Kang Li

SUGFW+ achieves state-of-the-art performance in medical image segmentation by seamlessly integrating sample selection and model training through uncertainty-guided feature weighting.

GeAR achieves unprecedented fidelity in reconstructing 3D scenes from classical paintings, outperforming traditional methods in both geometry and appearance.

LLMs can now navigate and reason over complex relational environments with significantly improved accuracy and efficiency thanks to a new reinforcement learning framework that treats graph learning as an agentic process.

Evolving interpretable composite features via Genetic Programming beats black-box deep learning at music tagging, revealing synergistic interactions and transformations that boost performance.

LLMs can now reason more effectively on multimodal graphs thanks to a new framework that aligns visual and textual features using graph structure and intelligently routes the most relevant modalities to the LLM.