I am interested in natural language processing. Here are some areas I have worked on:
- Language to code (executable semantic parsing): Convert textual descriptions into executable computer programs or database queries. Some common themes include learning from distant supervision, retrieval augmentation, and compositional generalization.
- Web interaction: Interact with web pages based on the given natural language commands.
- Retrieval-augmented models: Design models that can retrieve information and use it to guide predictions. The explicit retrieval step makes the model more interpretable, and we can update the model by modifying the retrieval index.
- Factuality: Verify the factual correctness of the given text, and prevent language models from generating factually incorrect claims.
I did my PhD at Stanford University, where I was advised by Percy Liang. I was part of the Stanford NLP Group. Before that, I received my bachelor's degree from MIT. I like ice cream, languages, and pencil puzzles.