AI benchmark for theoretical physics created by Raju and colleagues
Theoretical physicists are increasingly utilizing large language models to automate calculations and research, leading to a dramatic surge in paper submissions on arXiv. While some experts find AI helpful for catching up on literature and accelerating work, others express concern over the impact on students, the risk of getting scooped, and the challenge of navigating an endless stream of new research. Meanwhile, benchmark evaluations show that frontier AI models can solve complex theoretical physics problems with high accuracy, raising questions about the future role of human calculation in the field. Open article →