Anomaly Detection in CMS

With current advancements in computational resources and algorithmic developments, the CMS experiment at the LHC has been incorporating machine learning (ML) techniques to further enhance its physics potential. While ML offers powerful computational tools, the foundational building blocks remain rooted in the underlying physics phenomena. These advancements have significantly improved the search for new physics, allowing physicists to conduct more effective searches and measurements while enabling innovative approaches to data analysis. In these proceedings, we present our recent advancements in ML techniques applied to anomaly detection within the CMS experiment, focusing on both dijet final states and enhancements at the Level-1 (L1) trigger. Our work includes novel methods for identifying anomalous jet substructures, enhancing the discovery potential of new physics signatures that were previously unexplored. Additionally, we discuss the implementation of ML for anomaly detection at the L1 trigger, underscoring its potential to improve the early selection of interesting events. Our findings illustrate the efficacy of these approaches in maximizing sensitivity to rare events, contributing to the ongoing efforts to unravel the mysteries of the universe.