👤 About Me

🔬 Research

The long-term research goal of our team is to building general AI agents in both digital and physical worlds, contributing to the advancement of artificial general intelligence (AGI). Our core approach centers on neuro-symbolic learning, which bridges data-driven learning with symbolic-driven search and reasoning, often regarded as the hallmark of third-generation AI. The neural component provides grounding in perception and physical interaction, while the symbolic component augments reasoning and planning.
Recently, our research has primarily focused on building agents capable of thinking, reasoning, planning, and acting in virtual or physical worlds, such as game agents that master complex video games, embodied agents with long-horizon planning, or science agents that assist human researchers.
The core modules and representative applications in our research are illustrated below:

📰 News