OpenAI has published a new research note describing how its GPT-5.2 Pro model assisted physicists in deriving and verifying 'single-minus amplitudes' for gravitons — a step toward extending known techniques in scattering-amplitude calculations to quantum gravity.
Scattering amplitudes are mathematical expressions used to predict the outcomes of particle interactions; for gravitons, the hypothetical particles thought to mediate gravity, these calculations become extremely complex. The single-minus helicity amplitude is a simplified class of such calculations that researchers have used as a tractable entry point into the broader problem of unifying general relativity with quantum mechanics.
According to OpenAI's research page, GPT-5.2 Pro was used for two tasks in this work: helping derive the correct mathematical form of the graviton tree amplitudes through symbolic manipulation of the underlying equations, and helping verify the resulting formulas against known theoretical constraints.
OpenAI frames this as evidence that large language models can serve as active collaborators in abstract mathematical physics, not just as tools for data analysis or literature summarization. The published note does not detail the specific equations, error rates, or the extent of human oversight involved, so it should be read as an early demonstration rather than a completed physics result.