Princeton University Dark Cosmos Seminar
From Firmware to Final Inference: ML Driving Discovery at CMS
Abstract: Machine learning (ML) is an increasingly vital tool at the Compact Muon Solenoid (CMS) experiment at CERN’s Large Hadron Collider, from online data collection to offline interpretation. I will trace how ML is driving new-physics discovery end-to-end: first by reshaping data taking at the low level (Level-1) trigger, and then by pushing event-level reconstruction and analysis sensitivity beyond previous expectations. I begin with AXOL1TL (“Anomaly eXtraction Offline L1 Trigger Lightweight”), one of the first anomaly-detection, ML-based algorithms deployed in the CMS Level-1 trigger, and discuss how it informs the future of ML triggers at CMS. I then highlight analysis-level advances—ML-enhanced top and Higgs reconstruction, a sensitive H→WW measurement, and the first all-hadronic four-top analysis enabled by ML-driven background modeling. I close with an outlook to ML’s place in informing plans and challenges of the upcoming High Luminosity LHC upgrade and beyond. Together these pieces form a landscape of ML-driven discovery at CMS, from firmware to final inference.