Representative work spans large model prompting & selection, evolutionary feature learning, biomedical classification & pathway modeling, and edge/on-device inference. Publications appear in venues such as ICML, TKDE, TCBB, CSUR, KBS; several results received media coverage.
Research Experience
Research topics include self-evolution of AI models and agents under resource constraints, and evolutionary machine intelligence and its applications. Specific projects include:
- Reinforced Diverse Example Selector (RDES, ICML'25)
Research focuses on foundation models and collaborative intelligence between large and compact models, aiming to make powerful capabilities deployable under real-world constraints (latency, memory, privacy). Technically, combines evolutionary optimization, multi-task feature selection, and reinforcement learning with representation compression to enhance generalization across heterogeneous, high-dimensional biomedical and textual datasets.