research-paper-writing

Follow a phased pipeline: define a one-sentence contribution, map every experiment to a claim, run baselines with statistical tests, then draft using the venue's official LaTeX template. Finish with a simulated self-review and the venue's mandatory checklist before submitting.

Never generate BibTeX from memory—fetch it programmatically via DOI content negotiation from CrossRef or the Semantic Scholar API. Verify each paper exists in at least two sources, and mark anything unverifiable as a [CITATION NEEDED] placeholder.

Report error bars (stating std dev vs std error), 95% confidence intervals via bootstrapping, and McNemar's test for paired method comparisons. Include effect sizes like Cohen's h and state the number of runs or seeds for every key result.

Yes—NeurIPS desk-rejects submissions missing its 16-item checklist covering claims alignment, limitations, reproducibility, compute resources, and ethics. The checklist appears after references and does not count toward the page limit.

Use human evaluation when automated metrics miss what you claim—fluency, helpfulness, or preference in generation tasks, which ACL and EMNLP reviewers expect. Report annotator counts, Krippendorff's alpha agreement, compensation, and use pairwise comparisons over Likert scales for reliability.

NeurIPS allows 9 pages, ICML 8, ICLR 9, ACL 8 (long papers), AAAI 7, and COLM 9 for submissions, with camera-ready versions typically gaining one page. References and appendices are unlimited at all of these venues.