Unverified50% confidenceFactExact time
The Stanford Town project placed 25 Agents driven by large language models in a The Sims-like sandbox environment, with each Agent having an independent Memory Stream architecture.
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA generative agent paper published by Stanford University in 2023 was the first to systematically introduce memory streams, reflection, and planning mechanisms for LLM agents, demonstrating that agents with memory organization capabilities outperform baseline systems without memory management in behavioral coherence and goal achievement rate.67% similarVerifiedStanford's Smallville virtual social experiment had 25 agents simulate the interactions of small-town residents, revealing the possibility of emergent social behaviors in multi-agent collaboration.66% similarUnverified微软的AutoGen框架和斯坦福的Generative Agents论文（Smallville实验，让25个Agent模拟人类社会行为）是多智能体系统方向的标志性工作65% similarUnverified斯坦福2023年「虚拟小镇」（Generative Agents）实验以25个 GPT-4 实例模拟社会行为，展示了 LLM 多智能体的涌现协作能力64% similarUnverifiedIn Codex's multi-agent collaboration, each Agent runs in an isolated sandbox container with its own execution environment and context.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946472API
curl https://kongchang.com/api/v1/knowledge/claims/946472MCP
get_claim(id=946472)