The field of multimodal learning is rapidly advancing, with a focus on developing more efficient and effective methods for representing and processing multiple forms of data, such as text, images, and audio. A key direction in this area is the development of unified frameworks that can handle multiple modalities and tasks, such as e-commerce search and product understanding. These frameworks aim to learn shared representations that can be used across different tasks and modalities, reducing the need for task-specific training and improving overall performance. Another important area of research is the development of methods for generating and selecting high-quality creative images for advertising, which can enhance the shopping experience for users and increase revenue for advertisers. Noteworthy papers in this area include MM-R1, which introduces a framework for personalized image generation using multimodal large language models, and RefAdGen, which proposes a generation framework that achieves high fidelity through a decoupled design. Additionally, papers such as MOON and MOVER demonstrate the potential of generative multimodal large language models for improving product representation learning and multimodal optimal transport with volume-based embedding regularization.