hzwer/writingaipaper

This repository is a written guide for beginners trying to publish research papers at AI conferences. It is aimed at people who have done experiments and have results to share, but are not sure how to turn that work into a proper academic paper. The authors are researchers who have published at major venues and want to pass on practical advice that textbooks rarely cover. The guide covers two main areas. The first is how to build a paper from scratch: figuring out what your core contribution is, structuring the abstract and introduction, writing the related work section, and presenting your method and experiments clearly. The second area is about polishing the details: improving readability, making figures clear, checking the writing for common mistakes, and handling the revision and rebuttal process after a paper is submitted and reviewers respond. A recurring theme in the guide is the difference between publishing results and sharing genuine new knowledge. The authors push back against the idea that beating a benchmark score is enough. They argue that the most lasting papers are ones that surface a clear insight, introduce a new capability, or offer an explanation for something people did not understand before. They also spend time on the practical side of finishing a paper under deadline pressure, including when the first draft should be done and how to divide work among collaborators. The guide is written in English and was also published across several Chinese platforms, where it reached a large audience. It is not a software tool, it contains no code to run. The value is entirely in the text itself, which is structured as a long article with takeaway summaries after each section to make it easier to skim and revisit.