LLM-guided tree search enables autonomous disease forecasting

A new study on arXiv (2605.16238) introduces an autonomous system for multi-pathogen disease forecasting using Large Language Model (LLM)-guided tree search. The system iteratively generates, evaluates, and optimizes executable forecasting software without human intervention. In a prospective real-time evaluation during the 2025-2026 US respiratory season, it autonomously discovered models for influenza, COVID-19, and respiratory syncytial virus (RSV). The aggregated ensemble matched or outperformed the human-curated CDC hub ensembles, and successfully handled data-scarce 'cold start' scenarios for RSV. This approach addresses scalability bottlenecks in manual model curation, enabling granular geographic and emerging pathogen forecasting.