Human Robot Collaboration with Few-Shot LLM Robot Models

Maitrey Gramopadhye, Daniel Szafir

View the paper here

Large Language Models have had a notable surge in popularity in task planning for embodied robots. However, most approaches have the robot operating in isolation with minimal collaboration with humans. In this paper, we design a system that enables people to interact with an intelligent robot. We conduct a human subjects study to gain insights into the participants’ mental model, and whether the comprehensive abilities of LLMs encourage the users to adopt a collaborative role when working with the robot for long-horizon tasks.