Xiaowu Dai
Assistant Professor at UCLA Statistics and Data Science
Xiaowu Dai joined the UCLA statistics department as a tenure-track Assistant Professor in Fall 2022. It is now known as the UCLA Statistics & Data Science department. Previously, he did a postdoc at UC Berkeley from 2019-2022, advised by Professor Michael I. Jordan. He was a member of the Berkeley AI Research Lab, Consortium of Data Analytics in Risk, and the Simons Institute for the Theory of Computing. Before that, he earned his Ph.D. in Statistics from UW-Madison, advised by Professor Grace Wahba. His research focuses on statistical theory and methodology for real-world problems that blend computational, inferential, and economic considerations. During his free time, he enjoys running, reading history, traveling to new places, and meeting new people from diverse backgrounds.
Sessions
Training-Free Multi-Agent Language Models
Large Language Models are powerful but often unreliable. This talk introduces Peer Elicitation Games (PEG) — a novel, training-free method that uses game theory to make LLMs more truthful and consistent. By letting models evaluate each other’s responses, PEG encourages honesty without extra training data and significantly improves factual accuracy.