Multimodal Generative Flows for LHC Jets
By: Darius A. Faroughy, Manfred Opper, Cesar Ojeda
Potential Business Impact:
Makes computer simulations of particle collisions better.
Generative modeling of high-energy collisions at the Large Hadron Collider (LHC) offers a data-driven route to simulations, anomaly detection, among other applications. A central challenge lies in the hybrid nature of particle-cloud data: each particle carries continuous kinematic features and discrete quantum numbers such as charge and flavor. We introduce a transformer-based multimodal flow that extends flow-matching with a continuous-time Markov jump bridge to jointly model LHC jets with both modalities. Trained on CMS Open Data, our model can generate high fidelity jets with realistic kinematics, jet substructure and flavor composition.
Similar Papers
Machine-learning based particle-flow algorithm in CMS
High Energy Physics - Experiment
Helps scientists see tiny particles better.
Jet Image Generation in High Energy Physics Using Diffusion Models
High Energy Physics - Phenomenology
Creates realistic pictures of tiny particle collisions.
MusFlow: Multimodal Music Generation via Conditional Flow Matching
Sound
Makes music from pictures, stories, or words.