My Bio:

Shu Yang is a second-year Ph.D. student in Computer Science at King Abdullah University of Science and Technology (KAUST), advised by Prof. Di Wang in the PRADA Lab (Provable Responsible AI and Data Analytics Lab). She received her M.Sc. in Computational Linguistics from the University of Macau, where she was advised by Prof. Derek F. Wong. Shu has also been a visiting scholar at the University of Edinburgh and Tsinghua University. Her research interests lie in responsible AI, large language models (LLMs), and computational linguistics. She has published more than ten papers in top-tier conferences and journals, including ACL, EMNLP, NeurIPS, and CL. In addition, she is one of the organizers of the First PersonaLLM Workshop at NeurIPS 2025.

My research is centered on advancing the trustworthiness and alignment of Large Language Models. I tackle key challenges in AI safety and explainability, with a particular emphasis on two critical areas. Firstly, I investigate emergent deceptive behaviors such as specification gaming and reward hacking, aiming to understand their underlying mechanisms. Secondly, I leverage these insights to develop LLMs that are not only more aligned with complex human intent but are also designed for more effective monitoring and supervision, ensuring that their expressed reasoning accurately reflects their internal states.

Featured Research

CoLM 2024 | Model Autophagy Analysis to Explicate Self-consumption within Human-AI Interactions

https://openreview.net/pdf/259f10aec40499dd63119bf34937e1ec89729faa.pdf

Shu Yang*, Muhammad Asif Ali, Lu Yu, Lijie Hu, and Di Wang
keywords: XAI, Self-consumption, Human-AI Interactions
A progressive prevalence of model-generated synthetic information over time within training datasets compared to human-generated information.
The discernible tendency of large models, when acting as information transmitters across multiple iterations, to selectively modify or prioritize specific contents.
The potential for a reduction in the diversity of socially or human-generated information, leading to bottlenecks in the performance enhancement of large models and confining them to local optima.