Wenqi Cai

University of Toyama.
KiZuLab.
3190 Gofuku, Toyama 930-8555, Japan.
wenqicai297@gmail.com

I am currently an M.Sc. student at University of Toyama, with a research focus on computer vision and deep learning.

I come from Nanchang, Jiangxi, China. Now I’m living in Toyama, Japan, and actively learning Japanese. I’d be happy to make new friends here!

Outside of my studies, I’m passionate about gaming, especially challenging genres like soul-like, rogue-like, and star-hunting games, while also enjoying more relaxing ones such as Terraria, Dyson Sphere Program, and Slay the Spire. Beyond gaming and research, I also love spending time outdoors with friends—whether it’s walking, hiking, or exploring nature. I also enjoy listening to music in my free time.

Research Interests

Multimodal Learning

Learning to extract compact, task-relevant information from rich and often redundant multimodal signals.

Representation Learning

Learning general-purpose representations from the intrinsic structure of data that transfer effectively across diverse downstream tasks.

3D/4D Vision

Understanding how the physical world is structured in space and evolves over time, from dynamic scenes to human motion.

News

View all

I will present our work as a poster at CVPR 2026 in Denver, Colorado.

  • Wenqi Cai, Yawen Zou, Guang Li, Chunzhi Gu, Chao Zhang, EVLF: Early Vision-Language Fusion for Generative Dataset Distillation [arXiv] [GitHub]

My paper has been accepted to CVPR 2026.

  • Wenqi Cai, Yawen Zou, Guang Li, Chunzhi Gu, Chao Zhang, EVLF: Early Vision-Language Fusion for Generative Dataset Distillation [arXiv] [GitHub]

My paper has been officially published in the proceedings of QCAV 2025.

  • Wenqi Cai, Jun Yu, Yawen Zou, Chao Zhang, Improving the explainability of neural network by downstream post-optimization

I will give a presentation at ISID 2025.

  • Wenqi Cai, Chao Zhang, Improving the Explainability of Neural Network with Swarm Optimization