Speaker
Description
The identification of jets arising from heavy-flavor (bottom or charm) quarks is crucial for many analyses in CMS, such as in top quark and Higgs boson measurements. Heavy-flavor tagging algorithms rely on the kinematics of reconstructed charged particle tracks and secondary vertices. These algorithms are typically trained using simulations that are prone to mismodeling. Hence it is crucial to study the agreement between the simulated tagger inputs and output scores, and the corresponding properties in data.
In this poster, we present such comparisons using early Run-3 data recorded by the CMS experiment. The results provide a first estimate of flavor-tagging performance with Run-3 data and reliability of simulations and tagging algorithms. These studies were performed using a fast, computationally-efficient framework built upon state-of-the-art data analysis tools. The framework is designed to analyze vast amounts of collision data and perform several kinds of event selections, each enriched in a certain flavour of jets. The framework utilizes modern columnar analysis methods with integrated high-thoughput computing options and is expected to make future workflows more automated and prompt.
