bio
Short, medium, and long biography for press, talks, and introductions.
Feel free to use any of the bios below for conference introductions, press pieces, or speaker profiles.
Concise Bio (1 sentence)
An Vo is a Research Engineer at MBZUAI working on trustworthy Large Language Models and Vision Language Models, and lead author of VLMs are Biased (ICLR 2026).
Short Bio (2–3 sentences)
An Vo is a Research Engineer at the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), working with Prof. Thamar Solorio. His research focuses on making Large Language Models (LLMs) and Vision Language Models (VLMs) more trustworthy and explainable in edge cases. He is the lead author of VLMs are Biased (ICLR 2026), which was featured on the front page of Hacker News and cited by Google DeepMind and ByteDance.
Long Bio
An Vo is a Research Engineer at the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), where he works with Prof. Thamar Solorio. He obtained his MS from KAIST under the supervision of Prof. Anh Totti Nguyen and Prof. Daeyoung Kim, where he also collaborated closely with Dr. Mohammad Reza Taesiri. His MS was fully funded by the Hyundai CMK Global Scholarship, a prestigious award for graduate students in South Korea. Prior to joining KAIST, he graduated as valedictorian from Vietnam National University – Ho Chi Minh City (VNU-HCM) in 2023 under the supervision of Prof. Ngoc Hoang Luong.
His research is broadly focused on Large Language Models (LLMs) and Vision Language Models (VLMs), with a particular emphasis on trustworthiness, explainability, and behavior in hard or edge cases. In his earlier work, he contributed to the intersection of Evolutionary Computation, Multi-objective Optimization, and AutoML.
An is the lead author of VLMs are Biased (ICLR 2026), which introduced VLMBias — a benchmark revealing that state-of-the-art models (including o3, o4-mini, Gemini 2.5 Pro, and Claude 3.7 Sonnet) achieve near-perfect accuracy counting familiar objects but drop to ~17% accuracy in counterfactual scenarios. The paper was featured on the front page of Hacker News (top-5), covered by Gary Marcus in his Substack, and tweeted by Lucas Beyer and Yoav Goldberg. It has since been cited and used by Google DeepMind and ByteDance. His work has been accepted at ICLR, ICML, AAAI, GECCO, and other top venues.