research-paper-writing

Follow the phased pipeline: set up the workspace and contribution statement, run a breadth-then-depth literature search, map every experiment to a paper claim, analyze results with significance tests, then draft sections grounded in a structured experiment log. Finish with a simulated self-review and venue checklist before submission.

Never generate BibTeX from memory. Use the 5-step workflow: search Semantic Scholar, verify the paper exists in two sources, retrieve BibTeX via DOI content negotiation, validate the cited claim appears in the paper, then add it to the bibliography. Mark unverifiable entries as [CITATION NEEDED].

The templates directory includes official style files for ICML 2026, ICLR 2026, NeurIPS 2025, ACL, AAAI 2026, and COLM 2025, each with the .sty file, bibliography style, and example document. Compilation works via latexmk, pdflatex with bibtex, VS Code LaTeX Workshop, or Overleaf.

Yes. It covers annotator selection, pairwise versus Likert scale design, inter-annotator agreement metrics like Krippendorff's alpha and Cohen's kappa, crowdsourcing platform tradeoffs, compensation reporting, and IRB guidance required by ACL and similar venues.

Report McNemar's test for paired binary outcomes, bootstrapped 95% confidence intervals, and effect sizes such as Cohen's h. Always state sample sizes, number of runs, and whether error bars show standard deviation or standard error.

Use autoreason-style iterative refinement for mid-tier models and constrained tasks where the generation-evaluation gap is large. Use single-pass or critique-and-revise for frontier models on unconstrained tasks, template-filling tasks, or very weak models that cannot generate diverse candidates.