Data-Centric Intelligent Systems
Research on data-driven intelligent systems, reliability, and security.
We focus on Data-Centric Intelligent Systems, studying how data drives the capabilities and behaviors of intelligent systems while ensuring their reliability and security in real-world scenarios.
Data Construction
We study methods for building high-quality data, including the synthesis, representation, filtering, and evaluation of heterogeneous data, with the goal of improving the capability and generalization of LLMs and agentic systems.
Trustworthy Systems
We study system robustness and security mechanisms, focusing on model security, adversarial attacks and defenses, user behavior modeling, and anomaly detection to improve system reliability in real environments.
Scenario-Driven Applications
We build secure and reliable intelligent system solutions for complex scenarios such as smart healthcare, cybersecurity, and social media.