research-paper-writing

Follow the phased pipeline: set up the workspace and contribution statement, run an iterative literature search, map claims to experiments, analyze results with statistical tests, then draft using the venue's official LaTeX template. The skill provides templates and checklists for NeurIPS, ICML, ICLR, ACL, AAAI, and COLM.

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

Use McNemar's test for paired binary outcomes between two methods, bootstrapped 95% confidence intervals for key metrics, and Cohen's h for effect size. Always report sample sizes, number of runs, and whether error bars show standard deviation or standard error.

Yes, it bundles official templates for ICML 2026, ICLR 2026, NeurIPS 2025, ACL, AAAI 2026, and COLM 2025, each with the correct style files, bibliography styles, and example documents. Page limits and compilation instructions are documented per venue.

Use autoreason for mid-tier models (Haiku, Gemini Flash) and constrained tasks where the generation-evaluation gap is large. Use single pass or critique-and-revise for frontier models on unconstrained tasks, where autoreason suffers synthesis drift and fails to converge.

The pipeline assumes an empirical ML paper with runnable experiments; pure theory or survey papers need adapted phases. It also requires API access to Semantic Scholar and CrossRef for citation verification, and human evaluation studies still need external annotator recruitment and IRB approval.