research-paper-writing

Follow a phased pipeline: define a one-sentence contribution, map every experiment to a paper claim, run baselines with statistical tests, then draft using the venue's official LaTeX template. Complete the mandatory NeurIPS checklist or ICML reproducibility statement before submission to avoid desk rejection.

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 confirm the cited claim actually appears in the abstract before adding it to your bibliography.

Use McNemar's test for comparing two methods on the same problems, bootstrapped confidence intervals for key metrics, and Cohen's h for effect size on proportions. Always report sample sizes, number of runs, and whether error bars show standard deviation or standard error.

Use the official style files: iclr2026_conference.sty with the submission option for ICLR, and aaai2026.sty for AAAI, which enforces the strictest formatting rules. The templates directory includes ICML 2026, NeurIPS 2025, ACL, and COLM 2025 variants with example documents.

Use autoreason-style iterative refinement for mid-tier models on constrained tasks, where the generation-evaluation gap is largest. For frontier models on unconstrained tasks, single-pass or critique-and-revise performs better because synthesis drift prevents convergence.

NeurIPS allows 9 pages, ICML 8 pages, ICLR 9 pages, ACL 8 pages for long papers, AAAI 7 pages, and COLM 9 pages for main content. References and appendices are unlimited at all of these venues, and camera-ready versions typically get one extra page.