Quang Minh Nguyen

Logo

qm [dot] nguyen [at] kaist.ac.kr
View my GitHub profile
View my CV

Last update: January 2025

Introduction

My name is Quang Minh. I am a first-year master’s student at KAIST Graduate School of Data Science, where I work with Professor Taegyoon Kim. I previously obtained my bachelor’s degree from KAIST with a double major in Electrical Engineering and Mathematical Sciences.

My eventual goal is to build reasoning machines. This involves investigations of whether and how knowledge can be effectively stored and represented (memory and knowledge representation) and learned knowledge can be composed and applied to novel scenarios (reasoning mechanisms). Under this framework, my subjects of study are large language models (LLMs) and generative models with language modality, utilising natural language as a universal tool to inspect such representations and mechanisms.

Some ideas I am working on or interested in include:

  1. Reasoning under noise and biases
  2. Intrinsic capabilities of LLMs
  3. Inference-time algorithms
  4. Physics of LLMs

Each of these idea can be studied under the context of various reasoning tasks such as natural language understanding (e.g., stance detection and question answering), visual reasoning, computer use, mathematical reasoning, or coding.

I am open to collaboration with the goal of publishing at relevant venues such as NeurIPS, ICLR, ICML, ACL, EMNLP, COLM, etc. and their reasoning workshops. For such requests or just talks about interesting ideas in general, feel free to email me at qm [dot] nguyen [at] kaist.ac.kr and/or grab a coffee together if you pop by Daejeon, South Korea!


News


My Projects

I maintain a page which summarises my projects here.


Miscellaneous

I enjoy learning about history (especially East Asian history) as well as playing musical instruments (I play the piano, guitar, ukulele, and harmonica). We can be friends if you do too :)

I have YouTube channel here where I post music covers, presentations for my course projects, and possibly some vlogs in the future. Consider subscribing!