# Research Paper Writing

Pipeline for ML/AI research papers, lit review to LaTeX submission

- Repository: https://github.com/taracodlabs/aiden (868 stars)
- Category: Data and ML
- License: Apache-2.0
- Last updated: 2026-10-19
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add taracodlabs/aiden --skill research-paper-writing -g -a claude-code -y`
Codex: `npx skills add taracodlabs/aiden --skill research-paper-writing -g -a codex -y`
Cursor: `npx skills add taracodlabs/aiden --skill research-paper-writing -g -a cursor -y`
Gemini CLI: `npx skills add taracodlabs/aiden --skill research-paper-writing -g -a gemini-cli -y`

## Description

Pipeline for ML/AI research papers, lit review to LaTeX submission

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Research Paper Writing Pipeline

A structured, step-by-step pipeline for writing ML/AI research papers: from problem framing through literature review, experiment design, writing, and LaTeX formatting for arXiv submission.

## When to Use

- User wants to write or structure an academic research paper
- User wants help organizing experiments and results for a paper
- User wants to format a paper in LaTeX for arXiv or a conference
- User wants to do a literature review on a topic
- User wants to write an abstract, introduction, or related work section

## How to Use

### Phase 1: Frame the contribution

Define the paper's core claim before writing anything else.

```
1. One-sentence contribution: "We show that X outperforms Y on Z by doing W"
2. Key insight: what is non-obvious about your approach?
3. Research question: what question does the paper answer?
4. Limitations scope: what is explicitly out of scope?
```

### Phase 2: Literature review

Use the arXiv skill to find related work, then organize findings:

```
Search strategy:
- Start with 2-3 "seed" papers you know are relevant
- Find papers that cite them (via Semantic Scholar API)
- Search arXiv for your core keywords + recent date filter
- Organize into: direct predecessors, concurrent work, tangential work

For each related paper, note:
- Core method
- Dataset/benchmark used
- Key result number
- How your work differs
```

```python
# Semantic Scholar API, find papers citing a known paper
import requests
paper_id = "arXiv:2305.17333"
resp = requests.get(f"https://api.semanticscholar.org/graph/v1/paper/{paper_id}/citations?fields=title,year,authors,externalIds&limit=20")
for c in resp.json()["data"]:
  print(c["citingPaper"]["title"])
```

### Phase 3: Structure the paper

Standard ML/AI paper structure:

The rest: https://github.com/taracodlabs/aiden/blob/HEAD/skills/research-paper-writing/SKILL.md