StyleAR: Customizing Multimodal Autoregressive Model for Style-Aligned Text-to-Image Generation | Read Paper on Bytez