Speaker: Denis Peskoff (UC Berkeley iSchool)
UTA 5.522 or Zoom
Title: Laws & LLMs
Abstract: In a series of recent papers, we study LLM knowledge of the laws and create new datasets to improve it. The results are surprising. I am omitting details from the abstract, so come to the talk to learn more (or how to spend 1.5 million tokens answering 500 yes/no questions). In addition to this, I will touch upon past work of in-domain LLM evaluation, genAI detection, studying communications of the Fed, and the board game of Diplomacy.
Bio: Denis Peskoff is a Bellwether Postdoctoral Scholar at UC Berkeley with Professor Diag Davenport working on legal NLP. Previously he was at Northwestern, Princeton, Ludwig-Maximilians-Universität München, and the University of Maryland. His research interest is leveraging domain experts to create meaningful and accurate datasets for computational social science.
Lunch will be provided at noon