#GPT-5.2# OpenAI GPT-5.2 physics# # ##
IBNS-CMEDIA: A team of physicists from leading institutions has collaborated with OpenAI’s GPT-5.2 model to achieve a new result in theoretical physics, the AI company announced.
The research focuses on gluons—the particles that mediate the strong nuclear force. The findings have been posted as a preprint on arXiv and are being submitted for peer-reviewed publication.
Titled “Single-minus gluon tree amplitudes are nonzero,” the paper is authored by Alfredo Guevara (Institute for Advanced Study), Alex Lupsasca (Vanderbilt University and OpenAI), David Skinner (University of Cambridge), Andrew Strominger (Harvard University), and Kevin Weil (OpenAI), on behalf of OpenAI.
At the heart of the study is a fundamental concept in particle physics known as a scattering amplitude—the mathematical quantity used to calculate the probability that particles interact in a specific way.
For gluons, many scattering amplitudes at “tree level” (calculations involving only the simplest diagrams without quantum loops) take unexpectedly simple forms. Such simplifications have historically uncovered deeper mathematical structures within quantum field theory, the framework that unifies special relativity and quantum mechanics.
According to OpenAI, an internally scaffolded version of GPT-5.2 spent approximately 12 hours reasoning through the problem, independently arriving at the same formula and producing a formal proof of its validity.
The resulting equation was analytically verified to satisfy the Berends–Giele recursion relation—a standard iterative method for constructing multi-particle tree amplitudes from simpler components. It was also tested against the soft theorem, which governs how scattering amplitudes behave when one of the particles becomes “soft,” or carries very low energy.
OpenAI added that, with GPT-5.2’s assistance, the results have already been extended from gluons to gravitons, with further generalizations underway. Additional AI-assisted findings are expected to be reported separately.
