Simulation-rich training improves infectious disease forecasting with limited history
November 13, 2026
The Mantis 2 foundation model uses simulated outbreaks to improve disease forecasting. Research shows that simulation-based training is most effective when local observational history is limited and helps reduce exceptionally large forecast errors.
HOW THIS AFFECTS YOU
●
researcherYou can use mechanistic simulations to bolster foundation models in data-sparse environments.
●
healthThis method helps predict outbreaks more reliably when you don't have much local data to work with.