Researchers have introduced AI Tour Meeting, a novel framework for group travel planning that leverages multiple Large Language Model (LLM)-based agents. Each agent is assigned a distinct persona and engages in natural language discussions to collaboratively craft an itinerary that meets the group's constraints and preferences.
The framework is designed to facilitate easy and flexible orchestration of these multi-agent discussions. It provides interfaces for configuring agent personas, defining discussion workflows, monitoring progress, and deploying LLMs, making it adaptable to various planning scenarios.
AI Tour Meeting serves two primary use cases. First, it acts as a simulation tool for analyzing how multiple LLM agents behave during tour planning discussions. Second, it functions as a recommender system where persona-based agents represent group members who cannot participate, offering ideas from their perspectives.
The paper demonstrates the framework's utility through system validation and presents several analytical results. The code and documentation are openly available on GitHub, and the full paper is accessible on arXiv.