Jianxing Zhao
Hi there! This is Jianxing Zhao, also commonly known as Robin. I'm a first-year CS PhD student at UC Davis.
I'm grateful to be a member of VIA Lab and advised by Prof. Dongyu Liu.
Before my PhD journey, I earned my B.S. in Computer Science and a minor in Mathematics from UCLA in 2025.
Email /
Github /
LinkedIn
|
|
Research
My research interests broadly concern visualization-empowered AI systems, a direction I began to develop after starting my PhD.
I hope to build visual analytics systems that integrate AI techniques with human intelligence,
so that AI can use humans' input as a guardrail and produce understandable, acceptable output for humans in return.
More recently, my research directions have focused on tabular synthetic data and chart understanding using VLMs.
I have decent literacy in symbolic AI, causality, and NN verification as well. Feel free to talk to me about them. 🤓
|
|
|
Embedding Space Selection for Detecting Memorization and
Fingerprinting in Generative Models
Jack He,
Jianxing Zhao,
Andrew Bai,
Cho-Jui Hsieh
arXiv Preprint, 2024
We study memorization in generative models through layer-wise embeddings of Vision Transformers,
revealing that different layers capture distinct levels of memorized information—from low-level visual patterns to high-level semantics.
We show that each architecture exhibits a unique, inherent memorization profile across layers, and leverage this insight to propose a novel
fingerprinting method for model attribution. Our approach improves deepfake and malicious content identification accuracy by 30% over existing baselines.
|
|
(more vis related works are coming up... 🫣)
|
|
Writings
|
Reports, Manuscripts:
|
|
Books
|
My perspectives of CS reasarch are heavily influenced by the following books:
|
|
