The article introduces POLAR, a memory-augmented framework for personalized embodied agents that enhances their ability to assist users by leveraging accumulated interactions over time. This innovation is expected to significantly improve task execution and personalization in multimodal large language model applications.