AISysRev: LLM-Powered Tool Accelerates Academic Paper Screening

AISysRev is an innovative web application in a containerized format that leverages large language models to streamline the title-abstract screening process for systematic reviews. Users can upload CSV files with paper titles and abstracts, while specifying their own inclusion and exclusion criteria. The application is compatible with various LLMs, such as Gemini, Claude, Mistral, and ChatGPT via OpenRouter, as well as locally hosted models and any that work with the OpenAI SDK. It utilizes both zero-shot and few-shot prompting techniques and features interfaces that present LLM outputs to assist human reviewers during manual screening. By enabling parallel LLM calls, AISysRev can screen between 100 to 300 papers per minute, significantly alleviating the burdensome task of preparing systematic reviews. This tool effectively tackles the substantial number of papers researchers encounter during the selection phase of systematic reviews.