Overview

• PandemicLLM uses generative AI to reason about outbreak trajectories one to three weeks ahead with high precision.
• The model ingests state-level demographic, epidemiological, policy and genomic surveillance data to capture evolving outbreak drivers.
• Retrospective tests on 19 months of COVID-19 data across U.S. states showed it outperformed CDC-endorsed forecasting tools during volatile periods.
• Designed for adaptability, PandemicLLM can be customized for influenza, H5N1, RSV, monkeypox and other emerging pathogens.
• Researchers are now integrating behavioral decision-making simulations to enhance policy design and response planning.