🇻🇳 I visited Hue for the first time to present my work, "Benchmarking Open-Source Vietnamese-English Speech-to-Text Translation Systems," which will appear in the MAPR 2026 proceedings.
About
I study how large language models process language, how their behaviours emerge during training, and how interpretability can improve AI safety. My recent work examines whether LLMs internally route Vietnamese through English and how measurement tools can mislead for low-resource languages. My background in speech-to-speech translation and Vietnamese NLP informs my focus on interpretability across languages and reliable measurement.
Research interests
LLM interpretability · Training dynamics · AI safety · Multilingual & low-resource NLP · Speech processing
I’d be happy to discuss my research, potential collaborations, or related topics via email.
News
👨🏫 I joined SEAS-2026 again as a teaching assistant in Dong Hoi, Quang Tri. Many extraordinary people, many amazing memories.
🎉 Our multi-task work on Vietnamese audio-based toxic span detection was accepted to ICASSP 2026.
🌏 Our work on idiom-aware Vietnamese-English translation appeared at AACL-IJCNLP 2025. I really appreciate Assoc. Prof. Parisa Kordjamshidi for your heartfelt guidance and support.
🎙️ ViToSA, our work on toxic span detection in Vietnamese speech, appeared at INTERSPEECH 2025.
👨🏫 I joined SEAS-2025 for the first time as a teaching assistant, giving lectures and mentoring AI research projects for gifted, creative, and super-cool high school students in Dong Hoi, Quang Tri (Quang Binh).
📄 My master's thesis research on a zero-resource speech-to-speech translation approach was published in IEEE Access. Many thanks to my master's supervisor, Prof. Sakti Sakriana, for her supervision and support throughout my time at JAIST.
📊 Our dataset and benchmark for multi-dialect Vietnamese appeared at EMNLP 2024. See you in Miami, Florida!
🇺🇸 I landed in the United States for the first time and began a three-month visiting scholar period at the HLR Laboratory at Michigan State University under the supervision of Assoc. Prof. Parisa Kordjamshidi.
🔥🔥 I published my first A* paper at ACL based on my solo-author research, "ViHateT5: Our Unified Text-to-Text Approach to Vietnamese Hate Speech Detection."
💯 I completed my Master degree in Information Science at JAIST with a thesis grade of 100/100.
🎓 I joined the Department of Data Science at UIT, VNU-HCM, as a lecturer.