- Home
- Skills
- Orchestra Research
- Ai Research Skills
- 20 Ml Paper Writing
20-ml-paper-writing_skill
- TeX
5.2k
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill orchestra-research/ai-research-skills --skill 20-ml-paper-writing- SKILL.md37.8 KB
Overview
This skill creates publication-ready ML/AI papers tailored to top venues (NeurIPS, ICML, ICLR, ACL, AAAI, COLM). It combines a proactive drafting workflow, LaTeX templates, reviewer checklists, and programmatic citation verification to produce reproducible, submission-ready manuscripts. Use it to turn research repositories and experimental artifacts into polished first drafts and camera-ready submissions.
How this skill works
The skill inspects a research repo (code, results, configs, existing citations) to extract the main contribution and experimental evidence. It generates a complete LaTeX draft with structured sections, figure guidance, checklist items, and placeholders for any unverifiable citations. Citations are always fetched and verified via APIs; any unverified references are explicitly marked as placeholders for human confirmation.
When to use it
- Starting from a research repository to write a conference paper
- Drafting or revising specific sections (abstract, intro, methods, experiments)
- Verifying and fetching BibTeX citations programmatically
- Converting drafts to conference-specific LaTeX formats for submission
- Preparing camera-ready versions and checklist compliance
Best practices
- Be proactive: deliver a full first draft when the contribution and results are clear
- Never invent citations—mark unverifiable references as explicit placeholders
- Define the one-sentence contribution with the scientist before finalizing framing
- Include reproducibility details: hyperparameters, seeds, compute, and error bars
- Draft Figure 1 early: make it convey the core idea and stand alone
- Add a required limitations section to preempt reviewer concerns
Example use cases
- Transforming an experiments/outputs folder and README into a full NeurIPS draft
- Rewriting the Related Work section with verified BibTeX entries from Semantic Scholar
- Converting an ICLR submission to ACL format with updated templates and page limits
- Preparing a camera-ready paper with checklist compliance and figure file fixes
- Iterating on a draft after reviewer-style feedback with targeted revisions
FAQ
I will insert an explicit LaTeX placeholder comment and list the citation as requiring human verification in the draft notes.
How do you decide when to ask me questions versus drafting autonomously?
I default to drafting when the repo and results make the contribution clear; I ask 1-2 targeted questions only when framing is ambiguous or key results are missing.