This paper surveys the current state of Efficient Multimodal Learning (EML), a research area focused on optimizing multimodal models for computation, memory, and deployment. The authors introduce a structured taxonomy and provide insights into the vertical synergies between model, algorithm, and system layers, highlighting the importance of cross-layer co-design. The paper also presents a case study on Multimodal Large Language Models (MLLMs) and discusses future directions for EML research.