Research
Materials for courses you taught. Replace this text with your description.
[This site is currently under construction, but here are some bullet points that will be fleshed out]:
- I work with various cognitively / neurally plausible computational models (e.g. gradient symbolic computation, self-organized sentence processing, ACT-R, dynamical field theory) to simulate phenomena of syntactic processing and sentence production, currently focusing on structural priming.
- I work on behavioral evaluation and mechanistic interpretability on language language models (LLM), pursuing the hypothesis of treating neural networks as cognitive models.
- At the behavioral level: I have investigated to what extent LLMs show similar behavioral profiles as humans in structural priming and dynamic anaphora resolution;
- At the mechanistic level: I focus on the underlying mechanisms of in-context learning (ICL), drawing connections between ICL and human priming. I am also interested in using human-interpretable programs (e.g., the RASP language, tensor product representation) to interpret LLMs.
- I am also interested in human sentence production theories based on the Tree-Adjoining Grammar (TAG) formalism.