Unleashing the Emergent Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration

Human intelligence thrives on cognitive synergy, where collaboration amongdifferent minds yield superior outcomes compared to isolated individuals. Inthis work, we propose Solo Performance Prompting (SPP), which transforms asingle LLM into a cognitive synergist by engaging in multi-turnself-collaboration with multiple personas. A cognitive synergist is anintelligent agent that collaboratively combines multiple minds' strengths andknowledge to enhance problem-solving in complex tasks. By dynamicallyidentifying and simulating different personas based on task inputs, SPPunleashes the potential of cognitive synergy in LLMs. Our in-depth analysisshows that assigning multiple fine-grained personas in LLMs improvesproblem-solving abilities compared to using a single or fixed number ofpersonas. We evaluate SPP on three challenging tasks: Trivia Creative Writing,Codenames Collaborative, and Logic Grid Puzzle, encompassing bothknowledge-intensive and reasoning-intensive types. Unlike previous works, suchas Chain-of-Thought, that solely enhance the reasoning abilities in LLMs,experimental results demonstrate that SPP effectively reduces factualhallucination, and maintains strong reasoning capabilities. Additionally,comparative experiments show that cognitive synergy only emerges in GPT-4 anddoes not appear in less capable models, such as GPT-3.5-turbo andLlama2-13b-chat, which draws an interesting analogy to human development. Code,data, and prompts can be found at:https://github.com/MikeWangWZHL/Solo-Performance-Prompting.git.