GPT-Based Fast Simulation of CLAS12 Detector Hits via Conditional Autoregressive Generation
Researchers have developed a GPT-style autoregressive transformer model to simulate detector hits for the CLAS12 experiment at the Thomas Jefferson National Accelerator Facility. This model, conditioned on incident momentum, generates realistic detector hits across nine calorimeter layers, reproducing key physics characteristics. The generative approach achieves inference speeds over 700 events per second on a single GPU, significantly outperforming traditional Geant4-based simulations while maintaining essential physics fidelity for high-luminosity experiments. AI
IMPACT Accelerates scientific discovery by enabling faster, high-fidelity simulations for particle physics experiments.