JetPrism: Model Checkpoints and Dataset for Generative Simulation and Inverse Problems in Nuclear Physics

This repository contains the trained model checkpoints and MC-POM dataset for JetPrism, a Conditional Flow Matching (CFM) framework for kinematic event generation and detector unfolding in nuclear physics. The dataset comprises 8,000,000 events for the exclusive photoproduction reaction γp → ρ⁰p → π⁺π⁻p from Jefferson Lab, relevant to the forthcoming Electron-Ion Collider (EIC). The model checkpoints include both the unconditional generation network and conditional detector unfolding networks trained at various smearing scales (σ = 0.5, 1.0, 2.0).