cards Guide
• •Dataset Cards Guide
• •arXiv Submission Help