An AI-driven agent designed a fundamental physics formula for chemical bonds that beats the gold-standard humans spent decades perfecting.
Density functionals are the mathematical blueprints used to simulate how electrons behave in molecules and materials. For fifty years, the best versions were hand-crafted by the world's top theoretical physicists. This new system, called SAFS26-a, was discovered autonomously by a large language model and reduces errors in thermochemistry by nine percent compared to the previous best. This shift suggests that the era of human-designed physics constants is ending. We are moving toward a future where AI handles the deep math of discovery while humans focus on the applications. This could lead to a massive acceleration in how we design new drugs and materials.
Agentic Discovery of Exchange-Correlation Density Functionals
arXiv · 2605.05460
The development of accurate exchange-correlation (XC) functionals remains a longstanding challenge in density functional theory (DFT). The vast majority of XC functionals have been hand designed by human researchers combining physical insight, exact constraints, and empirical fitting. Recent advances in large language models enable a systematic, automated alternative to this human-driven design loop. This report presents an agentic search system in which an LLM proposes structured functional-for