I am a PhD student at the University of Virginia advised by Chirag Agarwal. Previously, I was a PhD student at the College of William and Mary advised by Antonio Mastropaolo. I am also fortunated to be mentored by Anh Totti Nguyen, Truong-Son Hy, and Thiago Serra.

I am interested in multimodal AI and AI safety: (1) evaluating and understanding LLMs/MLLMs and (2) making AI systems more robust and interpretable in high-stakes applications.

I was a Machine Learning Research Intern at CodaMetrix (Summer 2024 & 2025), where I developed LLM agents to (1) extract medical entities from EHRs and (2) evaluate and correct entities extracted by human experts and LLMs.

Selected Publications

♠ denotes equal contribution

ICLR 2026

Vision-Language Models are Biased

An Vo^♠, Khai-Nguyen Nguyen^♠, Mohammad Reza Taesiri, Vy Tuong Dang, Anh Totti Nguyen, Daeyoung Kim

Summary

We demonstrate that state-of-the-art MLLMs are strongly biased toward well-known patterns through extensive mechanistic analysis of these models on our proposed VLMsAreBiased benchmark.

paper / website / dataset / code

ASE 2026

Pattern over Pixels: Measuring Pattern Completion Bias in Multimodal Code Generation

Khai-Nguyen Nguyen, Oscar Chaparro, Antonio Mastropaolo

Summary

We introduce the first benchmark measuring pattern-completion bias in screenshot-to-code generation, showing that frontier MLLMs complete familiar UI patterns instead of reading what is actually on screen.

paper / website / dataset

TOSEM

Toward Explaining Large Language Models in Software Engineering Tasks

Antonio Vitale^♠, Khai-Nguyen Nguyen^♠, Denys Poshyvanyk, Rocco Oliveto, Simone Scalabrino, Antonio Mastropaolo

Summary

We present FeatureSHAP, an interpretability framework for software engineering tasks that attributes Shapley scores to input features based on their contributions to model output.

paper / code

ACL 2025 (Industry Track) Oral

Sentiment Reasoning for Healthcare

Khai-Nguyen Nguyen^♠, Khai Le-Duc^♠, Bach Phan Tat, Duy Le, Long Vo-Dang, Truong-Son Hy

Summary

We show that training LLMs on chain-of-thought reasoning improves their performance in sentiment analysis while enabling human-like explanations.

paper / code / dataset

Interspeech 2024 Oral

Real-time Speech Summarization for Medical Conversations

Khai Le-Duc^♠, Khai-Nguyen Nguyen^♠, Long Vo-Dang, Truong-Son Hy

Summary

We show that an even split in the annotation budget between synthetic and human-curated data collection yields the best dataset for medical speech summarization LLMs.

paper / dataset

CPAIOR 2023 ICLR Workshop

Getting Away with More Network Pruning

Jeffrey Cai^♠, Khai-Nguyen Nguyen^♠, Nishant Shrestha, Aidan Good, Ruisen Tu, Xin Yu, Shandian Zhe, Thiago Serra

Summary

We propose a mathematical theorem on the upper bound of the expressiveness of a neural network based on their geometric properties and apply it to model pruning.

paper / code

Selected Preprints

Prior Bias in VLMs on UML Diagram Interpretation

Under Review

Prior Bias in Vision Language Models on UML Diagram Interpretation

Zaiyu Cheng^♠, Khai-Nguyen Nguyen^♠, Antonio Mastropaolo

Summary

We show that VLMs interpret UML diagrams from pretrained priors rather than the diagram itself: simply reversing relation arrows drops open-source model accuracy by 33.48% on average.

paper

Under Review

S-Chain: Structured Visual Chain-of-Thought for Medicine

Khai Le-Duc, Phuong T.H. Trinh, Duy Minh Ho Nguyen, Tien-Phat Nguyen, Nghiem Tuong Diep, An Ngo, Tung Vu, Trinh Vuong, Anh-Tien Nguyen, Nguyen Dinh Mau, Van Trung Hoang, Khai-Nguyen Nguyen, Hy Nguyen, Chris Ngo, Anji Liu, Nhat Ho, Anne-Christin Hauschild, Khanh Xuan Nguyen, Thanh Nguyen-Tang, Pengtao Xie, Daniel Sonntag, James Zou, Mathias Niepert, Anh Totti Nguyen

Summary

We present S-Chain, a medical dataset for strutured visual reasoning. Training medical models on S-Chain improves their accuracy and explainability.

paper / code / dataset