An X-ray spectrum is a sound recording left by electrons in a magnetic material. Previously, scientists only listened to the volume, but now machine learning picks out the parts of two dancers: the electron's own spin and its orbital race. Using polarized light sensitive to the rotation direction, physicists trained a neural network on thousands of spectra created in computer simulations with varying internal field strengths and spin-orbit coupling. The network mastered reading the line shape as confidently as a conductor hears each instrument.

This approach turns quantum sensing into a super-precise measurer, paving the way for new magnetic devices.