Speaker
Description
The efficient identification of jets originating from heavy-flavour quarks is essential for many measurements in High-Energy Physics, such as Higgs boson and top quark measurements as well as searches for physics beyond the Standard Model. Nevertheless, differences between measured data and the simulation on which the algorithms were trained can cause performance deficits and prediction discrepancies. For precision measurements and searches for new physics, however, these flavour tagging algorithms must be well-calibrated to avoid such problems by mitigating any discrepancies arising from the application of the flavour tagging algorithms. We will present an overview of the calibration methods and first Run Ⅲ results for data recorded by the CMS Experiment.
Moreover, as the flavour tagging algorithms are continuously improved, the calibrations need to be refined accordingly to ensure reliable performance. These new, more powerful flavour tagging algorithms are particularly prone to capturing mismodelling in higher-order correlations of the simulation. The impact of the state-of-the-art algorithms on the calibration and the effect of the integrated mitigation strategy based on adversarial attacks in one of the algorithms will be demonstrated and compared. Additionally, we will illustrate concepts of novel calibration techniques, including the application of machine learning developed within the CMS Collaboration.
