🎓 I recently graduated with Distinction from the MSc in ACSE (Applied Computational Science and Engineering) at Imperial College London. Before this, I earned a BSc in Applied Mathematics from the TYUT (Taiyuan University of Technology), graduating with a GPA of 4.5/5.0 (2/40). During my BSc and MSc, I focused primarily on physics-inspired and scientific machine learning, with particular interests in operator learning, Koopman theory, and representation learning for dynamical systems.

🔍 Currently, I have been increasingly interested in how latent dynamics, representation learning, reinforcement learning, and world models can come together to enable machines to predict, reason about, and act in complex environments.

Recent Updates

• 🎉 [ICLR 2026 Oral] Information Shapes Koopman Representation has been accepted as an Oral at ICLR 2026. In this work, we study Koopman representation learning from an information-theoretic perspective, formalizing the trade-off between simplicity and expressiveness and translating it into an optimizable objective.

• 📝 A recent manuscript on integrating diffusion policy into a safe reinforcement learning framework is currently under review at ICML 2026.

• 📊 I am currently leading a benchmark study on Koopman-inspired models for data assimilation tasks.

• 🚀 I am also actively working on addressing the scaling bottleneck of RL world models with diffusion policy.

Open to Work

💼 I am especially interested in research-oriented industry positions at the intersection of world models, self-supervised representation learning, and machine learning for complex dynamical systems, particularly in building models that can predict, reason about, and plan within high-dimensional environments. I am also open to aligned PhD opportunities.

📮 Feel free to contact me at: wenxuan.yuan@qq.com