🗿 Short Bio

Hi! My name is Liangze Jiang (🇨🇳: 姜良泽). I’m a first-year PhD student in computer science at Swiss Federal Institute of Technology Lausanne (EPFL🇨🇭), where I also did my Master’s and worked with Prof. Amir Zamir and Prof. Martin Jaggi. Previously, I worked as a student researcher at Google Research with Dr. Claudiu Musat and Dr. Pedro Gonnet. I got my Bachelor degree from University of Electronic Science and Technology of China (UESTC).

I had fun research experiences on robustness and generalization under distribution shifts, graph neural networks, federated learning, etc. I am now broadly interested in foundation models and transfer learning, as well as understanding their underlying dynamics & generalization empirically and theoretically. I am also excited about the advancements of AI in natural sciences.

📃 Publications

(* denotes equal contribution)

Unraveling the Key Components of OOD Generalization via Diversification
H. L. Benoit*, L. Jiang*, A. Atanov*, O. F. Kar, M. Rigotti, A. Zamir
ICLR 2024
We explore the key factors (e.g., data distribution and its interaction with models) of multi-hypothesis OOD generalization methods.
Paper

Test-time Robust Personalization for Federated Learning
L. Jiang*, T. Lin*
ICLR 2023
We identify the pitfalls of PFL and propose an adaptive and robust test-time ensemble method under various distribution shifts.
Paper / Code

Understanding and Manipulating Agreement between Neural Networks
L. Jiang
Master's Thesis 2023
What does a high agreement between neural networks imply? How can agreement be utilized to tackle real-world problems?
Poster

TF-GNN: Graph Neural Networks in TensorFlow
O. Ferludin, A. Eigenwillig, and 25 others, including L. Jiang
arXiv 2023
A library to build Graph Neural Networks on the TensorFlow.
Paper / Code

🏆 Honors & Awards

• EPFL EDIC PhD Fellowship, 2023-2024
• Outstanding Graduate Award, 2020, Ranking: 6/244
• Undergraduate Academic Excellence Scholarship, 2016-2020, Top 10%
• China National Scholarship, 2016-2017, 2018-2019, Top 0.2%
    - The highest honor level scholarship across China for academic achievements.
• Meritorious Winner in MCM (International Mathematical Contest in Modeling), 2019, Top 7%
• Sekorm Scholarship, 2018, Top 3%
• Chinese Physics Olympiad, second prize in Shandong province

🔓 Open-Source Contribution

• Transformer Multi-head Attention in Tensorflow GNN