Empirical Study on Engineering Decisions in Multi-Agent Reinforcement Learning Systems

This replication package supports the paper “Empirical Study on Engineering Decisions in Multi-Agent Reinforcement Learning Systems,” accepted at SEAA 2026. The package contains the artifacts used for the empirical repository-mining study of open-source multi-agent reinforcement learning (MARL) repositories. It includes repository metadata, included and excluded repository lists, GitHub discovery queries, detector and analysis scripts, validation samples, manually labeled validation files, computed validation metrics, and artifacts used to reproduce the reported tables and figures. The dataset snapshot was collected in April 2026. The package is intended to support transparency, reproducibility, and independent inspection of the repository selection, detection, validation, and analysis steps reported in the paper.