Skip to main content

How Statistics Can Advance Large Language Models: Fairness Alignment and Watermarking

Abstract:

Large language models (LLMs) have rapidly emerged as a transformative innovation in machine learning. However, their increasing influence on human decision-making processes raises critical societal questions. In this talk, we will demonstrate how statistics can help address two key challenges: combating misinformation through watermarking and ensuring fairness for minority groups through alignment. First, we introduce a general statistical framework to analyze the efficiency of watermarking schemes for LLMs. We develop optimal detection rules for an important watermarking scheme recently developed at OpenAI and empirically demonstrate its superiority over the existing detection method. Second, we tackle the challenge of creating fair LLMs that equitably represent and serve diverse populations. We derive a regularization term that is both necessary and sufficient for aligning LLMs with human preferences, ensuring equitable outcomes across different demographics. Throughout the talk, we will showcase how statistical insights can not only address pressing challenges posed by LLMs but also unlock substantial opportunities for the field of statistics to drive responsible generative AI development.

 

Presented By: Weijie Su (University of Pennsylvania)