标签
The paper introduces CoSynFlow, a conformal symplectic neural flow that learns continuous-time solution maps for dissipative Hamiltonian dynamics while preserving the conformal symplectic structure and enabling cross-system prediction without retraining.