Anomaly Detection in the CMS Level 1 Trigger

AXOL1TL is a real-time, model-independent anomaly detection algorithm for the Compact Muon Solenoid (CMS) experiment at the Large Hadron Collider (LHC). It is designed to select rare, unexpected events without assuming a specific signal model. Operating at the 40 MHz LHC bunch-crossing rate, the CMS Level-1 Trigger must decide within 4 μs whether to keep or discard each event. We implement a machine-learning based autoencoder that flags unusual detector signatures while meeting strict latency and hardware constraints. First deployed in 2024, AXOL1TL enables CMS to record anomalous topologies for detailed offline analysis. We detail the low-latency implementation strategy, evaluate physics performance using simulated signal benchmarks, and summarize results from real-time monitoring and validation with proton collision data.