Transforming jet flavour tagging at ATLAS

TL;DR

This paper introduces GN2, a transformer-based jet flavour tagging algorithm for ATLAS that improves heavy-flavour jet identification by processing low-level tracking data with physics-informed training, validated in real collision data.

Contribution

The paper presents GN2, a novel transformer-based algorithm for jet flavour tagging that enhances interpretability and performance over previous methods in ATLAS.

Findings

C-jet rejection improved by a factor of 3.5 in data.

Light-jet rejection improved by a factor of 1.8 in data.

Validated performance in both simulation and collision data.

Abstract

Jet flavour tagging enables the identification of jets originating from heavy-flavour quarks in proton-proton collisions at the Large Hadron Collider, playing a critical role in its physics programmes. This paper presents GN2, a transformer-based flavour tagging algorithm deployed by the ATLAS Collaboration that represents a different methodology compared to previous approaches. Designed to classify jets based on the flavour of their constituent particles, GN2 processes low-level tracking information in an end-to-end architecture and incorporates physics-informed auxiliary training objectives to enhance both interpretability and performance. Its performance is validated in both simulation and collision data. The measured c-jet (light-jet) rejection in data is improved by a factor of 3.5 (1.8) for a 70% b-jet tagging efficiency, compared to the previous algorithm. GN2 provides substantial benefits for physics analyses involving heavy-flavour jets, such as measurements of Higgs boson pair production and the couplings of bottom and charm quarks to the Higgs boson, and demonstrates the impact of advanced machine learning methods in experimental particle physics.