Transforming literature screening: The emerging role of large language models in systematic reviews

摘要

Large language models (LLMs) can effectively be utilized for prefiltering scientific records for systematic reviews, leading to a substantial reduction in manual workload. Their performance in selecting scientific records based on titles and abstracts depends on the interplay between the inclusion and exclusion criteria and the LLM. Therefore, refining the formulation of the inclusion and exclusion criteria with the support of the LLM prior to title and abstract screening has the potential to improve the LLM’s performance in selecting relevant records.