I am a 1st year Ph.D. student in Computer Science at King Abdullah University of Science and Technology (KAUST) starting from Fall 2024. I am very fortunate to be advised by Prof. Di Wang in PRADA Lab (Provable Responsible AI and Data Analytics Lab). Before that, I received my Master’s degree in Computational Linguistics from the University of Macau, where I was advised by Derek F. Wong.

My research interests include Explainable AI (by reverse engineering), Bridge of AI learning and human learning, and Synthetic Data. I am committed to developing interpretable AI systems, ensuring that large models align with human values and effectively utilize synthetic data. My goal is to contribute to the creation of AI that is environmentally friendly, human-friendly, fair, and sustainable.

Additionally, my husband and I are running a startup, dedicated to providing controllable and secure AI software. Stay tuned for our upcoming products!

Projects I Want to Share

*Awesome-Sparse-Autoencoder* Collection of Reverse Engineering in Large Model (and Human Brain...)
**about-super-alignment** Feeling confused about super alignment? Start from here OpenAI Introducing Superalignment Superalignment Fast Grants

Featured Research

Arxiv 2024 | What makes your model a low-empathy or warmth person: Exploring the Origins of Personality in LLMs https://arxiv.org/pdf/2410.10863

Shu Yang*, Shenzhe Zhu*, Ruoxuan Bao, Liang Liu, Yu Chen, Lijie Hu, Mengdi Li, Di Wang
keywords: Personality in LLMs, SAE, XAI
Drawing on the theory of social determinism, we investigate how long-term background factors, such as family environment and cultural norms, interact with short-term pressures like external instructions, shaping and influencing LLMs' personality traits.

Arxiv2024 | Understanding Reasoning in Chain-of-Thought from the Hopfieldian View https://arxiv.org/html/2410.03595v1

Lijie Hu*, Liang Liu*, Shu Yang*, Xin Chen, Zhen Tan, Muhammad Asif Ali, Mengdi Li, Di Wang
keywords: XAI, Reasoning, CoT, Representation
We establish a connection between CoT reasoning and key cognitive elements such as stimuli, actions, neural populations, and representation spaces.
We propose the Representation-of-Thought (RoT) framework, which leverages the robustness of low-dimensional representation spaces to enhance the robustness of the reasoning process in CoTs.