Model-Agnostic Tagging of Quenched Jets in Heavy-Ion Collisions and Data-Driven Discovery of Quenching Observables

Measurements of jet substructure in ultra-relativistic heavy-ion collisions indicate that interactions with the quark-gluon plasma quench the jet showering process. Modern data-driven methods have shown promise in probing these modifications in the jet’s hard substructure. In this study, we present a machine learning framework to identify quenched jets while accounting for pileup, uncorrelated soft particle background, and detector effects — a more experimentally realistic and challenging scenario than previously addressed.

Our approach leverages an interpretable sequential attention-based model that integrates representations of individual jet constituents alongside global jet observables as features. The framework sets a new benchmark for tagging quenched jets with reduced model dependence, achieving a classification area under the curve (AUC) of 0.95, which significantly outperforms state-of-the-art techniques, including CNNs, DeepSets, and LSTM-based models, which typically achieve AUC values between 0.67 and 0.85.

Aided by the interpretability of our model, we also introduce "Quenchedness", a new closed-form substructure observable that encodes the model’s learned emphasis on hard-soft particle correlations and large-angle fragmentation, and is a transparent, one-dimensional discriminator with competitive performance, providing direct physical insight into jet–medium interactions while also matching existing state-of-the-art machine learning algorithms in classification ability.