Chuancheng Shi

Hi! I’m currently a researcher at the University of Sydney 🏫, where my research focuses on Text-to-Image (T2I) generation models and Large Language Models (LLMs). I’m particularly interested in the internal mechanisms of modern AI systems, and my recent work investigates neuron-level representation, interpretability, and semantic structure within these models.

My broader goal is to deepen our understanding of how generative models encode, organize, and transform information—from fine-grained neuron-level representations to high-level semantic behavior. By studying these internal mechanisms 🧠, I aim to uncover principles that can help us design models that behave more predictably, align more closely with human intent, and remain robust under diverse conditions.

Ultimately, I hope to contribute to the development of AI systems that are more transparent, controllable, and trustworthy 🔍🤝. This includes not only improving interpretability and safety, but also enabling richer and more reliable creative and reasoning capabilities in next-generation generative models.

Thank you for visiting my website 😊!