Teaching Codex and Claude to Write ML Papers Like a Professor

What it does

This repository turns Prof. Peng Sida’s open study notes on research writing into a loadable skill for AI coding assistants like Codex, Claude Code, and Gemini. It provides structured rules, section-specific templates, and workflow guidance in SKILL.md and references/ so the model can help draft or rewrite paper sections, improve paragraph flow, check that claims match evidence, and run a pre-submission review from a reviewer mindset. The author’s stated contribution is curation and packaging, not the underlying methodology.

The interesting bit

The project treats research-writing heuristics as executable agent capabilities rather than static blog posts. By wrapping academic mentorship into SKILL.md rules and agent metadata, it effectively gives your coding assistant a temporary graduate advisor specialized in ML/CV/NLP rhetoric.

Key highlights

• Curated from Prof. Peng Sida’s open notes and repackaged as reusable AI skills
• Covers the full paper lifecycle: drafting, rewriting, logic flow, and reviewer-style self-critique
• Section-specific guides for Abstract, Introduction, Method, Experiments, and Conclusion
• Compatible with Codex, Claude Code, and Gemini via standard skills directories
• About 4,700 GitHub stars, suggesting strong appetite for structured academic writing help

Caveats

• The repository is primarily organizational glue: the writing methodology itself comes entirely from Prof. Peng Sida’s existing open notes
• Only one skill package (research-paper-writing/) is currently available
• The README does not reveal the actual content of the templates or specific heuristics inside references/

Verdict

Grab it if you want your AI assistant to stop generating boilerplate and start critiquing your paper’s argument structure. Look elsewhere if you need original technical methodology or domain-specific research insights rather than writing craft.