Generative Agents: Interactive Simulacra of Human Behavior
Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, Michael S. Bernstein
Abstract
We introduce generative agents — computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, head to work, paint, write, form opinions, notice each other, and initiate conversations. We describe an architecture that extends a large language model to store a complete record of the agent's experiences, synthesize those memories into higher-level reflections, and retrieve them dynamically to plan behavior.
Key Findings
- 1Created believable simulations of human behavior using LLM-based agents
- 2Introduced memory, reflection, and planning architectures for AI agents
- 3Demonstrated emergent social behaviors in a simulated town of 25 agents
- 4Showed agents forming relationships, spreading information, and coordinating
- 5Established a framework for memory-augmented autonomous agents
Impact & Significance
This paper captured public imagination about AI agent capabilities and provided architectural patterns (memory, reflection, planning) that became standard in AI agent frameworks like AutoGPT, CrewAI, and LangGraph.
Related Papers
The Llama 3 Herd of Models
Meta AI
Qwen2 Technical Report
Alibaba Cloud / Qwen Team
DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
DeepSeek AI
The Claude 3 Model Family: Opus, Sonnet, and Haiku
Anthropic