Research Artifacts for Edge-Deployed Multi-Agent LLM System for Emergency Lighting Fault Diagnosis and Remediation

This record provides the research artifacts accompanying the following paper: A. Gijón, F. Rincón, X. del Toro, F. J. Villanueva, M. J. Santofimia, and J. C. López, "Edge-Deployed Multi-Agent LLM System for Emergency Lighting Fault Diagnosis and Remediation," submitted to IEEE Internet of Things Journal, 2026. It facilitates the reproducibility of the experimental evaluation presented in the paper by making available all non-proprietary research artifacts: the system prompts of both agents and of the LLM-as-a-Judge, the input telemetry of the eight fault scenarios, the complete raw outputs and metrics of the twenty evaluated models over five independent replicas each, the judge assessments, and the replica-stability analysis.