ML Paper Writing for Top AI Conferences

Expert-level guidance for writing publication-ready papers targeting NeurIPS, ICML, ICLR, ACL, AAAI, and COLM. This skill combines writing philosophy from top researchers (Nanda, Farquhar, Karpathy, Lipton, Steinhardt) with practical tools: LaTeX templates, citation verification APIs, and conference checklists.

Default operating order

Use this skill in the following order unless the task is unusually narrow:

1. lock the operating mode from references/OPERATING-MODES.md,
2. understand the repo or draft context,
3. use references/citation-workflow.md as the canonical citation authority,
4. load venue- or template-specific references only after the main writing path is clear.

Google Scholar may still help with manual discovery, but it is not the canonical verification authority in this skill. Default verification should use programmatic sources such as Semantic Scholar, CrossRef, and arXiv.

Core Philosophy: Collaborative Writing

Paper writing is collaborative, but Claude should be proactive in delivering drafts.

The typical workflow starts with a research repository containing code, results, and experimental artifacts. Claude's role is to:

1. Understand the project by exploring the repo, results, and existing documentation
2. Deliver a complete first draft when confident about the contribution
3. Search literature using web search and APIs to find relevant citations
4. Refine through feedback cycles when the scientist provides input
5. Ask for clarification only when genuinely uncertain about key decisions

Key Principle: Be proactive. If the repo and results are clear, deliver a full draft. Don't block waiting for feedback on every section—scientists are busy. Produce something concrete they can react to, then iterate based on their response.

⚠️ CRITICAL: Never Hallucinate Citations

This is the most important rule in academic writing with AI assistance.

The Problem

AI-generated citations have a ~40% error rate. Hallucinated references—papers that don't exist, wrong authors, incorrect years, fabricated DOIs—are a serious form of academic misconduct that can result in desk rejection or retraction.

The Rule

NEVER generate BibTeX entries from memory. ALWAYS fetch programmatically.

When You Can't Verify a Citation

If you cannot programmatically verify a citation, you MUST:

% EXPLICIT PLACEHOLDER - requires human verification
\cite{PLACEHOLDER_author2024_verify_this}  % TODO: Verify this citation exists

Always tell the scientist: "I've marked [X] citations as placeholders that need verification. I could not confirm these papers exist."

Recommended: Install Exa MCP for Paper Search

For the best paper search experience, install Exa MCP which provides real-time academic search:

Claude Code:

claude mcp add exa -- npx -y mcp-remote "https://mcp.exa.ai/mcp"

Cursor / VS Code (add to MCP settings):

{
  "mcpServers": {
    "exa": {
      "type": "http",
      "url": "https://mcp.exa.ai/mcp"
    }
  }
}

Exa MCP enables searches like:

• "Find papers on RLHF for language models published after 2023"
• "Search for transformer architecture papers by Vaswani"
• "Get recent work on sparse autoencoders for interpretability"

Then verify results with Semantic Scholar API and fetch BibTeX via DOI.

Workflow 0: Starting from a Research Repository

When beginning paper writing, start by understanding the project:

Project Understanding:
- [ ] Step 1: Explore the repository structure
- [ ] Step 2: Read README, existing docs, and key results
- [ ] Step 3: Identify the main contribution with the scientist
- [ ] Step 4: Find papers already cited in the codebase
- [ ] Step 5: Search for additional relevant literature
- [ ] Step 6: Outline the paper structure together
- [ ] Step 7: Draft sections iteratively with feedback

Step 1: Explore the Repository

# Understand project structure
ls -la
find . -name "*.py" | head -20
find . -name "*.md" -o -name "*.txt" | xargs grep -l -i "result\|conclusion\|finding"

Look for:

• README.md - Project overview and claims
• results/, outputs/, experiments/ - Key findings
• configs/ - Experimental settings
• Existing .bib files or citation references
• Any draft documents or notes

Step 2: Identify Existing Citations

Check for papers already referenced in the codebase:

# Find existing citations
grep -r "arxiv\|doi\|cite" --include="*.md" --include="*.bib" --include="*.py"
find . -name "*.bib"

These are high-signal starting points for Related Work—the scientist has already deemed them relevant.

Step 3: Clarify the Contribution

Before writing, explicitly confirm with the scientist:

"Based on my understanding of the repo, the main contribution appears to be [X]. The key results show [Y]. Is this the framing you want for the paper, or should we emphasize different aspects?"

Never assume the narrative—always verify with the human.

Step 4: Search for Additional Literature

Use web search to find relevant papers:

Search queries to try:
- "[main technique] + [application domain]"
- "[baseline method] comparison"
- "[problem name] state-of-the-art"
- Author names from existing citations

Then verify and retrieve BibTeX using the citation workflow below.

Step 5: Deliver a First Draft

Be proactive—deliver a complete draft rather than asking permission for each section.

If the repo provides clear results and the contribution is apparent:

1. Write the full first draft end-to-end
2. Present the complete draft for feedback
3. Iterate based on scientist's response

If genuinely uncertain about framing or major claims:

1. Draft what you can confidently
2. Flag specific uncertainties: "I framed X as the main contribution—let me know if you'd prefer to emphasize Y instead"
3. Continue with the draft rather than blocking

Questions to include with the draft (not before):

• "I emphasized X as the main contribution—adjust if needed"
• "I highlighted results A, B, C—let me know if others are more important"
• "Related work section includes [papers]—add any I missed"

When to Use This Skill

Use this skill when:

• Starting from a research repo to write a paper
• Drafting or revising specific sections
• Conducting literature reviews and finding related work
• Discovering recent papers in your research area
• Finding and verifying citations for related work
• Formatting for conference submission
• Resubmitting to a different venue (format conversion)
• Iterating on drafts with scientist feedback

Always remember: First drafts are starting points for discussion, not final outputs.

Workflow: Literature Research & Paper Discovery

When conducting literature reviews, finding related work, or discovering recent papers, use this workflow to systematically search, evaluate, and select ML papers.

Workflow 5: Finding and Evaluating Papers

Literature Research Process:
- [ ] Step 1: Define search scope and keywords
- [ ] Step 2: Search arXiv and academic databases
- [ ] Step 3: Screen papers by title/abstract
- [ ] Step 4: Evaluate paper quality (5 dimensions)
- [ ] Step 5: Select top papers and extract citations
- [ ] Step 6: Verify citations programmatically

Step 1: Define Search Scope

Identify specific research areas, methods, or applications:

• Technique-focused: transformer architecture, graph neural networks, self-supervised learning
• Application-focused: medical image analysis, reinforcement learning for robotics, language model alignment
• Problem-focused: out-of-distribution generalization, continual learning, fairness in ML

Step 2: Search arXiv

Use arXiv search with targeted keywords:

URL Pattern:
https://arxiv.org/search/?searchtype=all&query=KEYWORDS&abstracts=show&order=-announced_date_first

Example Searches:
- https://arxiv.org/search/?searchtype=all&query=graph+neural+networks&abstracts=show&order=-announced_date_first
- https://arxiv.org/search/?cat:cs.LG+AND+all:transformer&abstracts=show&order=-announced_date_first

Tips:

• Combine keywords with + for AND
• Filter by categories: cs.LG, cs.AI, cs.CV, cs.CL
• Sort by announced_date_first for recent papers
• Use Chrome MCP tools when available for automation

Step 3: Screen Papers

Quick screening by title and abstract:

• Relevance to research topic
• Novelty of contribution
• Venue/reputation of authors
• Code availability (check for GitHub links)

Step 4: Evaluate Quality

Use the 5-dimension quality criteria:

Scoring: Rate each dimension 1-5, calculate weighted total

Step 5: Select and Extract

• Rank papers by total score
• Select top papers for detailed review
• Extract metadata: title, authors, arXiv ID, abstract
• Note code repository links

Step 6: Verify Citations

For selected papers, verify citations using Semantic Scholar API:

• Fetch BibTeX programmatically via DOI
• Mark unverified citations as [CITATION NEEDED]
• Store in bibliography with verification status

When to Use Literature Research

Use this workflow when:

• Starting a new project: Find related work and baselines
• Writing Related Work section: Discover recent papers in your area
• Staying updated: Track recent publications in your field
• Finding baselines: Identify state-of-the-art methods for comparison
• Literature review: Comprehensive survey of research area

Quality Thresholds

• Excellent: 4.0+ (include definitely)
• Good: 3.5-3.9 (include if relevant)
• Fair: 3.0-3.4 (include if highly relevant)
• Poor: <3.0 (exclude unless essential)

Reference Files

For detailed literature research guidance:

• references/literature-research/arxiv-search-guide.md - arXiv search strategies and URL patterns
• references/literature-research/paper-quality-criteria.md - Detailed 5-dimension evaluation rubrics

Knowledge Base: Paper-Miner Global Writing Memory

This skill consumes a single canonical writing memory maintained by paper-miner:

• references/knowledge/paper-miner-writing-memory.md

This memory is global, not project-specific.

Even when paper-miner is invoked while working inside a specific repository, it still writes mined writing knowledge only into this one global memory. It does not maintain project-local writing memory.

Canonical memory structure

The maintained memory contains these sections:

How the memory is maintained

The paper-miner agent reads papers and merges reusable writing knowledge into this one file:

You: "Learn writing patterns from this paper: path/to/paper.pdf"
↓
paper-miner analyzes the paper
↓
Extracts reusable writing signals
↓
Updates paper-miner-writing-memory.md
↓
ml-paper-writing reuses that memory later

When to use this memory

Use the global paper-miner memory when you need:

• structure inspiration for intros, methods, results, or discussion,
• reusable transition phrases or framing templates,
• venue-facing writing signals,
• rebuttal phrasing and response structure ideas,
• examples of how strong papers support and sequence claims.

Default read order

When drafting or revising with ml-paper-writing, read this memory before writing if the task involves:

• introduction framing,
• related work organization,
• method exposition style,
• results narration,
• discussion framing,
• venue-facing polishing.

Use this read order:

1. references/knowledge/paper-miner-writing-memory.md
2. repo-local evidence and experiment artifacts
3. cited papers or notes if needed
4. venue template and formatting constraints

Read narrowly, not exhaustively:

• first scan How this helps our writing,
• then check Writing patterns mined and Structure signals,
• then inspect Reusable phrasing only for concrete wording help,
• use Venue-specific signals when targeting a known venue.

Contribution rule

Every paper mined by paper-miner should improve the same global memory.

Do not scatter newly mined knowledge across multiple maintained files. Do not create project-specific paper-miner memory. Do not duplicate near-identical patterns from the same source.

See references/knowledge/README.md for the detailed knowledge-base contract.

Balancing Proactivity and Collaboration

Default: Be proactive. Deliver drafts, then iterate.

Draft first, ask with the draft (not before):

Only block for input when:

• Контур статьи ещё не выбран (это меняет рамку, а не только объём)
• Multiple contradictory framings seem equally valid
• Results seem incomplete or inconsistent
• Explicit request to review before continuing

Don't block for:

• Word choice decisions
• Section ordering
• Which specific results to show (make a choice, flag it)
• Citation completeness (draft with what you find, note gaps)

The Narrative Principle

The single most critical insight: Your paper is not a collection of experiments—it's a story with one clear contribution supported by evidence.

Every successful ML paper centers on what Neel Nanda calls "the narrative": a short, rigorous, evidence-based technical story with a takeaway readers care about.

Three Pillars (must be crystal clear by end of introduction):

If you cannot state your contribution in one sentence, you don't yet have a paper.

Paper Structure Workflow

Workflow 1: Writing a Complete Paper (Iterative)

Copy this checklist and track progress. Each step involves drafting → feedback → revision:

Paper Writing Progress:
- [ ] Step 1: Define the one-sentence contribution (with scientist)
- [ ] Step 2: Draft Figure 1 → get feedback → revise
- [ ] Step 3: Draft abstract → get feedback → revise
- [ ] Step 4: Draft introduction → get feedback → revise
- [ ] Step 5: Draft methods → get feedback → revise
- [ ] Step 6: Draft experiments → get feedback → revise
- [ ] Step 7: Draft related work → get feedback → revise
- [ ] Step 8: Draft limitations → get feedback → revise
- [ ] Step 9: Complete paper checklist (required)
- [ ] Step 10: Final review cycle and submission

Step 1: Define the One-Sentence Contribution

This step requires explicit confirmation from the scientist.

Before writing anything, articulate and verify:

• What is the single thing your paper contributes?
• What was not obvious or present before your work?

"I propose framing the contribution as: '[one sentence]'. Does this capture what you see as the main takeaway? Should we adjust the emphasis?"

Step 2: Draft Figure 1

Figure 1 deserves special attention—many readers skip directly to it.

• Convey core idea, approach, or most compelling result
• Use vector graphics (PDF/EPS for plots)
• Write captions that stand alone without main text
• Ensure readability in black-and-white (8% of men have color vision deficiency)

Step 3: Write Abstract (5-Sentence Formula)

From Sebastian Farquhar (DeepMind):

1. What you achieved: "We introduce...", "We prove...", "We demonstrate..."
2. Why this is hard and important
3. How you do it (with specialist keywords for discoverability)
4. What evidence you have
5. Your most remarkable number/result

Delete generic openings like "Large language models have achieved remarkable success..."

Step 4: Write Introduction (1-1.5 pages max)

Must include:

• 2-4 bullet contribution list (max 1-2 lines each in two-column format)
• Clear problem statement
• Brief approach overview
• Methods should start by page 2-3 maximum

Step 5: Methods Section

Enable reimplementation:

• Conceptual outline or pseudocode
• All hyperparameters listed
• Architectural details sufficient for reproduction
• Present final design decisions; ablations go in experiments

Step 6: Experiments Section

For each experiment, explicitly state:

• What claim it supports
• How it connects to main contribution
• Experimental setting (details in appendix)
• What to observe: "the blue line shows X, which demonstrates Y"

Requirements:

• Error bars with methodology (standard deviation vs standard error)
• Hyperparameter search ranges
• Compute infrastructure (GPU type, total hours)
• Seed-setting methods

Step 7: Related Work

Organize methodologically, not paper-by-paper:

Good: "One line of work uses Floogledoodle's assumption [refs] whereas we use Doobersnoddle's assumption because..."

Bad: "Snap et al. introduced X while Crackle et al. introduced Y."

Cite generously—reviewers likely authored relevant papers.

Step 8: Limitations Section (REQUIRED)

All major conferences require this. Counter-intuitively, honesty helps:

• Reviewers are instructed not to penalize honest limitation acknowledgment
• Pre-empt criticisms by identifying weaknesses first
• Explain why limitations don't undermine core claims

Step 9: Paper Checklist

NeurIPS, ICML, and ICLR all require paper checklists. See references/checklists.md.

Writing Philosophy for Top ML Conferences

This section distills the most important writing principles from leading ML researchers. These aren't optional style suggestions—they're what separates accepted papers from rejected ones.

"A paper is a short, rigorous, evidence-based technical story with a takeaway readers care about." — Neel Nanda

The Sources Behind This Guidance

This skill synthesizes writing philosophy from researchers who have published extensively at top venues:

For deeper dives into any of these, see:

• references/writing-guide.md - Full explanations with examples
• references/sources.md - Complete bibliography

Time Allocation (From Neel Nanda)

Spend approximately equal time on each of:

1. The abstract
2. The introduction
3. The figures
4. Everything else combined

Why? Most reviewers form judgments before reaching your methods. Readers encounter your paper as: title → abstract → introduction → figures → maybe the rest.

Writing Style Guidelines

Sentence-Level Clarity (Gopen & Swan's 7 Principles)

These principles are based on how readers actually process prose. Violating them forces readers to spend cognitive effort on structure rather than content.

Full 7 principles with detailed examples: See references/writing-guide.md

Micro-Level Tips (Ethan Perez)

These small changes accumulate into significantly clearer prose:

• Minimize pronouns: ❌ "This shows..." → ✅ "This result shows..."
• Verbs early: Position verbs near sentence start
• Unfold apostrophes: ❌ "X's Y" → ✅ "The Y of X" (when awkward)
• Delete filler words: "actually," "a bit," "very," "really," "basically," "quite," "essentially"

Full micro-tips with examples: See references/writing-guide.md

Word Choice (Zachary Lipton)

• Be specific: ❌ "performance" → ✅ "accuracy" or "latency" (say what you mean)
• Eliminate hedging: Drop "may" and "can" unless genuinely uncertain
• Avoid incremental vocabulary: ❌ "combine," "modify," "expand" → ✅ "develop," "propose," "introduce"
• Delete intensifiers: ❌ "provides very tight approximation" → ✅ "provides tight approximation"

Precision Over Brevity (Jacob Steinhardt)

• Consistent terminology: Different terms for same concept creates confusion. Pick one and stick with it.
• State assumptions formally: Before theorems, list all assumptions explicitly
• Intuition + rigor: Provide intuitive explanations alongside formal proofs

What Reviewers Actually Read

Understanding reviewer behavior helps prioritize your effort:

Bottom line: If your abstract and intro don't hook reviewers, they may never read your brilliant methods section.

Citation Workflow (Hallucination Prevention)

⚠️ CRITICAL: AI-generated citations are a high-risk failure mode. Never write BibTeX from memory.

Canonical authority

Use references/citation-workflow.md as the default authority for citation verification.

The default verification path is:

1. Search programmatically with Semantic Scholar / CrossRef / arXiv / OpenAlex when appropriate.
2. Verify existence in two sources when the claim is important.
3. Retrieve BibTeX programmatically from DOI or a trusted source.
4. Validate the claim against the actual paper content when the citation supports a specific statement.
5. Add the citation only after the metadata and claim are verified.

The golden rule

IF you cannot verify a citation programmatically:
    -> mark it as [CITATION NEEDED] or [PLACEHOLDER - VERIFY]
    -> tell the scientist explicitly
    -> NEVER invent a plausible-sounding reference

Workflow 2: Adding citations

Citation verification:
- [ ] Step 1: Search with Semantic Scholar / CrossRef / arXiv / OpenAlex as appropriate
- [ ] Step 2: Confirm title, authors, year, and venue
- [ ] Step 3: Retrieve BibTeX from DOI, arXiv, or another trusted export path
- [ ] Step 4: Verify that the claim being cited actually appears in the source
- [ ] Step 5: Add verified BibTeX to the bibliography
- [ ] Step 6: If any step fails -> mark as placeholder and report it explicitly

Discovery vs authority

• Programmatic APIs are the canonical verification path.
• Google Scholar may still be used as a manual discovery surface when coverage is weak, but not as the primary authority.
• If Google Scholar finds something that the canonical APIs do not, treat it as a lead that still requires explicit verification.

Summary: citation rules

🚨 NEVER generate BibTeX from memory. Use the programmatic workflow in references/citation-workflow.md. 🚨

Complete Citation Workflow Example

Scenario: You need to cite the Transformer paper.

Step 1: Search programmatically
- Semantic Scholar query: "Attention is All You Need Vaswani 2017"
- Result: title, authors, year, and DOI align

Step 2: Verify existence
- CrossRef confirms DOI metadata
- Semantic Scholar record matches the same paper

Step 3: Retrieve BibTeX
- Fetch BibTeX from the DOI / trusted export path

Step 4: Verify the claim
- Read the abstract or paper section that supports the cited statement
- Confirm that the claim being cited is actually present

Step 5: Add to bibliography
- Paste verified BibTeX into the .bib file
- Cite with the verified key

Step 6: If any step fails
- mark the citation as [PLACEHOLDER - VERIFY]
- tell the scientist explicitly what remains unverified

Common Issues and Solutions

Issue: Abstract too generic

Delete first sentence if it could be prepended to any ML paper. Start with your specific contribution.

Issue: Introduction exceeds 1.5 pages

Split background into Related Work. Front-load contribution bullets. Methods should start by page 2-3.

Issue: Experiments lack explicit claims

Add sentence before each experiment: "This experiment tests whether [specific claim]..."

Issue: Reviewers find paper hard to follow

• Add explicit signposting: "In this section, we show X"
• Use consistent terminology throughout
• Include figure captions that stand alone

Issue: Missing statistical significance

Always include:

• Error bars (specify: std dev or std error)
• Number of runs
• Statistical tests if comparing methods

Reviewer Evaluation Criteria

Reviewers assess papers on four dimensions:

Scoring (NeurIPS 6-point scale):

• 6: Strong Accept - Groundbreaking, flawless
• 5: Accept - Technically solid, high impact
• 4: Borderline Accept - Solid, limited evaluation
• 3: Borderline Reject - Solid but weaknesses outweigh
• 2: Reject - Technical flaws
• 1: Strong Reject - Known results or ethics issues

See references/reviewer-guidelines.md for detailed reviewer instructions.

Tables and Figures

Tables

Use booktabs LaTeX package for professional tables:

\usepackage{booktabs}
\begin{tabular}{lcc}
\toprule
Method & Accuracy ↑ & Latency ↓ \\
\midrule
Baseline & 85.2 & 45ms \\
\textbf{Ours} & \textbf{92.1} & 38ms \\
\bottomrule
\end{tabular}

Rules:

• Bold best value per metric
• Include direction symbols (↑ higher is better, ↓ lower is better)
• Right-align numerical columns
• Consistent decimal precision

Figures

• Vector graphics (PDF, EPS) for all plots and diagrams
• Raster (PNG 600 DPI) only for photographs
• Use colorblind-safe palettes (Okabe-Ito or Paul Tol)
• Verify grayscale readability (8% of men have color vision deficiency)
• No title inside figure—the caption serves this function
• Self-contained captions—reader should understand without main text

References & Resources

Reference Documents (Deep Dives)

LaTeX Templates

Compiling to PDF:

• VS Code/Cursor: Install LaTeX Workshop extension + TeX Live → Save to auto-compile
• Command line: latexmk -pdf main.tex or pdflatex + bibtex workflow
• Online: Upload to Overleaf

Key External Sources

Writing Philosophy:

• Neel Nanda: How to Write ML Papers - Narrative, "What/Why/So What"
• Farquhar: How to Write ML Papers - 5-sentence abstract
• Gopen & Swan: Science of Scientific Writing - 7 reader expectation principles
• Lipton: Heuristics for Scientific Writing - Word choice
• Perez: Easy Paper Writing Tips - Micro-level clarity

APIs: Semantic Scholar | CrossRef | arXiv