Our lab focuses on advancing machine learning, particularly in the fields of Multimodal Learning, Self-supervised Learning, and Social Artificial Intelligence. However, our interests are not limited to these areas. We are open to a wide range of machine learning topics that enable machines to achieve human-level intelligence or beyond. Please refer to the publications for further information.
We develop machines capable of understanding and integrating information from multiple modalities such as visual, language, and audio cues. By leveraging the complementary nature of these diverse modalities, we address complex challenges in our multifaceted real-world environments.
Visual-Language learning / Audio-Visual learning / Multimodal foundation models
We explore self-supervised learning paradigms that enable machines to learn meaningful feature representations from weakly-labeled or unlabeled data. By reducing the reliance on labeled data and leveraging readily available large-scale unlabeled data, we develop more efficient and scalable learning systems.
Weakly-labeled data / Unlabeled data / Representation learning
We develop machines with social intelligence, enabling them to recognize, interpret, and appropriately respond to human social behaviors. By modeling and simulating social dynamics, we aim to create socially intelligent machines that can interact seamlessly with humans across diverse social contexts.
Social perception / Social reasoning / Social AI agents