UROP Proceedings 2022-23

School of Science Department of Mathematics 52 Quantum Information Theory and Error-correcting Codes Supervisor: XIONG, Maosheng / MATH Student: JI, Wenzhao / MATH-IRE WANG, Geqi / MATH-PMA Course: UROP1100, Spring UROP1100, Spring Classical information theory is established in early 20th century by Harry Nyquist, Ralph Hartley and Von Neumann. Information theory is applied in many fields, including telecommunication, computer science and electronic engineering. The key concept of information theory is entropy, describing the uncertainty of a piece of information. In quantum information theory, there is a decisive variation on form of information, from bits to qubits. Thus techniques and conclusion in classical information theory need to be improved to fit in quantum information theory. In our essay, we will give a brief introduction to classical information theory, quantum information theory and some improvements of famous theorems in classical information theory. Research in AI and Machine Learning Supervisor: ZHANG, Tong / MATH Student: TAN, Weile / DSCT Course: UROP1100, Summer The field of natural language processing has been undergoing a revolution in recent years due to the rapid advancement of Large Language Models (LLMs). These highly effective models have shown tremendous potential in addressing a wide range of NLP tasks, including natural language understanding and generation tasks. Some experts even believe that they could pave the way to achieving Artificial General Intelligence (AGI). However, to utilize the LLMs efficiently and effectively, we need to have an in-depth understanding of their capabilities and limitations. In the UROP this summer, I got a chance to study the topic of LLMs from starch. I started by reading a few important research papers, and in this report, the summaries of three articles are included. After I got sufficient basic knowledge, I started to work on finetuning tasks based on the LMFlow - an extensible, convenient, and efficient toolbox for finetuning large machine-learning models. The details about the finetuning tasks are also included after the summaries of the papers.