Huteng Dai

Huteng Dai

CV
North Quad 3417, 105 South State Street, Ann Arbor, MI 48109

Linguistics · Weinberg Cognitive Science · AI Lab · MIDAS · MCAIM
prev. PhD Rutgers

My name is Huteng Dai ([xu˨˩ tʰəŋ˧˥ tai˥˨] ≈ who-tongue dye; meaning). I'm a assistant professor in computational linguistics & phonology at the U‑M Linguistics Department.

Research Interests

Memory Acquisition Interpretability Language Models Formal Language Theory

I take a principle-centered approach to cognitive science and computational linguistics (cf. Smolensky et al. 1992), building on a few first principles.

See the principles
  • Language makes infinite use of finite means.
  • Working memory has finite capacity.
  • Language processing is predictive.
  • Generalization is formed over distributed representations, e.g. phonological features.
  • Learning is possible from positive evidence alone.

What I study is the class of theories and models these principles point to: memory-bounded and incrementally predictive, symbolic in structure yet distributed in representation—properties that constitute the inductive biases that make learning possible.

News

Older news
  • Invited talk, Peking University, "ML for linguists".
  • Talk, EMNLP 2025 (main conference) "Mind the Gap: How 'Baby' Language Models Learn Long-Distance Filler-Gap Dependencies." See you in Suzhou!
  • Invited talk, HCAI'25: The University of Michigan Symposium on Human-Centered AI [link].
  • Invited talk, The Education University of Hong Kong [news].
  • Paper "An Exception-Filtering Approach to Phonotactic Learning" published in Phonology [article] [podcast].
  • Invited talk, LSA organized session "Computational Models of Learnability and Acquisition of Morphology and Phonology."
  • Teaching LING 702: Language & Information.
  • Teaching LING 123: The ABCs of Python.
  • Launching the Computation, Acquisition, and Processing (CAP) Lab [website]!
  • Teaching "The Linguistics of Speech Recognition" at U-M.
  • Serving as Area Chair of CoNLL 2024.
  • Accepted a tenure-track assistant professor position at the University of Michigan, Ann Arbor [link]. Go Blue!
  • Tutorial "Doing Formal Language Theory" for MathLing [slides][code].
  • Served as a graduate mentor at Rutgers Honors College.
  • Secured the University & Louis Bevier Dissertation Completion Fellowship.
  • Talk with Connor Mayer (UCI) and Richard Futrell (UCI) "Rethinking representations: A log-bilinear model of phonotactics" at SCiL 2023 [paper][slides].