Stephen Cheng
I am a first-year CS PhD student at the University of Maryland co-advised by Prof. Sarah Wiegreffe and Prof. Dinesh Manocha. My research interests lie broadly in understanding LLM behavior and post-training methodologies, with the goal of applying these findings to improve model performance and alignment. Currently, I am work on mechanistic interpretability methods for studying reasoning models.
News
- August 2026: 1 first author paper accepted to EMNLP 2026!
- March 2026: Now co-advised by Prof. Sarah Wiegreffe!
- September 2025: Began PhD in Computer Science at UMD, advised by Prof. Dinesh Manocha!
- June 2025: Graduated magna cum laude from Northwestern University with a BS in Electrical Engineering and a MS in Computer Science!
- September 2024: Researched various deep learning topics, advised by Prof. Han Liu!
Selected Publications
- Stephen Cheng, Sarah Wiegreffe*, Dinesh Manocha*. What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal. (EMNLP 2026) arXiv Code
- Weijian Li*, Stephen Cheng*, Lining Mao*, Jigyasa Kumari, Alex Pyo, Mehak Kawatra, Jialong Li, Jiayi Wang, Ammar Gilani, Jingya Xun, Jerry Yao-Chieh Hu, Han Liu. A benchmark study for limit order book (lob) models and time series forecasting models on lob data. (preprint 2024) arXiv