Bio
Welcome to my homepage! I am Tianxiang Zhao, an Assistant Professor in the AI thrust at the Hong Kong University of Science and Technology (Guangzhou), starting in February 2026. Prior to joining HKUST(GZ), I worked as an LLM Researcher at Huawei through the TopMinds (天才少年) program in 2025. I graduated from the Class of Gifted Young at the University of Science and Technology of China (USTC) in 2017, and in 2019, I left my Master’s program at USTC to begin my Ph.D. journey. I received my Ph.D. in 2024 from the Pennsylvania State University, where I was co-advised by Prof. Xiang Zhang and Prof. Suhang Wang.
Research
My research focuses on data-centric machine learning. Recently, my interests include the following directions:
- RL in PostTraining:
- Stable and efficient on-policy distillation.
- Agentic RL
- Efficiency of Vision-Language models in reasoning, optimization of agentic systems
- Trustworthy in Machine Learning:
- Learning from imperfect supervison (e.g., noisy or unlabeled data);
- Model debias
- Explanablity approaches for understanding model behaviors and learning dynamics
I am broadly interested in topics related to machine learning and cross-disciplinary applications. Feel free to email me if you’d like to connect.
I am looking for multiple PhD students to start in Fall 2026, and also have openings for Interns/Research Assistants at the moment. If you are interested, please send your resume/CV to me.
Education
- 2019 - 2024, Ph.D, Information Sciences and Technology, The Pennsylvania State University
- 2017 - 2019, Master student, University of Science and Technology of China
- 2013 - 2017, Bachelor in Computer Science, Class of The Gifted Young, University of Science and Technology of China
Recent News
- 05/2024: One paper, “Multi-source Unsupervised Domain Adaptation on Graphs with Transferability Modeling”, is accepted by KDD-2024
- 02/2024: One paper, “Towards Inductive and Efficient Explanations for Graph Neural Networks”, is accepted by TPAMI
- 01/2024：One paper, “Disambiguated Node Classification with Graph Neural Networks”, is accepted by WebConf-2024
- 10/2023：Two papers, “Interpretable Imitation Learning with Dynamic Causal Relations” and “Distribution Consistency based Self-Training for Graph Neural Networks with Sparse Labels” are accepted by WSDM-2024
- 08/2023, One paper, “T-SaS: Toward Shift-aware Dynamic Adaptation for Streaming Data”, is accepted by CIKM-2023
- 07/2023, One paper, “Faithful and Consistent Graph Neural Network Explanations with Rationale Alignment”, is accepted by ACM TIST
- 05/2023: One paper, “Skill Disentanglement for Imitation Learning from Suboptimal Demonstrations”, is accepted by KDD-2023
- 05/2023: Glad to join Microsoft Research as a research intern for the summer of 2023
- 11/2022: One paper, “TopoImb: Toward Topology-level Imbalance in Learning from Graphs”, is accepted by LOG-2022
- 10/2022: One paper, “Towards Faithful and Consistent Explanations for Graph Neural Networks”, is accepted by WSDM-2023
- 05/2022: Glad to return to NEC Labs America as a research intern for the summer of 2022
- 01/2022: One paper, “Exploring Edge Disentanglement for Node Classification”, is accepted by WebConf-2022
- …