Salvato in:
Dettagli Bibliografici
Autori principali: Hildebrandt, Richard, Kourlitis, Evangelos, Hashemi, Baran, Bünstorf, Manuel, Meyer, Thierry, Boskov, Nikola, Kagan, Michael, Rosenbaum, Dan, Ganguly, Sanmay, Heinrich, Lukas
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2605.06591
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911658317709312
author Hildebrandt, Richard
Kourlitis, Evangelos
Hashemi, Baran
Bünstorf, Manuel
Meyer, Thierry
Boskov, Nikola
Kagan, Michael
Rosenbaum, Dan
Ganguly, Sanmay
Heinrich, Lukas
author_facet Hildebrandt, Richard
Kourlitis, Evangelos
Hashemi, Baran
Bünstorf, Manuel
Meyer, Thierry
Boskov, Nikola
Kagan, Michael
Rosenbaum, Dan
Ganguly, Sanmay
Heinrich, Lukas
contents We introduce a new strategy for compositional neural surrogates for radiation-matter interactions, a key task spanning domains from particle physics through nuclear and space engineering to medical physics. Exploiting the locality and the Markov nature of particle interactions, we create a \emph{next-particle prediction} kernel using hybrid discrete-continuous transformer models based on Riemannian Flow Matching on product manifolds. The model generates variable-sized typed sets of particles and radiation side effects that are the result of the interaction of an incident particle with a material volume. The resulting kernel can be composed to simulate unseen large-scale material distributions in a zero-shot manner. Unlike mechanistic simulators, our model is designed to be differentiable, provides tractable likelihoods for future downstream applications. A significant computational speed-up on GPU compared to CPU-bound mechanistic simulation is observed for single-kernel execution. We evaluate the model at the kernel level and demonstrate predictive stability over multi-round autoregressive rollouts. We additionally release a novel 20M-event radiation-matter interaction dataset for further research.
format Preprint
id arxiv_https___arxiv_org_abs_2605_06591
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle BRICKS: Compositional Neural Markov Kernels for Zero-Shot Radiation-Matter Simulation
Hildebrandt, Richard
Kourlitis, Evangelos
Hashemi, Baran
Bünstorf, Manuel
Meyer, Thierry
Boskov, Nikola
Kagan, Michael
Rosenbaum, Dan
Ganguly, Sanmay
Heinrich, Lukas
Machine Learning
High Energy Physics - Phenomenology
We introduce a new strategy for compositional neural surrogates for radiation-matter interactions, a key task spanning domains from particle physics through nuclear and space engineering to medical physics. Exploiting the locality and the Markov nature of particle interactions, we create a \emph{next-particle prediction} kernel using hybrid discrete-continuous transformer models based on Riemannian Flow Matching on product manifolds. The model generates variable-sized typed sets of particles and radiation side effects that are the result of the interaction of an incident particle with a material volume. The resulting kernel can be composed to simulate unseen large-scale material distributions in a zero-shot manner. Unlike mechanistic simulators, our model is designed to be differentiable, provides tractable likelihoods for future downstream applications. A significant computational speed-up on GPU compared to CPU-bound mechanistic simulation is observed for single-kernel execution. We evaluate the model at the kernel level and demonstrate predictive stability over multi-round autoregressive rollouts. We additionally release a novel 20M-event radiation-matter interaction dataset for further research.
title BRICKS: Compositional Neural Markov Kernels for Zero-Shot Radiation-Matter Simulation
topic Machine Learning
High Energy Physics - Phenomenology
url https://arxiv.org/abs/2605.06591