Speaker
Description
We present a search for strongly coupled dark matter in the form of semi-visible jets (SVJs) using the full Run~2 CMS dataset. SVJs arise in Hidden Valley models with a confining dark QCD sector, where dark quarks produced via \textit{t}-channel scalar exchange hadronize into dark hadrons---some decaying to Standard Model particles and others escaping detection. This results in jets with both visible and invisible components, and missing transverse momentum ($p_T^{\text{miss}}$) that is aligned with a jet rather than isotropic. After a set of event preselection requirements, an event-level deep neural network (DNN) trained with a Lagrange multiplier to ensure independence from $p_T^{\text{miss}}$ is applied to select candidate events. This DNN output is used to define the ABCD plane for data-driven background estimation. Within each selected event, individual jets are tagged using both supervised and unsupervised SVJ taggers. The ABCD method is used to estimate Standard Model backgrounds and validate closure directly in data using low-signal control regions, ensuring reliable background predictions in the presence of semi-visible signals.
| Progress report area | Lightning round |
|---|
