Title: Training Large Language Models for Human Robot Interaction with Reinforcement Learning

Presented by Enoch Levandovsky, Computing PhD student, Computer Science emphasis

Abstract

Large Language Models (LLMs) are widely used in many areas of science and technology. In this paper, I focus on how Reinforcement Learning (RL) and Auto-Regressive (AR) LLM training come together. Firstly, I introduce key concepts of a traditional RL Algorithm.

Second, I describe how RL is used to train Auto-regressive LLMs.

Third, I describe how LLMs are used for human-robot interaction dialogue management and planning. Finally, I identify some missing research gaps in using LLMs and RL for human-robot interaction.

Advisor: Dr. Casey Kennington

Committee Members: Dr. Jun Zhuang, Dr. Tim Andersen

External Examiner: Dr. Yu Zhang