About

I am currently a postdoctoral scholar at the University of Pennsylvania, advised by Prof. Konrad Kording. Previously, I received my Ph.D. from the Chinese University of Hong Kong, where I was a member of Multimedia Lab (MMLab) and was supervised by Prof. Xiaogang Wang and Prof. Hongsheng Li. I received my B.Eng degree from the College of Intelligence and Computing, Tianjin University (TJU) in 2017. At that time, I also minored in Finance.

I study what intelligent systems optimize, what they ought to optimize given their tasks and constraints, and when their objectives should change.

Three questions guide this work, for artificial systems and humans alike. Where do objectives come from, and how do they shape learning and behavior? What would an ideal system achieve, what is achievable under realistic constraints, and how does the actual system compare? Which of these gaps reveal properties of the system, and which are artifacts of our assumptions or evaluation methods?

I build explanations from decision theory, optimization, and the normative models of perception and action used in neuroscience and cognitive science, and test them through interventions and predictions under new conditions. My earlier work developed objectives and optimization methods for visual recognition and unsupervised segmentation, and inferred objectives from human sensorimotor behavior, such as where people look and how they decide. My current work brings this approach to general-purpose learning systems, asking whether explanations in terms of internal goals or world models predict behavior under intervention, or merely account for it after the fact.

Ultimately, I aim to understand and design the objectives of intelligent systems that adapt and evolve on their own. Central to this is evidence-driven objective revision: recognizing when goals no longer serve the task, and revising them using evidence the system cannot manipulate, while remaining open to external correction.

Publications

Falcon: Fast Proximal Linearization of Normalized Cuts for Unsupervised Image Segmentation
Xiao Zhang, Xiangyu Han, Xiwen Lai, Yao Sun, Pei Zhang, Xia Liu, Konrad Kording
2026 International Conference on Learning Representations (ICLR)

When Goldfish Start Divining: Algorithmic Divination and the Paradox of Control in a Computer Vision–Driven Generative BioArt Installation
Xia Liu, Xiao Zhang
2026 International Symposium on Electronic Art (ISEA) — Academic Track
Related artwork: For Whose Will, This Coronation — CVPR 2026 Art Gallery Shortlisted

Who Streams Whom? The Loop of Existence in Algorithmic Vision
Xia Liu, Xiao Zhang
2026 International Symposium on Electronic Art (ISEA) — Academic Track

Rhythm Gate: Invisible Conversations in the Elevator — Echoes of Material, Behavior, Memory and Transformation
Xia Liu, Xiao Zhang
2025 ACM International Conference on Multimedia (ACM MM)

Refining Pseudo Labels with Clustering Consensus over Generations for Unsupervised Object Re-Identification
Xiao Zhang, Yixiao Ge, Yu Qiao, Hongsheng Li
2021 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

RBF-Softmax: Learning Deep Representative Prototypes with Radial Basis Function Softmax
Xiao Zhang, Rui Zhao, Yu Qiao, Hongsheng Li
2020 European Conference on Computer Vision (ECCV)

AdaCos: Adaptively Scaling Cosine Logits for Effectively Learning Deep Face Representations
Xiao Zhang, Rui Zhao, Yu Qiao, Xiaogang Wang, Hongsheng Li
2019 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Oral

P2SGrad: Refined Gradients for Optimizing Deep Face Models
Xiao Zhang, Rui Zhao, Junjie Yan, Mengya Gao, Yu Qiao, Xiaogang Wang, Hongsheng Li
2019 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

Range Loss for Deep Face Recognition with Long-Tailed Training Data
Xiao Zhang, Zhiyuan Fang, Yandong Wen, Zhifeng Li, Yu Qiao
2017 IEEE International Conference on Computer Vision (ICCV)

Research Experience

• Research Intern at NVIDIA. (Jul. 2021 to Jul. 2022)
  Worked on self-supervised representation learning with Dr. Charles Cheung

• Computer Vision Research Assistant at the Chinese University of Hong Kong. (Jul. 2017 to Jul. 2019)
  Worked with Prof. Hongsheng Li and Dr. Junjie Yan

• Visiting Student at MMLab-SIAT, Chinese Academy of Sciences. (Jul. 2016 to Jul. 2017)
  Advisor: Prof. Yu Qiao

Entrepreneurship

• Co-Founder, Stealth Startup. San Francisco, USA (2024–2025)
  Developed intelligent exercise equipment that combines human motion sensing and motor control to provide assistance responsive to users’ movements.

• Co-Founder & CTO, Stealth Startup. Shenzhen, China (2020–2022)
  Led the development and deployment of multi-camera systems for real-time 3D human motion perception in fitness spaces.

Services

I have served as a reviewer for the following conferences and journals:

Computer Vision

• ICCV: 2025, 2023, 2021, 2019
• CVPR: 2023, 2022, 2021, 2019, 2018
• ECCV: 2022, 2020

Machine Learning and AI

• ICML: 2026
• ICLR: 2026, 2025, 2024
• NeurIPS: 2024, 2023
• AISTATS: 2025

Robotics

• ICRA: 2019

Journals

• IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI)
• International Journal of Computer Vision (IJCV)
• Neurocomputing
• Computer Vision and Image Understanding (CVIU)

Teaching

Teaching assistant for the following courses at CUHK:

• CUHK ENGG 5202, Pattern Recognition, Spring 2022
• CUHK ELEG 5760, Machine Learning for Signal Processing Applications, Fall 2019
• CUHK ENGG 2030, Signals and Systems, Fall 2022, Fall 2021, Spring 2021, Fall 2020, Spring 2020, Fall 2019