Preview

PDF, English Download (4MB) | Terms of use

Abstract

The simulation chain of LHC physics is a well-established and indispensable toolkit for conducting precision measurements at general-purpose detectors at the LHC. While not all components are derived from first-principle physics, the simulation chain as a whole has demonstrated remarkable accuracy and reliability over the past decade. To keep pace with increasing experimental demands and growing data statistics, the simulation chain undergoes continuous refinement. In recent years, the rise of machine learning has opened new avenues for further advancing the simulation and analysis pipeline of high energy physics. In this work, we explore the integration of state-of-the-art generative machine learning algorithms into different stages of the LHC simulation chain. First, we test their ability to enhance forward simulations by improving the generation speed, particularly in computationally intensive steps. Second, we apply generative machine learning models to the inverse problem of unfolding detector effects, offering an alternative to traditional techniques. We show that both tasks can be solved using machine learning with high precision and accuracy, demonstrating the potential of these approaches to significantly improve the scalability and robustness of future LHC analyses.