PRISMA-AI: Systematic Review Dataset with Human Annotations

PRISMA-AI Dataset This dataset contains 108 Creative Commons licensed systematic review papers with human annotations for PRISMA 2020 guideline compliance evaluation. Contents: 40 Emergency Medicine papers (Suda et al., 2025) 68 Rehabilitation papers (Tsuge et al., 2025) 257 human annotations (PRISMA 2020 checklist evaluations) Dataset Statistics: Papers: 108 CC-licensed systematic reviews Each paper includes detailed license information (e.g., CC-BY, CC-BY-ND) in its metadata Annotations: 257 human evaluations including other non-CC systematic reviews PRISMA Items: main body + abstract items Publication: Kataoka Y, So R, Banno M, Tsujimoto Y, Takayama T, Yamagishi Y, Tsuge T, Yamamoto N, Suda C, Furukawa TA. Large language models for automated PRISMA 2020 adherence checking. arXiv. 2025. Available from: http://arxiv.org/abs/2511.16707 Funding; This work was supported by a JSPS Grant-in-Aid for Scientific Research (Grant No. 25K13585) provided to Y.K.