AI4Math - Math Problem Solving

PremiLab-Math is a sub-group of PremiLab at XJTLU, focusing on leveraging LLMs to solve mathematical problems, including Math Word Problems, Geometry Problems and Table Math Word Problems. We also investigate the interpretability of LLMs in math problem solving. Our group page is at PremiLab-Math.

TeLL Icon
MathCheck
(ICLR 2025)
TeLL Icon
TeLL
(AAAI 2025)
TeLL Icon
GNS
(AAAI 2025)

3D Scene

The research group focuses on advancing 3D scene understanding, particularly in zero-shot learning for 3D object recognition. Their recent work explores open-pose 3D zero-shot classification, developing new benchmarks and methods to recognize objects in arbitrary orientations using techniques like angle refinement and diffusion-based knowledge transfer.

Open-Pose 3D Icon
Open-Pose 3D
(Neural Networks 2025)

Digital Human

We explore digital human modeling and understanding, with a focus on reconstructing and analyzing human body shape, pose, and motion. The recent work addresses challenges in creating realistic and controllable digital humans, integrating vision, graphics, and learning-based methods to enable applications such as virtual try-on, performance capture, and human–AI interaction.

KDTalker Icon
KDTalker
(IJCV 2025)

Image Editing

This research group focuses on developing intelligent and controllable image editing techniques by leveraging deep learning and generative models for realistic and high-quality visual content creation.

KDTalker Icon
Point2Pix
(2025)