Research Artifacts: Class Imbalance and Batch Effects in LLM-Based Screening for Systematic Reviews

This record contains artifacts associated with a manuscript accepted at ENIAC 2026 (XXIII Encontro Nacional de Inteligência Artificial e Computacional), on class imbalance and batch effects in LLM-based title–abstract screening for systematic reviews. The artifact package includes spreadsheet tables and Jupyter notebooks used to prepare the data, process model outputs, and analyze the experimental results. The study follows a 2×2 factorial design comparing individual versus batch processing and prompts with versus without screening metadata. The underlying systematic review data were obtained from the SESR-Eval dataset repository (Huotala, Kuutila & Mäntylä, 2025; https://doi.org/10.5281/zenodo.16408882).