| CMS Detector Performance Summaries | |
| Report number | CMS-DP-2025-009 ; CERN-CMS-DP-2025-009 |
| Title | Performance of the ParticleNet b and bb-tagging algorithms in the CMS High-Level Trigger in Run 3 |
| Corporate author(s) | CMS Collaboration |
| Publication | 2025 |
| Collaboration | CMS Collaboration |
| Imprint | 08 Mar 2025 |
| Number of pages | 12 |
| Subject category | Detectors and Experimental Techniques |
| Accelerator/Facility, Experiment | CERN LHC ; CMS |
| Keywords | Trigger |
| Abstract | Dedicated ParticleNet AK4 b-tagging and AK8 bb-tagging algorithms, based on graph neural networks (GNN), for the jets reconstructed at the High-Level Trigger (HLT) were trained for the very first time in 2022 using simulated events. New triggers were developed based on these algorithms that have been integrated into the CMS HLT menu since the start of Run 3 data-taking. This note presents measurements of the ParticleNet b- and bb-tagging efficiencies using 2024 data, along with a detailed comparison of the performance with respect to that achieved under the 2022 and 2023 data-taking conditions. Additionally, the efficiency of the relevant triggers on simulated $HH \rightarrow 4b$ events is highlighted for both the resolved and high-$p_{T}$ merged topologies. |
Record created 2025-03-10, last modified 2025-03-10
