Energy-Weighted Flow Matching for Offline Reinforcement Learning

Shiyuan Zhang, Weitong Zhang, Quanquan Gu

This paper investigates energy guidance in generative modeling, where the target distribution is defined as $q(\mathbf{x})\propto p(\mathbf{x})\exp (-{\hat{I}}^2\mathcal{E}(\mathbf{x}))$, with…