Mutian He

Mutian He is a PhD candidate and Doctoral Assistant at the École Polytechnique Fédérale de Lausanne (EPFL), Switzerland, affiliated with the Idiap Research Institute and the School of Engineering. He is pursuing his doctoral studies in Electrical Engineering under the supervision of Phil Garner. He holds a B.E. from Beihang University (BUAA) and an MPhil from the Hong Kong University of Science and Technology (HKUST).

• B.E., Beihang University (BUAA), 2019
• MPhil, Hong Kong University of Science and Technology, 2022
• PhD Candidate, École Polytechnique Fédérale de Lausanne (EPFL), ongoing

His research focuses on spoken language understanding, speech synthesis, and the intersection of speech and language processing with machine learning. He explores efficient model architectures, pretraining strategies, multilingual and low-resource modeling, and the use of large language models in speech tasks. His work spans both theoretical and applied aspects, including distillation to linear-complexity models, robust TTS, and commonsense reasoning via conceptualization.

His recent publications at top venues such as ICLR, EMNLP, Interspeech, and KDD demonstrate a strong trend towards efficient and scalable models for speech and language, with increasing emphasis on multilingualism, knowledge transfer, and real-world deployment in low-resource settings. He has also contributed to open-source implementations and community tools like Speech Rankings.

• Joint Fine-tuning and Conversion of Pretrained Speech and Language Models towards Linear Complexity (ICLR 2025)
• Acquiring and Modelling Abstract Commonsense Knowledge via Conceptualization (AIJ 2024)
• The Interpreter Understands Your Meaning: End-to-end Spoken Language Understanding Aided by Speech Translation (Findings of EMNLP 2023)
• Can ChatGPT Detect Intent? Evaluating Large Language Models for Spoken Language Understanding (Interspeech 2023)
• Multilingual Byte2Speech Models for Scalable Low-resource Speech Synthesis (2022)

Mutian He has served as a teaching assistant for courses including Introduction to Natural Language Processing at HKUST and Introduction to Speech Processing at Idiap. He has also worked on speech synthesis at Microsoft, focusing on robustness and multilingual conditions. He is actively involved in research advising under Phil Garner and has collaborated with multiple researchers across institutions.

He is affiliated with the LIDIAP (Laboratory of Intelligent Data Analysis and Pattern Recognition) at EPFL, where he contributes to research in deep learning for speech and language. His lab work involves developing novel neural architectures, conducting experiments on multilingual datasets, and open-sourcing code to promote reproducibility.