GPT-Based Fast Simulation of CLAS12 Detector Hits via Conditional Autoregressive Generation

2026-06-14

AI Summary

This work addresses the high computational cost of traditional Geant4-based particle detector simulations, which struggle to meet the demands of high-luminosity experiments requiring both efficiency and high fidelity. The authors propose the first application of a GPT-like conditional autoregressive Transformer to high-energy physics detector simulation, conditioning on incident particle momentum to encode the nine-layer response of the CLAS12 electromagnetic calorimeter as sequences of strip, ADC, and TDC tokens. The model autoregressively generates realistic hit signals layer by layer. By combining sequential data representation with GPU-accelerated inference, the method achieves over 700 events per second on a single GPU while accurately reproducing key physical characteristics—including hit multiplicity, spatial distributions, energy deposition, and energy–momentum response—demonstrating significantly improved simulation efficiency without sacrificing fidelity.

Abstract

Modern particles physics experiments have demonstrated an increasing need for fast, high-fidelity detector simulation as detector components have improved and subsequent computational requirements approach the limits of available resources. Recently, deep generative models have emerged as a promising alternative to traditional Monte-Carlo methods, with recent works drawing inspiration from large language models (LLMs) and self-supervised next-token prediction methods. In this work, we present an application of a GPT-style autoregressive transformer as a fast surrogate model for the calorimeter inside the CLAS12 experiment at the Thomas Jefferson National Accelerator Facility. The model is conditioned on incident momentum and generates realistic detector hits autoregressively across all nine calorimeter layers as sequences of strip, ADC, and TDC tokens. We demonstrate that the model faithfully reproduces hit multiplicity, spatial distributions, energy deposits, and the energy-momentum response of the electromagnetic calorimeter. The generator achieves inference rates exceeding 700 events per second on a single GPU, providing a substantial speedup over traditional Geant4-based simulations while maintaining physics fidelity essential for high-luminosity experimental programs.

Problem

• fast simulation
• detector simulation
• high-fidelity
• computational efficiency
• particle physics

Innovation

• autoregressive transformer
• fast detector simulation
• conditional generation
• GPT-based surrogate model
• calorimeter hit generation