Research
Autonomous LLM-Guided Tree Search Outperforms Human Epidemiologists in Multi-Pathogen Forecasting
An autonomous system using LLM-guided tree search iteratively generated, evaluated, and optimized executable forecasting software for infectious diseases. In a fully prospective, real-time evaluation during the 2025–2026 season, the system matched or exceeded expert human modeling teams across multiple pathogens — without any domain-specific training or human model curation. This demonstrates LLM agents autonomously generating and validating scientific software that outperforms manual expert approaches.
Source
↳ Follow the thread