Triggering the Unknown: Model-Independent Discovery with Intelligent Detectors

At the Large Hadron Collider (LHC), the coming decade will see discovery potential determined not only by luminosity but by the intelligence embedded inside the detector itself. In the absence of any clear signs for physics beyond the standard model, two options remain: either new physics is beyond the reach at the LHC or we are not looking in the right places. In response, new colliders have been proposed to take us beyond the LHC physics reach and new theoretical models (e.g. dark sector) have been proposed with signatures that could be hiding amongst the trillions upon trillions of proton interactions that take place at the LHC. The Compact Muon Solenoid (CMS) is a general-purpose detector at the LHC that sees petabytes worth of data each second during proton-proton collisions. At CMS a real-time trigger system built upon FPGAs is used to select interesting interactions. Recently this system has begun to incorporate unsupervised machine learning to flag anomalous events within nanoseconds for further analysis. This talk explores recent developments in trigger-level anomaly detection, including FPGA-deployed networks and the broader integration of graph networks, quantized transformers, and ASICs. These innovations point the way toward scalable, intelligent data selection for the HL-LHC, Future Circular Collider (RCC), and future muon colliders, enabling a broader, more agnostic approach to discovery.

Speaker: Isabel Rose Ojalvo, Princeton University