Tag
This paper proposes a two-stage fine-tuning pipeline combining domain-adaptive fine-tuning and reinforcement learning to generate protein sequences that match a desired amino-acid composition profile while maintaining sequence quality.