Hello World :)

Huy Nghiem
Huy Nghiem
PhD Student in AI, NLP and Computational Sociology

I am a third-year PhD student in Computer Science at the University of Maryland, College Park under the advise of Professor Hal Daumé III. My research focuses on Safety, Robustness and Trustworthiness in AI, with a focus on application of NLP systems to solve social problems. My recent and upcoming works investigate and enhancing the robustness of Large Language Models (LLMs) in tasks that require complex reasoning. Before UMD, I earned my Master's degree at the University of Southern California and conducted research in social media under Dr. Fred Morstatter and Dr. Emilio Ferrara. I also worked at the Children's Data Network as a Data Scientist, investigating and enhancing public policy in the child welfare system.

In my free time, I enjoy casual hiking along scenic routes 🌲, exploring quaint towns ⛰️ and electrifying cities 🗽, sipping matcha 🍵 while reading manga or scientific articles...

If you are interested in collaborating in causal reasoning or using AI agents for social good, please feel free to reach out.

News & Updates

📝 Blogs

April 06, 2026
Why Some Models Shatter and Others Don't: The Geometry of Emergent Misalignment Across Architectures
Authors: Huy Nghiem
AI safetyalignmentemergent misalignment

All blog posts →

Selected Publications

Bias in the Tails: How Name-conditioned Evaluative Framing in Resume Summaries Destabilizes LLM-based Hiring
arXiv 2026
Bias in the Tails: How Name-conditioned Evaluative Framing in Resume Summaries Destabilizes LLM-based Hiring
Huy Nghiem, Phuong-Anh Nguyen-Le, Sy-Tuyen Ho, Hal Daumé III
Subtle race-gender evaluative framing in LLM-generated resume summaries undermines hiring fairness.
paper
SMARTER: A Data-efficient Framework to Improve Toxicity Detection with Explanation via Self-augmenting Large Language Models
ACL 2026
SMARTER: A Data-efficient Framework to Improve Toxicity Detection with Explanation via Self-augmenting Large Language Models
Huy Nghiem, Advik Sachdeva, Hal Daumé III
Data-efficient toxicity detection via LLM-based synthetic explanations.
paper
'Rich Dad, Poor Lad': How do Large Language Models Contextualize Socioeconomic Factors in College Admission ?
EMNLP 2025
'Rich Dad, Poor Lad': How do Large Language Models Contextualize Socioeconomic Factors in College Admission ?
Huy Nghiem, Phuong Anh Nguyen-Le, John Prindle, Rachel Rudinger, Hal Daumé III
How LLMs contextualize socioeconomic factors in college admissions.
papercode
You Gotta be a Doctor, Lin
EMNLP 2024
You Gotta be a Doctor, Lin
Huy Nghiem, John Prindle, Jieyu Zhao, Hal Daumé III
Our empirical results indicate a preference among these models for hiring candidates with White female-sounding names over other demographic groups across 40 occupations
papercode
A Survey of State of the Art Large Vision Language Models: Alignment, Benchmark, Evaluations and Challenges
CVPR Workshop 2025
A Survey of State of the Art Large Vision Language Models: Alignment, Benchmark, Evaluations and Challenges
Zongxia Li, Xiyang Wu, Hongyang Du, Fuxiao Liu, Huy Nghiem, Guangyao Shi
A survey on Large Vision-Langauge Models.
papercode

All publications →