Large Language Model and AI Agents

🤖 Foundation Models

We aim to develop next-generation foundation models capable of handling long-context and multimodal inputs with high computational efficiency. Additionally, we are focused on creating an efficient AI infrastructure to ensure the scalability of these novel architectures and to facilitate effective computation across large-scale GPU clusters. We actively collaborate with industry partners to develop high-performance AI infrastructures and prototype our algorithms in cutting-edge AI applications.

🧠 Brain-inspired AI Agents

We aim to develop AI agents that can be flexibly configured to execute a broad range of tasks grounded in systems neuroscience, enabling them to learn continuously from their experiences without catastrophic forgetting. These agents will actively perceive diverse sensory signals from their environment and refine their internal states—such as memory, world models, goals, and emotional states. Consequently, they will be able to reason about purposeful actions, both internal and external, allowing them to autonomously pursue complex, long-term objectives.

🌍 Applications

Building on advanced foundation models and AI infrastructure, our AI agents are designed to transform key sectors of the economy and society, unlocking new levels of productivity. For individuals, these agents collaborate with humans and provide services in areas such as personal healthcare and AI assistance. At the societal level, they are poised to revolutionize the robotics industry and drive a new wave of industrial automation. Ultimately, our mission is to enhance the health, well-being, and prosperity of all humanity.

📝 Representative Publications