The field of robotic manipulation is moving towards leveraging multimodal learning to improve task performance and generalization. Recent research has focused on integrating vision, language, and other modalities such as audio and muscle signals to enable robots to learn from human demonstrations and adapt to new tasks. This has led to the development of more robust and generalizable models that can perform complex manipulation tasks in real-world environments. Notably, advances in vision-language-action models have enabled robots to perform long-horizon and dexterous manipulation skills, such as cleaning a kitchen or bedroom, in entirely new homes. Additionally, techniques such as co-training and hybrid multi-modal examples have been shown to be essential for effective generalization. Some notable papers include: Chain-of-Modality, which introduces a prompting strategy that enables Vision Language Models to reason about multimodal human demonstration data, resulting in a threefold improvement in accuracy for extracting task plans and control parameters. Text-to-Decision Agent, which proposes a simple and scalable framework that supervises generalist policy learning with natural language, facilitating high-capacity zero-shot generalization. ManipDreamer, which introduces an advanced world model based on the action tree and visual guidance, significantly boosting the instruction-following ability and visual quality of robotic manipulation video synthesis.