Erlangen 2026 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 56: Data, AI, Computing, Electronics VI

T 56.3: Vortrag

Mittwoch, 18. März 2026, 16:45–17:00, KH 02.014

Foundation Model Based Approaches to Jet Tagging Using JEPA — Mirac Noyan Özdemir, Alexander Schmidt, •Sergio Schöneberg, and Ulrich Willemsen — III. Physikalisches Institut A, RWTH Aachen University

Foundation Models have gained significant traction in Natural Language Processing and Computer Vision and have recently become the focus of intensive research in the field of high-energy physics as they provide highly adaptable models that require less task-specific training and data. Moreover, their ability to leverage unlabelled data and their capacity to derive meaningful representations are promising opportunities for improving model robustness.

This talk presents the application of the Joint-Embedding Predictive Architecture (JEPA) to jet physics, capabilities of few-shot learning are demonstrated and the viability for jet flavor tagging is discussed. Finally, the broader implications for AI safety within high-energy physics are addressed.

Keywords: Foundation Models; Jet Tagging; AI Safety; Machine Learning