• View
• Edit
• History
• Print

Research Areas

We develop efficient, adaptive, trustworthy, and secure AI systems for real-world applications.

Research

• Efficient foundation models, state-space models, and long-context memory
• Conversational and personalized AI, including AI companions and dialogue compression
• Federated, distributed, and privacy-preserving learning
• AI security and autonomous cyber defense
• AI for communication networks and energy systems

Featured Project: Naju

Naju is a native discrete state-space model for long-sequence memory. It independently controls memory retention and new information writing, enabling long-term retention and selective overwriting.

• Naju Repository
• Naju Paper