Orateur
Description
The High-Luminosity Large Hadron Collider (HL-LHC) will open new opportunities for the study of rare processes, including Higgs boson pair (di-Higgs) production. Flavour tagging will play a central role in enabling these measurements. This talk presents the expected challenges for flavour tagging in the HL-LHC environment, characterized by unprecedented instantaneous luminosity, high pile-up conditions, and the upgraded ATLAS detector. Particular emphasis is placed on the expected performance of the transformer neural network GN2 when trained and evaluated on simulated Run 4 datasets. In addition, we explore the potential of exploiting relaxed track selections within the GN2 architecture to further enhance flavour-tagging performance, discussing the trade-offs between potential gains in tagging performance and the increased computational resources required for track reconstruction. Finally, the implications for future di-Higgs analyses are discussed, highlighting the critical importance of flavour-tagging performance and reconstruction strategies for achieving sensitivity to this key measurement in the Higgs sector.