AI and the World's Hard Problems · Physics · arXiv · 2026-10-23
Machine Learning Neutrino-Nucleus Cross Sections II: More Leptons
Neutrino-nucleus cross sections are essential for long-baseline oscillation experiments but hard to model precisely, and black-box data-driven models do not generalise between near and far detector fluxes. The authors extended a structure-function framework that imposes Standard Model symmetries to multiple lepton flavours and both beam modes, including electron appearance. The learned model recovered the injected oscillation parameters without significant bias, with constraints comparable to those from the exact cross section, and flux nuisance parameters had to be included for accurate results.