Generative AI for Simulations in NP

Generative AI methods offer a promising alternative to traditional simulation chains. This tutorial presents a practical introduction to applying generative models for fast simulation of detector responses. Using a toy simulation of a pixel-based detector, participants will explore how models are trained on particle-level features to generate realistic hit patterns under noise and resolution effects. The session will focus on data preprocessing, architecture design principles and evaluation metrics such fidelity scores. The tutorial targets nuclear physics researchers aiming to accelerate simulation pipelines with learned surrogates that maintain physics fidelity.