Comprehensive reconstruction of collider events with hypergraph representation learning and graph-conditioned diffusion
VyPER treats collider reconstruction as a joint hypergraph assignment and diffusion problem.
The paper splits reconstruction into assigning jets and charged leptons to parent particles, then predicting unseen neutrino kinematics. Its framework uses physics-shaped hypergraphs for the assignment side and a graph-conditioned diffusion model for neutrino prediction. The authors test it across several proton-proton collision processes and compare it with analytical and machine-learning reconstruction methods. They argue the results support accurate reconstruction across multiple Standard Model processes relevant to Higgs, electroweak, and top-quark measurements. ArXiv · AI/CL/LG's note
The paper splits reconstruction into assigning jets and charged leptons to parent particles, then predicting unseen neutrino kinematics. Its framework uses physics-shaped hypergraphs for the assignment side and a graph-conditioned diffusion model for neutrino prediction. The authors test it across several proton-proton collision processes and compare it with analytical and machine-learning reconstruction methods. They argue the results support accurate reconstruction across multiple Standard Model processes relevant to Higgs, electroweak, and top-quark measurements. ArXiv · AI/CL/LG's note
score 4