Paper

Zheyuan Zhang, Yuchen Wang, Jianfei Yu

Writing literature reviews is a time-consuming but essential part of academic research. We present AutoSurvey, a system that fully automates the generation of structured literature surveys. Given a research topic, AutoSurvey (1) retrieves relevant papers from academic databases, (2) extracts key claims, methods, and results, (3) organizes papers into thematic categories, and (4) synthesizes a coherent, citation-accurate survey. Human evaluation by 12 domain experts shows that AutoSurvey-generated surveys achieve a quality score of 4.1/5.0 compared to 4.4/5.0 for human-written surveys, while reducing production time from weeks to hours.