| name | neurips-submission |
| description | Use when auditing a NeurIPS main-track submission for current-cycle CFP, OpenReview, formatting, anonymity, track, contribution-type, checklist, code/data, dual-submission, and LLM/agent policy compliance. |
NeurIPS Submission
Use this skill for a submission-readiness audit before a paper enters OpenReview. Treat every
annual rule as volatile: reopen the current CFP, Main Track Handbook, paper template, and submission
portal before final advice.
Core checks
- Target the right track. NeurIPS main track, Evaluations & Datasets, Position Papers,
Reproducibility/MLRC, competitions, workshops, and Journal-to-Conference tracks have separate
calls and portals; do not assume transfer is possible after submission.
- Confirm that all authors have active OpenReview profiles and that author names are entered by the
abstract deadline. After that, only author order may change under the 2026 handbook.
- Verify the yearly LaTeX style file, PDF-only submission format, content-page limit, file-size
limit, and whether references, text appendices, checklist, and code/data are handled separately.
- Confirm double-blind compliance across the main PDF, supplement, repository links, uploaded ZIPs,
acknowledgments, self-citations, prompts, demos, model cards, dataset cards, and metadata.
- Check dual-submission exposure. Overlapping archival submissions can cause desk rejection; do not
submit the same work to another archival venue while it is under NeurIPS review.
NeurIPS-specific risks
- Missing checklist is a desk-rejection risk.
- A submission that has strong benchmark wins but weak scientific explanation is vulnerable.
- A paper with copied, hallucinated, or unverifiable citations creates academic-integrity risk.
- Public preprints are allowed in the 2026 policy, but aggressive advertising of a submitted paper
can violate the spirit of double-blind review.
- Non-standard agent or LLM use that is part of the method should be documented in the experimental
setup; ordinary editing and basic code assistance generally do not need a methods statement.
Audit workflow
- Identify the intended track and contribution type: General, Theory, Use-Inspired,
Concept & Feasibility, Negative Results, or the current-cycle equivalent.
- Compare the PDF to the live template and page rules.
- Walk every link, repository, dataset, artifact, and supplement for anonymity.
- Confirm code/data ZIP handling and whether the artifact is review-only or intended for public
release after acceptance.
- Inspect the checklist answers for unsupported "yes" responses, missing justifications, and
claims that are broader than the evidence.
- Record all unresolved current-cycle items in the output.
Output format
[Submission readiness] Ready / Needs fixes / Not ready
[Track] Main / E&D / Position / MLRC / other
[Contribution type] General / Theory / Use-Inspired / Concept & Feasibility / Negative Results / current-cycle label
[Blocking official checks] <CFP/template/OpenReview/anonymity/checklist/code-data/dual-submission>
[Highest desk-reject risk] <one issue>
[Fix list] <ordered fixes before submission>
