Hongao Zhu

Hongao Zhu

PhD student in computational linguistics at UC San Diego

I am Hongao Zhu (IPA: /xʊŋ35 ɑʊ35/ /ʈʂu55/, Mandarin: 朱虹翱, pinyin: ZHŪ HÓNG ÁO).

I am a second-year PhD student in computational linguistics at UC San Diego, advised by Alex Warstadt in the Learning, Meaning, and Natural Language Lab (LeM🍋N Lab). Before UCSD, I received a B.A. in Linguistics from Shanghai Jiao Tong University.

I use language models to study how people learn and process language: whether models can show the critical-period effects seen in human second-language learners, when language model surprisal predicts human reading times across languages, and how to test what models know about grammar, as with the minimal-pair benchmark ZhoBLiMP. At SJTU, I worked with Hai Hu on how language models acquire the Ba construction, and since 2024 I have also worked with William Schuler at Ohio State on language model surprisal as a predictor of human reading times. I also co-organized the first ChineseBabyLM Challenge.

I’m looking for a research internship for summer 2027, especially on how people understand and interact with language models, on model evaluation, or on computational psycholinguistics. See the Contact page.

Curious what I have been thinking about lately? My Puzzles page covers questions such as whether language models can show a critical period, what makes a construction learnable, how to benchmark what vision-language models understand about sign language, when a voice model should take its turn, whether humans are LLMs, and how to detect AI impersonation.

To know more about me and my research, please visit the About page.

Posts