Taoyong Cui 崔涛镛

AI FOR SCIENCE & PHYSICS

Taoyong Cui 崔涛镛

Scientific World Models

Ph.D. student, The Chinese University of Hong Kong
Visiting Ph.D. researcher, Stanford University

I study AI for Physics, World Models, and Recursive Self-Improvement (RSI) to understand physical and scientific systems and advance scientific discovery.

Google Scholar GitHub Email
Portrait of Taoyong Cui
AI for Physics · World Models · RSI

ABOUT

Research &
Background.

I work at the intersection of machine learning, science, and physics. My research explores AI for Physics, World Models, and Recursive Self-Improvement (RSI), with applications spanning atomic and macroscopic physical systems, scientific discovery, and embodied environments.

At CUHK, I work with the AI4LS Laboratory and MMLab, co-supervised by Prof. Pheng Ann Heng and Prof. Wanli Ouyang. I also collaborate with Prof. Le Cong at Stanford.

Background Master's in Biomedical Engineering, Tsinghua University; previously at Microsoft Research Asia and Shanghai AI Laboratory.

LATEST

News

  1. JEPA-Anything released. Featured by QbitAI (量子位) ↗.

  2. Nature Communications paper accepted.

  3. Scientific Data paper accepted.

  4. Current Opinion in Structural Biology paper accepted.

Earlier updates

  1. Advanced Science paper accepted.

  2. Nature Communications paper accepted.

  3. VLDB paper accepted.

  4. Nature Machine Intelligence paper accepted.

Research Directions

Predictive models across physical and scientific systems.

AI for Physics

Understanding the physical world, from atomic-scale processes to macroscopic systems.

World Models

Predictive models across scientific and physical systems, including embodied environments.

RSI

Recursive self-improvement through iterative learning, interaction, evaluation, and feedback.

Selected Publications

Scientific world models, generalization, and reliable atomistic simulation.

Overview figure for JEPA-Anything: Learning Predictive Models across Different Worlds

arXiv · 2026 Preprint

JEPA-Anything: Learning Predictive Models across Different Worlds

Taoyong Cui, Zhongyao Wang, Xinyue Xu, Weiyang Liu, Zhaochen Yu, Yuying Zhang, Qiang Gao, Mengyue Yang, Wanli Ouyang, Pheng Ann Heng, Yingcheng Wu, Zhenfei Yin, Ling Yang

A shared predictive framework for world models across vision, biology, clinical trajectories, control, molecular dynamics, physical fields, and weather.

Read paper Code
Overview figure for Evidential deep learning for interatomic potentials

Nature Communications · 2026

Evidential deep learning for interatomic potentials

Han Xu†, Taoyong Cui†, Chenyu Tang†, Jinzhe Ma, Dongzhan Zhou, Yuqiang Li, Xiang Gao, Xingao Gong, Wanli Ouyang, Shufei Zhang, Mao Su

Efficient uncertainty quantification for interatomic potentials.

Read paper
Overview figure for Online test-time adaptation for better generalization of interatomic potentials to out-of-distribution data

Nature Communications · 2025

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

Test-time adaptation for out-of-distribution atomic structures.

Read paper
Overview figure for Geometry-enhanced pretraining on interatomic potentials

Nature Machine Intelligence · 2024

Geometry-enhanced pretraining on interatomic potentials

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

Geometry-aware pretraining for interatomic potentials.

WAIC 2024 Outstanding Thesis Award · GBA AI for Science Ph.D. Forum Best Paper

Read paper

† Equal contribution. Figures link to the original papers.

All publications on Google Scholar

Talks & Workshops

Research exchange and community building.

Workshop Organization

Scientific world models: from predictive representations to interventions and experimental feedback

ICLR 2027 WORKSHOP PROPOSAL · ORGANIZER

Scientific World Models:
From Prediction to Intervention and Discovery (Proposal under review)

Connecting representation learning, scientific machine learning, and embodied intelligence through models of scientific systems.

Workshop website

Selected Talks

Global Artificial Intelligence Technology Conference

Guangdong–Hong Kong–Macao Greater Bay Area AI for Science Ph.D. Forum

World Artificial Intelligence Conference

China Academic Forum on Interdisciplinary Innovation for Graduate Students in Materials Science

COMMUNITY

Academic Service

Reviewer for journals and conferences in machine learning and scientific AI.

  • Nature Communications
  • AISTATS
  • ICML
  • ICLR
  • NeurIPS
  • AAAI

CONTACT

Let's talk science.

AI for Science collaborations and research conversations are welcome.

cty21@tsinghua.org.cn