PaperPanorama

arXiv:2509.26218·v2·High Energy Physics — Experiment

Event Tokenization and Masked-Token Prediction for Anomaly Detection at the Large Hadron Collider

Ambre Visive (1 and 2)🇳🇱 · Polina Moskvitina (3 and 2)🇳🇱 · Clara Nellist (1 and 2)🇳🇱 · Roberto Ruiz de Austri (4)🇪🇸 · Sascha Caron (3 and 2) ((1) Institute of Physics, University of Amsterdam, Amsterdam, The Netherlands, (2) Nikhef, Dutch National Institute for Subatomic Physics, Amsterdam, The Netherlands, (3) High Energy Physics, Radboud University, Nijmegen, The Netherlands, (4) Instituto de Física Corpuscular, IFIC-UV/CSIC, Paterna, Spain)🇳🇱

PDFarXivINSPIRE

Abstract

We propose a novel use of Large Language Models (LLMs) as unsupervised anomaly detectors in particle physics. Using lightweight LLM-like networks with encoder-based architectures trained to reconstruct background events via masked-token prediction, our method identifies anomalies through deviations in reconstruction performance, without prior knowledge of signal characteristics. Applied to searches for simultaneous four-top-quark production, this token-based approach shows competitive performance against established unsupervised methods and effectively captures subtle discrepancies in collider data, suggesting a promising direction for model-independent searches for new physics.

Comments: 5 pages, 3 figures, to be submitted to SciPost Physics Proceedings

Citation historyopen in Citation History ↗