NeurIPS 2026 · Checklist

NeurIPS 2026 Paper Checklist

Mandatory for every submission. Appended after references. Doesn't count toward the 9-page limit. Here's every question with guidance on how to answer well.

Open template with checklist.tex pre-filled

What is checklist.tex?

checklist.tex is the standardized NeurIPS Paper Checklist that all submissions must include. It asks ~15 questions covering reproducibility, ethics, limitations, broader impact, and experimental rigor. Each question gets one of three answers: Yes, No, or NA, plus a short justification.

The checklist sits at the end of your paper (after the bibliography). It does not count toward the 9-page main paper limit — fill it in completely without worrying about space.

Checklist questions, by section

1. Claims
  • Do the main claims made in the abstract and introduction accurately reflect the paper's contributions and scope?
How to answer well: Answer 'Yes' if your abstract claims match what the paper actually proves/shows. If you over-claim in the abstract, fix it before submission — reviewers will catch it and downgrade your score.
2. Limitations
  • Does the paper discuss the limitations of the work performed by the authors?
How to answer well: You must include a Limitations section (or paragraph). 'No' is rarely acceptable. Be honest — every method has failure cases, and reviewers respect papers that name them.
3. Theory & Proofs
  • For each theoretical result, does the paper provide the full set of assumptions and a complete (and correct) proof?
How to answer well: Proofs go in the appendix if they don't fit. Reference the appendix proof in the main body. 'NA' if your paper is empirical-only.
4. Reproducibility
  • Does the paper fully disclose all the information needed to reproduce the main experimental results of the paper to the extent that it affects the main claims?
How to answer well: List datasets, model architectures, hyperparameter ranges, optimizer settings, and compute requirements. Code release is not strictly required but heavily recommended.
5. Code & Data
  • Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results?
How to answer well: Anonymous code repositories (anonymous.4open.science, supplementary .zip) work for double-blind. 'No' is acceptable but lowers your reproducibility score.
6. Experimental Details
  • Does the paper specify all the training and test details necessary to understand the results?
How to answer well: Be specific: dataset version, train/val/test split, batch size, learning rate, # of epochs, # of seeds, hardware. Vague answers tank reviewer scores.
7. Statistical Significance
  • Does the paper report error bars suitably and correctly defined or other appropriate information about the statistical significance of the experiments?
How to answer well: Run multiple seeds, report mean ± std (or 95% CI). Single-seed results without error bars are a common reviewer complaint.
8. Compute Resources
  • For each experiment, does the paper provide sufficient information on the computer resources (type of compute workers, memory, time of execution) needed to reproduce the experiments?
How to answer well: Specify GPU type (e.g., 8x A100 40GB), wall-clock time per run, and total compute used. This helps reviewers assess feasibility and the broader impact of compute cost.
9. Code of Ethics
  • Does the research conducted in the paper conform, in every respect, with the NeurIPS Code of Ethics?
How to answer well: Read the NeurIPS Code of Ethics before answering. Most papers answer 'Yes'. If your work has potential dual-use or societal harm, address it explicitly.
10. Broader Impacts
  • Does the paper discuss both potential positive societal impacts and negative societal impacts of the work performed?
How to answer well: A short Broader Impact section (or subsection) is standard. Discuss real impacts — generic 'AI could be used for good or evil' answers don't pass.

Frequently asked questions

Is the NeurIPS Paper Checklist mandatory?

Yes. Every submission must include a complete checklist. Submissions without a checklist are desk-rejected. The checklist appears at the end of your paper, after references and acknowledgments.

Does checklist.tex count toward the 9-page limit?

No. The NeurIPS Paper Checklist is appended after references and does not count toward the 9-page main paper limit. You can fill it in completely without worrying about page space.

Where do I get checklist.tex?

checklist.tex is bundled in the official NeurIPS author kit at media.neurips.cc and is pre-loaded in the TypeTeX NeurIPS 2026 template. Open the template and checklist.tex is already in place — just answer each question.

Can I answer 'NA' to questions?

Yes, when the question genuinely doesn't apply. For example, a purely empirical paper can answer 'NA' to the theoretical proofs question. Avoid using 'NA' as an escape hatch — reviewers can tell.

What if I answer 'No' to a checklist question?

It's allowed but you must justify why. 'No, code is proprietary' or 'No, this is preliminary work' is acceptable. Unjustified 'No' answers correlate with lower review scores.

Does the checklist count against me in review?

Reviewers see your answers and may use them to assess reproducibility and rigor. Honest, complete answers help — they show you've thought about your work carefully. Vague or evasive answers hurt.

More NeurIPS 2026 resources