Artificial intelligence (AI) methods are increasingly transforming data analysis in high-energy physics. Within the CMS experiment, AI is being investigated not only to enhance the efficiency and accuracy of established analyses, but also to enable novel strategies that move beyond traditional approaches. In particular, anomaly detection has drawn significant attention, as it enables model-independent searches for unexpected phenomena in collision data. Applications to jet analysis are being actively explored, offering promising avenues for uncovering subtle deviations from Standard Model expectations. Approaches such as autoencoders, normalizing flows, and weakly supervised learning are under systematic study and validation with both simulated and real CMS datasets. This talk will review the current progress of these developments, highlighting their potential impact on searches for new physics. I will also discuss their implications for online algorithms in the CMS trigger system, where AI-based anomaly detection could help identify unusual event signatures in real time.

castaned@cern.ch

Alfredo Martin Castañeda Hernandez