RePIC: Reinforced Post-Training for Personalizing Multi-Modal Language Models | Read Paper on Bytez