Yuxin Wang

Yuxin Wang

MSE Data Science @ JHU | SciML Research

About Me

I am a Master’s student in Data Science (MSE) at the Department of Applied Mathematics and Statistics, Johns Hopkins University (Jan 2026 – Present). [Resume]

I am a curious person drawn to ideas that go beyond understanding and solving a single problem. I enjoy finding stable and reusable patterns across models and tasks, understanding why they emerge, and transferring them to new settings. I am also interested in developing analysis tools that can be applied efficiently across different problems. I believe that seemingly complex model behaviors may often share simpler underlying causes, and that uncovering these causes can help us identify, understand, and address a broader range of failures.

My research interests include AI for Science, model diagnostics, transfer learning, and training dynamics. I am also beginning to explore theoretical perspectives on learning dynamics and generalization.

My research experience spans several related settings:

  • With Prof. Yaoqing Yang and Pu Ren, I work on Scientific Machine Learning (SciML) and automated research, studying loss landscapes, model diagnostics, and transfer learning across models, tasks, and capacities.
  • With Xing Han, we identified failure modes in language-model reasoning and used counterfactual self-evolution to improve evidence-grounded reasoning.
  • In my research engineering work with William Walden and Prof. Benjamin Van Durme, I build multi-agent infrastructure for scientific claim verification, including evaluation pipelines that diagnose evidence and reasoning failures in long-horizon research.

Previously, I received my B.E. in Data Science and Big Data Technology from Tongji University (2019–2023). Then I worked as a Software Development Engineer at Beijing Dajia Internet Information Technology (Kuaishou) (2023–2025), where I built performance analysis tools and maintained a JS container for large-scale mobile systems.

News

Publications

Counterfactual Self-Evolving Agents for Evidence-Grounded Reasoning
Xing Han, Yuxin Wang, Chen Chen, Wei Dai, Gautham Krishna Gudur, Shijun Li, Hsing-Huan Chung, Gregory D. Hager, Joydeep Ghosh, Paul Pu Liang, Suchi Saria
arXiv preprint, 2026
[Paper]

Unveiling Multi-regime Patterns in SciML: Distinct Failure Modes and Regime-specific Optimization
Yuxin Wang*, Yuanzhe Hu*, Xiaokun Zhong*, Xiaopeng Wang*, Haiquan Lu, Tianyu Pang, Michael W. Mahoney, Yujun Yan, Pu Ren, Yaoqing Yang
International Conference on Machine Learning (ICML), 2026
[Paper] [Code]

Contact

ruangyx@gmail.com / ywan1150@jh.edu

Other

I enjoy reading various kinds of novels! My happiness comes from building imagination within reality and seeking reality within those fantasies. Actually, I’m trying to make my life an adventure novel and my future research a detective novel.