Unleashing the Power of Granular-ball for LLM-basedMulti-Modality Recommendation
About
My current research focuses on multimodal LLM-based recommendation. I am familiar with related technologies such as multimodal learning, LLM, and granular-ball computing.
Publications
Competitions
-
10.2025
“Huawei Cup” The 22nd China Graduate Mathematical Modeling Competition
-
4.2025
The 15th MathorCup Mathematical Application Challenge, 2025
-
11.2025
The 6th MathorCup Big Data Competition, 2025
-
6.2025
The 7th Zhongqing Cup National College Student Mathematical Modeling Competition, 2025
Projects
For the requirements of injury assessment in complex trauma scenarios, I participated in the development of an LLM-based multimodal intelligent diagnosis system, enabling collaborative analysis and decision support across physiological signals, medical imaging, and semantic information. Built a multimodal injury assessment model using PyTorch, where LLMs were leveraged to generate semantic descriptions of physiological data, and a Transformer-based architecture was adopted to fuse textual representations with physiological signal features. For the image branch, ViT was employed to extract features from ultrasound images, while Fourier transform was further incorporated to enhance frequency-domain information representation, improving the model’s feature modeling capability in complex trauma scenarios.