Algorithmic Prompt Generation for Diverse Human-like Teaming and Communication with Large Language Models
Researchers have developed a novel method for generating diverse, human-like team behaviors in large language model (LLM) agents. By combining Quality Diversity (QD) optimization with LLM-powered agents, the approach iteratively searches for prompts that elicit varied coordination and communication strategies in collaborative, long-horizon tasks. A human-subjects experiment confirmed the diversity of human behavior in the chosen domain, and subsequent studies demonstrated that the generated LLM behaviors are both human-like and capture patterns difficult to observe without extensive data collection. AI
IMPACT This research offers a new method for generating diverse agent behaviors, potentially accelerating studies in human-agent teaming and AI-assisted decision-making.