Beier Luo
a.k.a. Bale · PhD student, College of Computing and Data Science (CCDS), Nanyang Technological University
Hi, I'm Beier. My Chinese name 贝尔 is pronounced like “Bale”, so I also go by Bale. With my surname Luo in front, my name sounds like “Nobel”, which I take as a good omen for my research career.
I am a first-year PhD student at the College of Computing and Data Science (CCDS), Nanyang Technological University, supervised by Dr. Sean Du. Previously, I spent two wonderful years as a graduate student in the group of Dr. Hongxin Wei at Southern University of Science and Technology (SUSTech), where I also earned my Bachelor's degree in Statistics.
news
research
My research focuses on building reliable and trustworthy large language models. I am broadly interested in understanding, modeling, and improving the confidence and uncertainty behavior of modern machine learning systems, particularly under distribution shift and open-ended generation. Current directions:
- confidence calibration and uncertainty estimation for LLMs
- detection and mitigation of hallucinations in LLMs
- theoretical and empirical foundations of reliable machine learning
selected publications
-
Your Pre-trained LLM is Secretly an Unsupervised Confidence CalibratorNeural Information Processing Systems (NeurIPS), 2025
working papers
† denotes equal contribution
-
Unlocking the Pre-Trained Model as a Dual-Alignment Calibrator for Post-Trained LLMs
-
On the Pitfalls of Verbalized Confidence Priors for Calibrating Large Reasoning Models
-
Cross-Entropy Loss Falls Short: What Cross-Entropy Loss Reveals about Large-Language-Model Performance?
education
writing
contact
I'm always excited to explore new research directions and open to collaborating. If you're interested in my research or just want to chat, reach me at LUOB0007@e.ntu.edu.sg.