Learning Standard Model structure from LHC data with Riemannian flow matching
Midori Kato, Kevin A. Urquía-Calderón, Inar Timiryasov, Oleg Ruchayskiy
Read on arXiv →
Key claim
Generative model captures extensive Standard Model features from data.
In plain English
In particle physics, accurately modeling events across a wide range of energies is challenging, as existing methods often rely on limited Monte Carlo samples. Current approaches struggle to capture the full complexity of interactions observed in high-energy collisions. This work introduces extsc{ShellFlow}, a generative model that learns directly from a vast dataset of real proton-proton collision events, enabling it to reproduce key features of the Standard Model. Builders in the field of physics data analysis might find this approach valuable for improving simulations and understanding particle interactions more comprehensively.
Introduces a novel generative model for particle physics that captures a wide range of invariant mass structures.
Trained on a large dataset of real collision events, demonstrating robust results across multiple physics phenomena.
Deep reliability assessment
The methodology supports the claim that the model can learn a substantial fraction of the Standard Model directly from collision data, but it may overclaim the generalizability across different collider environments without further validation.
Reproducibility
No open source code or dataset URL is mentioned in the paper.
Key figure
Figure 1 likely illustrates the architecture of the ShellFlow model, showing how it generates particles on their on-shell manifold using a Riemannian flow matching approach.