Learning Disentangled Identifiers for Action-Customized Text-to-Image Generation | Read Paper on Bytez