A first-year undergraduate in Computer Science at China University of Geosciences (Beijing), working on information theory of learning systems, mechanistic interpretability, and AI safety.

Research Interests

• Theoretical foundations of model collapse and generative degradation
• Observability and recoverability of features in sparse autoencoders
• Safety and failure mode analysis for Vision-Language-Action models

News

• Apr 2026  ·  Papers in preparation.
• Apr 2026  ·  Leading VLA safety project at the 9th “Cradle Cup” Innovation Competition, CUGB.
• Sep 2025  ·  Started undergraduate studies at CUGB.

About

I play piano, read broadly, and spend too much time thinking about open questions — in research and in life. I’ve never been able to walk past one. Research is what happens when you stop pretending you can.

I’m always happy to talk — about research, ideas, or anything interesting. Feel free to reach out.

Contact

lancer20060105@gmail.com  ·  GitHub  ·  Beijing