Shifang Zhao

I'm a master student in the Institute of Information Science at the Beijing Jiaotong University(BJTU), supervised by Prof. Yunchao Wei. Previously, I received the B. Sc. degree in Automation Engineering from Beijing University of Technology(BJUT) in 2023.

Email  /  CV  /  Scholar  /  Wechat  /  Zhihu

Research

I'm interested in Multimodal Large Language Model, Image Generation and AI4Medical. My research is about inspect and understanding the visual information.

This was once a passive interest shaped by the trend. After watching Saining Xie's talk, I clarified a long-term research direction:

Create to inspire creation.

I currently focus on building agentic systems that help people express ideas through existing tools faster and more clearly. Instead of asking users to master fragmented software stacks, my goal is to let them state intent in natural language and have agents plan, orchestrate, and execute the right toolchain.

An approach for enhancing MLLMs in anomaly detection and understanding by unified perception and reasoning.

Bridging the gap between the textual and visual priors for robust zero-shot personalized image generation.

A plug-and-play framework for addressing data imbalance in class incremental learning.

The temporal correlations can be used to detect open-set anomalies in open-ended nature of surveillance.

Experience

About Me More

Thanks to GPT-5.3-Codex