This research introduces a breakthrough approach to multi-task Bayesian optimization by leveraging large language models to learn from thousands of previous optimization processes.

• Scales to approximately 2,000 tasks - far beyond previous methods
• Dramatically improves efficiency of optimizing new tasks across domains
• Successfully applied to antimicrobial peptide design in biology
• Represents a significant advancement for drug discovery and other biological optimization problems

For biology applications, this means faster identification of effective antimicrobial compounds, accelerated drug development cycles, and more efficient exploration of complex biological design spaces.

Large Scale Multi-Task Bayesian Optimization with Large Language Models