spinaote (Université de Geneve) will work on track classification tools for FTag algorithms in release 22. He will start from reviewing the tools available on the market or being proposed, comparing their standalone performance. The next step would then be to modify the FTag SS to make them interchangeable, so that their effect on the lo- and high-level algorithms can be tested. In the process, he will also perform a crosscheck on the coherent use of the same track labelling scheme throughout the FTag performance studies (from PhysVal to algorithm tuning tools). Some data/MC comparisons for the output of track classifiers will be another interesting byproduct. In addition, Sebastian will study the impact of adopting attention mechanisms for the tracks given in input to the Umami training framework; this will pave the way to a deeper interplay between track classification and taggers using tracks as input. All the developed procedures and corresponding software will be presented at FTag Software and Algorithms subgroup meetings, while code and its documentation will be made available in GitLab.