Scholar
Alexander H. Liu
Google Scholar ID: LIiCDa0AAAAJ
Massachusetts Institute of Technology
Citations & Impact
Citations
2,354
H-index
20
i10-index
27
Publications
20
Co-authors
11
list available
Publications
Cited
0
Cited
0
Cited
4
Cited
0
Cited
0
Resume (English only)
Academic Achievements
- 1. SHuBERT: Self-Supervised Sign Language Representation Learning via Multi-Stream Cluster Prediction, ACL 2025 (Oral)
- 2. UniWav: Towards Unified Pre-training for Speech Representation Learning and Generation, ICLR 2025
- 3. Generative Speech Foundation Model Pretraining for High-Quality Speech Extraction and Restoration, ICASSP 2025
- 4. Generative Pre-training for Speech with Flow Matching, ICLR 2024
- 5. Revisiting Self-supervised Learning of Speech Representation from a Mutual Information Perspective, ICASSP 2024
- 6. DinoSR: Self-Distillation and Online Clustering for Self-supervised Speech Representation Learning, NeurIPS 2023
- 7. Joint Audio and Speech Understanding, ASRU 2023
- 8. Listen, Think, and Understand, ICLR 2024
Research Experience
- 1. Member of the Spoken Language System (SLS) Group at MIT, working on natural language and speech processing
- 2. Member of the Speech Processing Lab at NTU, working on machine learning and speech processing
- 3. Research intern at Facebook AI Research (now FAIR at Meta AI) and Nvidia Applied Deep Learning Research (ADLR)
Education
- 1. Ph.D. candidate at Massachusetts Institute of Technology (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL), Advisor: Dr. James Glass
- 2. M.S. in Computer Science & Information Engineering (CSIE) from National Taiwan University (NTU), Advisors: Lin-shan Lee, Prof. Hung-yi Lee
- 3. B.S. in Computer Science & Information Engineering (CSIE) from National Taiwan University (NTU), Advisor: Yu-Chiang Frank Wang
Background
- Research interests include natural language and speech processing, with the goal of building machines that can seamlessly interact with humans through voice. Specific areas include multimodal audio representation learning, multimodal alignment, large language models, and generative models for audio.
Miscellany
- 1. Currently on leave until Spring 2026, working at Mistral AI to build frontier open audio models such as Voxtral
- 2. Inspired by Wei-Chiu Ma, committed to providing 1-2 hours per week for suggestions and/or mentorships to junior students in need, especially those from underrepresented groups
