Taoyong Cui (崔涛镛)

🌐 Welcome to my personal web page!

I am currently a Ph.D. student at the AI4LS Laboratory and Multimedia Laboratory (MMLab) in The Chinese University of Hong Kong (CUHK), co-supervised by Prof. Pheng Ann Heng and Prof. Wanli Ouyang. Before joining CUHK, I received my Academic Master’s degree in Biomedical Engineering from Tsinghua University. I served as a research intern at Microsoft Research and Shanghai Artificial Intelligence Laboratory, and have served as a reviewer for Nature Communications, AISTATS, ICML, ICLR, NeurIPS, and AAAI.

My research interests are in AI for science — in particular, machine learning force field, deep generative models, and LLM applications. Feel free to contact me if you are interested in my research.

📢 News and Highlights

• [July 2025] One paper is accepted by Advanced Science.
• [February 2025] One paper is accepted by Nature Communications.
• [November 2024] One paper is accepted by VLDB.
• [April 2024] One paper is accepted by Nature Machine Intelligence.

🔍 Research

Online test-time adaptation for better generalization of interatomic potentials to out-of-distribution data

Taoyong Cui, Chenyu Tang, Dongzhan Zhou, Yuqiang Li, Xingao Gong, Wanli Ouyang, Mao Su, Shufei Zhang

• Description: In this work, we propose an online Test-time Adaptation Interatomic Potential (TAIP) framework to improve the generalization on test data. Specifically, we design a dual-level self-supervised learning approach that leverages global structure and atomic local environment information to align the model with the test data.
• Note: This paper has been accepted at Nature Communications.

MSGNN: Masked Schema based Graph Neural Networks

Hao Liu, Qianwen Yang, Taoyong Cui, Wei Wang

• Description:In this work, we introduce Masked Schema based Graph Neural Networks (MSGNN), which combines schema instances with bi-level self-supervised learning and mask technique to acquire effective context representations. Furthermore, we propose a decomposition-reconstruction schema instance retrieval strategy to ensure efficient instance searching.
• Note: This paper has been accepted at VLDB.

Geometry-enhanced pretraining on interatomic potentials

Taoyong Cui, Chenyu Tang, Mao Su, Shufei Zhang, Yuqiang Li, Lei Bai, Yuhan Dong, Xingao Gong, Wanli Ouyang

• Description: In this work, we propose a geometric structure learning framework that leverages unlabelled configurations to improve the performance of MLIPs. Our framework consists of two stages: first, using classical molecular dynamics simulations to generate unlabelled configurations of the target molecular system; and second, applying geometry-enhanced self-supervised learning techniques, including masking, denoising and contrastive learning, to capture structural information.
• Note: This paper has been accepted at Nature Machine Intelligence.
• Honors: The work was awarded 🏅Outstanding Thesis Award for Young Researchers in World Artificial Intelligence Conference 2024 and 🏅Best Paper Award at the Guangdong-Hong Kong-Macao Greater Bay Area AI for Science PhD Forum 2024.

🗣️ Talks

• World Artificial Intelligence Conference (July 2024)
• Guangdong-Hong Kong-Macao Greater Bay Area AI for Science PhD Forum (October 2024)
• China Academic Forum on Interdisciplinary Innovation for Graduate Students in Materials Science (December 2024)
• World Artificial Intelligence Conference (July 2025)
• Guangdong-Hong Kong-Macao Greater Bay Area AI for Science PhD Forum (September 2025)

📬 Contact

• Email: cty21@tsinghua.org.cn