Collider Physics Measurements in High Jet Multiplicity Final States

The CMS Experiment has collected an unprecedented amount of data during the 2016-2018 data taking period. Given the efficient lepton reconstruction, measurements targeting final states with electrons and muons are often preferred. On the other hand, final states without leptons are often more abundant due to the large cross sections from Quantum Chromodynamics (QCD) production, while providing complementary information from measurements targeting leptonic final states. Novel data analysis methods, such as machine learning, can improve the experimental reach from fully hadronic final states by combining physics knowledge and modern data analysis methods. In this work, new machine learning methods are developed and applied to analyses using the data collected by the CMS experiment between 2016 and 2018 data taking. The methods include supervised, weakly-supervised, and unsupervised classifiers. These methods are then applied to two results: the first measurement of the top quark pair production cross section in association with additional b-quarks ($\ttbb$) and the first search of vector-like leptons to third generation fermions. In the former, a weakly supervised classifier is developed to separate the signal process from the QCD background by training directly on data. This strategy resulted in the first determination of the $\ttbb$ cross section in the all-jets channel, measured to be 5.5 $\pm 0.3$ (stat) $^{+1.6}_{-1.3}$ (syst) pb in the full phase space. In the second result, machine learning is used to extract information from a high jet multiplicity final state through the usage of a graph neural network that learns the kinematic properties of the signal through the correlations between the final state particles. Expected exclusion limits on the vector-like lepton production through electroweak interactions are derived. Vector-like lepton masses up to 700 GeV are expected to be excluded with 95\% confidence interval in the absence of signal.