Reward Score Matching: Unifying Reward-based Fine-tuning for Flow and Diffusion Models
概要
arXiv:2604.17415v2 Announce Type: replace-cross Abstract: Reward-based fine-tuning steers a pretrained diffusion or flow-based generative model toward higher-reward samples while remaining close to the pretrained model. Although existing methods are derived from different perspectives, we show that…
