Polarising Perspectives: Unveiling the Exotic in LHC Hadronic Jets

One of the most rigorously tested theories in physics and our most reliable framework for understanding the universe, the Standard Model, still raises unanswered questions and remains incomplete. Classifying signal versus background events and differentiating vector boson polarisation states in the fully hadronic VVjj channel may uncover evidence of new exotic particles and provide deeper insights into the dynamics of particle interactions. Through detailed feature analyses and advanced machine learning techniques, key jet features such as the DisCo score, jet mass, number of tracks, and D2 score are identified as significant enhancers of signal vs. background discrimination. For polarisation states, jet substructure variables, including Zcut12 , Split12, Angularity, and KtDR, prove most effective. Robust classification models, specifically boosted decision trees and deep neural networks, achieve receiver operating characteristic (ROC) area under the curve (AUC) scores of 0.93 and 0.90, respectively, in the signal vs. background classification task. However, polarisation classification remains challenging, with ROC AUC scores around 0.73, highlighting the need for innovative techniques like decay angle regression. Preliminary results suggest that this method may improve experimental calibration, though challenges persist with lower cosθ predictions. Overall, the findings lay a solid foundation for future advancements in polarisation tagging by integrating machine learning with a comprehensive feature analysis.