Reproduce “HARBOR: Exploring Persona Dynamics in Multi-Agent Competition” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kenan Jiang, Li Xiong, Fei Liu
Researchers at Emory University developed HARBOR, a framework to investigate how assigned personas influence Large Language Model agents' bidding behavior and how agents can leverage Theory of Mind for strategic advantage in competitive house auctions. They found that personas cause agents to deviate from purely profit-driven behavior, while accurate competitor profiling combined with strategic modules significantly enhances competitive performance.
Key insights
- Assigning personas to LLM agents significantly influences their decision-making, causing them to deviate from purely rational, profit-maximizing behavior towards preference adherence.
- LLM agents can effectively profile competitors' personas from observed bidding behaviors, with accuracy improving as more persona-aligned data becomes available.
- Leveraging inferred competitor personas through Theory of Mind, particularly when combined with expert-crafted strategic modules, substantially enhances an agent's competitive performance in multi-agent auctions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.