✳flâneur — a map of the web's best reading

MetaGPT: Meta Programming for a Multi-Agent Collaborative Framework

arxiv.org · 10,626 words · saved by 1 readers

This is experimental HTML to improve accessibility. We invite you to report rendering errors. Use Alt+Y to toggle on accessible reporting links and Alt+Shift+Y to toggle off. Learn more about this project and help improve conversions. Remarkable progress has been made on automated problem solving through societies of agents based on large language models (LLMs). Existing LLM-based multi-agent systems can already solve simple dialogue tasks. Solutions to more complex tasks, however, are complicated through logic inconsistencies due to cascading hallucinations caused by naively chaining LLMs. Here we introduce MetaGPT, an innovative meta-programming framework incorporating efficient human workflows into LLM-based multi-agent collaborations. MetaGPT encodes Standardized Operating Procedures (SOPs) into prompt sequences for more streamlined workflows, thus allowing agents with human-like domain expertise to verify intermediate results and reduce errors. MetaGPT utilizes an assembly line p

MetaGPT: Meta Programming for a Multi-Agent Collaborative Framework Sirui Hong 1 1 1 footnotemark: 1 , Mingchen Zhuge 2 , Jiaqi Chen 1 , Xiawu Zheng 3 , Yuheng Cheng 4 , Ceyao Zhang 4 , Jinlin Wang 1 , Zili Wang , Steven Ka Shing Yau 5 , Zijuan Lin 4 , Liyang Zhou 6 , Chenyu Ran 1 , Lingfeng Xiao 1,7 , Chenglin Wu 1 , Jürgen Schmidhuber 2,8 1 DeepWisdom, 2 AI Initiative, King Abdullah University of Science and Technology, 3 Xiamen University, 4 The Chinese University of Hong Kong, Shenzhen, 5 Nanjing University, 6 University of Pennsylvania, 7 University of California, Berkeley, 8 The Swiss AI

Explore this link on the map →

saved by

related reading