Exploring jets: substructure and flavour tagging in CMS and ATLAS

TL;DR

This paper reviews advanced jet identification and calibration techniques used by CMS and ATLAS, emphasizing neural networks, attention mechanisms, and adversarial training to improve particle physics analyses.

Contribution

It provides a comprehensive comparison of the latest methods in jet substructure and flavor tagging, highlighting innovations and limitations in current approaches.

Findings

Neural network architectures enhance jet identification accuracy.

Attention mechanisms improve feature extraction in jet analysis.

Adversarial training helps reduce systematic uncertainties.

Abstract

The identification and characterization of jets are crucial tasks for effectively probing fundamental particle interactions. The ATLAS and CMS experiments have developed cutting-edge techniques to improve jet identification and calibration, employing innovative approaches including advanced neural network architectures, attention-based mechanisms, and adversarial training. These proceedings provide a comprehensive review of the state-of-the-art methods employed by both collaborations, highlighting their similarities, unique strengths, and limitations through a comparative analysis.