Reward Score Matching: Unifying Reward-based Fine-tuning for Flow and Diffusion Models

ArXi:2604.17415v1 Announce Type: new Reward-based fine-tuning aims to steer a pretrained diffusion or flow-based generative model toward higher-reward samples while remaining close to the pretrained model. Although existing methods are motivated by different perspectives such as Soft RL, GFlowNets, etc., we show that many can be written under a common framework, which we call reward score matching