Writing tips for modern Academic Writing
A curated list of resources for writing papers in AI, computer vision, and related fields.
🧭 Principles
- Simple and direct. A paper exists to transfer information, not to show off prose. Use short sentences, plain words, and as few nested clauses as possible.
- Figures first. If an idea can be shown as a figure, show it. A reader should get the core message of a figure within 30 seconds. See Figure 1 of HED and FCN, and Figures 1–2 of Attention Is All You Need.
- Math is a simplifier. If a process can be written as an equation, write it as one.
- Fixed content in fixed places. Contributions at the end of the introduction, implementation details at the start of experiments, ablations at the end. Reviewers know where to look.
- Cite every claim. Each claim is supported either by your own experiments or by a reference. Prefer the formally published version over the arXiv version.
Figures
- figures4papers — Python scripts for high-quality figures in top AI conferences and journals
📖 Classic Guides
- How to Write a Great Research Paper — Simon Peyton Jones
- Tips for Writing Technical Papers — Jennifer Widom
- How to Do Research — Bill Freeman
- How to Get Your SIGGRAPH Paper Rejected — Jim Kajiya
- The Science of Scientific Writing — George Gopen & Judith Swan
- Troubling Trends in Machine Learning Scholarship — Zachary Lipton & Jacob Steinhardt
✒️ Style
- The Elements of Style — William Strunk Jr.
- Politics and the English Language — George Orwell
📝 Writing & Editing Tools
- Overleaf — collaborative LaTeX editor
- Claude — drafting, polishing, and self-review before submission
- DeepL Write — rephrasing and grammar
- LanguageTool — open-source grammar checker
- Detexify — draw a symbol, get its LaTeX command
- Mathpix — convert equations in images and PDFs to LaTeX
- Tables Generator — build LaTeX tables visually
- booktabs — the standard package for clean, professional tables
📚 Literature & References
- Zotero — reference manager
- Semantic Scholar — paper search with citation context
- Google Scholar — paper search
- Connected Papers — explore related work as a graph
- dblp — clean, consistent BibTeX for CS venues
- rebiber — replace arXiv BibTeX entries with their published versions
🎨 Figure Tools
- SciDraw — free scientific illustrations
- draw.io — diagrams and pipeline figures
- Excalidraw — quick hand-drawn style sketches
- Inkscape — vector editing for final figures
- Figma — polished overview figures
- SciencePlots — publication-ready Matplotlib styles
- ColorBrewer — colorblind-safe and print-friendly palettes
✅ Before Submission
- NeurIPS Paper Checklist — a good self-review list for any ML paper
- arxiv-latex-cleaner — strip comments and unused files before uploading to arXiv