Huck Yang

about

Learning to scale, evolve, and hypothesize for the well-being of silicon and organic life; along the way, I’ve spent time at nvidia, amazon, google (ex-speech/brain); tsmc. I received my grad degrees at georgia tech with the wallace coulter fellowship and my undergrad at national taiwan university.

interests

• scaling laws & science: neural scaling laws for time series [ICLR 25] and multilingual ASR/translation in OWLS [ICML 25]. time-series reasoning: TimeOmni-1 [ICLR 26], open grpo models. LLMs in the scientific hypotheses generation [Nature npj AI 25] and network motifs [Nature Comm 25].
• multimodal post-training: co-led omni/audio traces for Nemotron-Omni-30B [GTC 26] and long-horizon alignment in OmniVinci-9B [ICLR 26]; RL post-training on audio [ICLR 25, 24]. multilingual audio post-training in Google [ICASSP 23]. The first voice prompting for frozen acoustic models [ICML 21].
• voice interactive agents: the first n-best generative ASR correction (GER) [ASRU 23], Whispering-LLaMA [EMNLP 23], HyPoradise [NeurIPS 23], audio post-training [ICLR 24]. best industry paper honorable mention [ACL 25]. Speech-Hands [ACL 26] and Voice Memory [SLT 26].

out of curiosity, on quantum machine learning, I created the first variational circuit based speech [ICASSP 21] and text classification [ICASSP 22] and received the Xanadu AI Quantum ML Award in 2019; recently, on quantum parameter adaptation for LLMs in [ICLR 25].