Oanh N. Tran
Hello, my name is Oanh N. Tran (also Oanh Tran or Tran Ngoc Oanh). I am a research assistant at VinUniversity, advised by Prof. Khoa D. Doan. My research studies failure modes and behavioral reliability in multimodal AI systems, with a focus on understanding why these systems behave unreliably and how to mitigate their failures.
Beyond AI, I enjoy playing Teamfight Tactics (TFT) and Ultimate Disc 🥏, as well as trying new recipes through cooking and baking 🍳🧁—though definitely not washing dishes 🫧.
News
- 06.2026 ViRGo, was accepted to the ICML 2026 Workshop on Efficient Multimodal Question Answering.
- 11.2024 I graduated from HCMUT with a GPA of 3.6/4.0.
- 08.2024 I started working as a Research Assistant at the MAIL Research Group.
- 06.2024 Our paper, “Cross-Data Knowledge Graph Construction for LLM-enabled Educational Question-Answering System: A Case Study at HCMUT,” was accepted to AIQAM ’24.
Publications
Look Before You Zoom: Adaptive Routing for the Resolution-Context Trade-off in Visual RAG EMM-QA @ ICML 2026
ViRGo adaptively routes visual question answering between global perception, patch-based retrieval, and attention-based retrieval to improve the accuracy-efficiency trade-off.
Cross-Data Knowledge Graph Construction for LLM-enabled Educational Question-Answering System: A Case Study at HCMUT AIQAM '24 @ ACM Multimedia 2024
A method for constructing a Knowledge Graph from cross-data educational sources and applying it with LLMs for domain-specific question answering.