About Me
I am a first-year Computer Science Ph.D. student at the University of Wisconsin-Madison.
I received dual bachelor's degrees from a joint program of the University of Electronic Science and Technology of China (UESTC) and the University of Glasgow.
My research focuses on developing smarter and safer large language models, with an emphasis on the post-training and inference stages. A central challenge is ensuring that LLMs not only produce factually correct answers but also follow transparent and reliable reasoning processes, thereby reducing the risk of harmful or misleading outputs.
I aim to enhance both the intelligence and safety of LLMs through improved alignment techniques. I believe that alignment is fundamental to achieving AGI. In fact, the earliest version of ChatGPT emerged from an alignment-driven iteration of GPT-3. Alignment not only ensures instruction-following and safety but may also enhance a model’s general capabilities, making it more reliable and intelligent.