OmniJet-${\alpha}_C$: Learning point cloud calorimeter simulations using generative transformers

TL;DR

This paper introduces OmniJet-α_C, a generative transformer model that creates realistic calorimeter shower simulations as point clouds, capturing complex geometries without fixed grid constraints.

Contribution

It is the first application of generative transformers to calorimeter shower simulation as point clouds, enabling variable-length, geometry-aware modeling of detector hits.

Findings

Successfully generates realistic calorimeter showers

Supports variable-length sequences for shower development

Learns complex geometries without voxel grid restrictions

Abstract

We show the first use of generative transformers for generating calorimeter showers as point clouds in a high-granularity calorimeter. Using the tokenizer and generative part of the OmniJet-${\alpha}$ model, we represent the hits in the detector as sequences of integers. This model allows variable-length sequences, which means that it supports realistic shower development and does not need to be conditioned on the number of hits. Since the tokenization represents the showers as point clouds, the model learns the geometry of the showers without being restricted to any particular voxel grid.