Researchers introduce CoRe, a co-evolving reward framework for mitigating latent reward hacking in video diffusion models. CoRe refits the reward model on the generator's current samples while anchoring it to real-video preferences, improving generation quality and avoiding quality collapse.