Forecasting Generative Amplification
This paper introduces two complementary methods, averaging and differential amplification, to estimate the statistical precision of generative networks for LHC simulations without large holdout datasets, demonstrating that event amplification is feasible in specific phase-space regions.