Learning Standard Model structure from LHC data with Riemannian flow matching
In this work we demonstrate that a single transformer-based generative model can capture Standard Model structure spanning five decades of invariant mass, from the sub-GeV regime to the TeV continuum, a range that no single Monte Carlo sample covers. To achieve this we design \textsc{ShellFlow}, a Riemannian conditional flow matching model that, given the recorded event composi
Reading in Plain English mode — toggle below to switch
In this work we demonstrate that a single transformer-based generative model can capture Standard Model structure spanning five decades of invariant mass, from the sub-GeV regime to the TeV continuum, a range that no single Monte Carlo sample covers. To achieve this we design \textsc{ShellFlow}, a Riemannian conditional flow matching model that, given the recorded event composi
This model release advances AI systems and production ML, directly affecting how engineers build, optimize, and ship AI systems. Teams running production LLM workloads need to understand these techniques to stay competitive in hiring and architecture decisions.