Wei-Ning Hsu
Research scientist at Meta AI (formerly FAIR); first author of HuBERT
- Research Scientist, Meta AI (FAIR)123
Wei-Ning Hsu is a research scientist at Meta AI (formerly FAIR), working on representation learning, self-supervised learning, and generative modeling for speech, including multimodal speech. Hsu is first author of HuBERT, a self-supervised speech representation model that predicts hidden units for masked regions, and Hsu's homepage also highlights Audio-Visual HuBERT, data2vec, wav2vec-U, and textless speech-to-speech translation. Meta's people page lists Hsu as a co-author of more recent work such as SAM Audio: Segment Anything in Audio (2025).1245
Work
HuBERT (external site: arxiv.org)Published research
First author of HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units (2021), which uses an offline clustering step to provide target labels for a BERT-like prediction loss over masked speech.42
data2vec (external site: arxiv.org)Published research
Co-author of data2vec (2022), a self-supervised learning framework that uses the same method for speech, vision, and language; the paper says models and code are available in the fairseq repository.562
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.