awesome-ai-for-math-paper

name	awesome-ai-for-math-paper
description	Add paper links to the awesome-ai-for-math repository README. Use when the user gives an arXiv, DOI, proceedings, journal, or paper URL and asks to add it to awesome-ai-for-math, update the README paper table, classify topics, find associated resources such as GitHub repositories, update an existing paper-add PR, or commit the change with a Codex co-author trailer.

Awesome AI For Math Paper

Workflow

Use this workflow for the awesome-ai-for-math repository unless the user gives different instructions.

1. Inspect state first.

   • Run git status -sb.
   • Read the relevant README rows around the alphabetical insertion point.
   • If a PR branch is already active for paper additions, continue on it. If starting from main, create a codex/... branch before committing.

2. Fetch paper metadata from primary sources.

   • For arXiv links, read the arXiv abstract page or Atom API record.
   • Extract title, submission year, abstract, arXiv subjects, comments, and visible resource links.
   • Do not rely only on the user-provided title.

3. Choose README fields.

   • Title: exact paper title as rendered by the source, linked to the paper URL.
   • Subject(s): prefer existing README tags. Use mathematical domain tags first, then method tags such as LLM, ATP, RL, Transformer, Neural Network, or Benchmark.
   • Venue & Year: use arXiv YYYY unless a published venue is confirmed.
   • Links & Resources: include official resources discovered from the paper page, abstract, paper PDF, repository search, author pages, or clearly related existing project repos.
   • Add an [arXiv] link to the last column only when the title links to a different published version (journal or conference) and the arXiv link would not duplicate it. If the title already links to the arXiv page, do not repeat an [arXiv] link in the last column.

4. Search for associated resources.

   • Check the paper page and PDF for URLs such as GitHub, GitLab, project pages, datasets, notebooks, Lean artifacts, or chat logs.
   • Search the web for the exact title plus GitHub, code, and named systems from the abstract.
   • Verify candidate repositories exist before adding them.
   • If a paper has a dedicated repo, prefer it over a shared system repo. Example: use [Code](https://github.com/frenzymath/iteris) for Iteris rather than the shared Rethlas repo.
   • If a known shared project has an existing README label style and no more-specific repo is confirmed, reuse it. Example: use [Code (Rethlas)](https://github.com/frenzymath/Rethlas) for Rethlas-based papers.
   • Use Code only for actual executable code or formal artifacts. Use Chat Logs for transcripts or prompt/response records.

5. Edit README only unless the user asks for generated/site data updates.

   • Bump the paper count in the intro.
   • Insert the row alphabetically by title.
   • Keep markdown table formatting consistent.
   • Do not edit assets/papers.json or charts when the user says README-only.

6. Validate.

   • Run rg -c '^\\| \\*\\*\\[' README.md and confirm the count matches the intro.
   • Run git diff -- README.md and inspect the exact row(s).
   • Confirm git status -sb shows only intended files.

7. Commit and publish when requested.

   • Stage explicit files.
   • Include a Codex co-author trailer in commit messages:

Co-authored-by: Codex <[e-mail]>

• For multi-line commit messages, use git commit -m "Summary" -m "Co-authored-by: Codex <[e-mail]>".
• Push the branch and update or create the draft PR.
• Keep PR title/body aligned with the final set of paper additions and validation.

Classification Guidance

• Use Numerical Analysis, LLM for LLM-assisted computational mathematics papers centered on numerical methods.
• Use Algebraic Geometry, Number Theory for p-adic/nonabelian Hodge or arithmetic geometry papers unless the paper explicitly emphasizes formal proof or LLM methodology.
• Add ATP only when the paper uses automated/formal theorem proving or proof assistant artifacts.
• Add LLM when the abstract or resources describe language models, agentic AI systems, AI-generated proof drafts, or AI-assisted research workflow.