> Activities > WP3
Task 3.6: Practical real-time AI for Level 1 Trigger and L1 Scouting
Task leads: Sioni Summers (CERN)
For this task, we propose to research and develop methods to make optimal use of the information that is available in the trigger system, and a system to deploy models with robust provenance tracking and reproducibility. We anticipate that Machine Learning will be prevalent throughout the CMS L1T during the HL-LHC era, with around 20 models and 50 billion inferences per second already accounted for. Developing and operating the experiment with this large amount of ML in the data acquisition pipeline is a new frontier for CMS.
Publications and other resources
- CMS Collaboration, Detector Performance Note 22/10/2024, “Electron Reconstruction and Identification in the CMS Phase-2 Level-1 Trigger”
- Piero Viscone, Presentation at CHEP 23/10/2024, “Low-latency AI for triggering on electrons at High Luminosity LHC with the CMS Level-1 hardware Trigger”
- Sioni Summers, Presentation at INFN Rome 17/03/2025, “Next Generation Triggers for CMS — Level 1 Trigger and Scouting”
- CMS Collaboration, Detector Performance Note 18/06/2025, “Multiclass object tagging in the CMS Phase-2 Level-1 Correlator Trigger”
- Stella Felice Shaefer, Presentation at EuCAIFCon 19/06/2025, “Advancing the CMS Level-1 Trigger: Jet Tagging with DeepSets at the HL-LHC“
- Duc Minh Hoang, Poster at EPS-HEP 09/07/2025, “Advancing the CMS Level-1 Trigger: Jet Tagging with DeepSets at the HL-LHC”