I'm a fourth-year PhD student working on Machine Learning and Natural Language Processing at UC Irvine advised by Sameer Singh. My research is motivated towards designing systems that can robustly respond to real-world queries by operating over large amounts of information and producing arbitrarily complex natural language responses.
Natural Language Generation: The ability to generate responses is imperative for creating systems that can handle arbitrarily complex queries. We have worked on designing better evaluation metrics in the form of learned metrics as well as studied the factors that lead to hallucinations and proposed effective mitigation strategies.
Open Domain Language Reasoning: Given the scale of textual data, retrieval is required for efficient reasoning. We showed retrievers are not robust on the long tail of natural language and measured how interactions between retrievers and readers lead to hallucinations.