HuggingGPT
HuggingGPT是一个基于大型语言模型的智能代理（Agent）框架，由微软研究院提出。它以ChatGPT等大语言模型作为任务规划与决策核心，通过调用Hugging Face平台上托管的多种专业AI模型，协同完成自然语言处理、图像识别、语音理解等跨模态复杂任务，实现了"以语言模型为控制器、多专业模型为执行者"的多模型协作机制。
Timeline (last 90 days)
In 2023, several important papers including Toolformer and HuggingGPT explored methods for enabling large models to autonomously select and call external tools; OpenAI's Function Calling mechanism standardized this capability.
The theoretical foundations of tool-augmented large language models can be traced back to research such as Toolformer, the ReAct framework, and HuggingGPT from around 2023.
All Facts (3)
HuggingGPT的架构是一个大模型指挥多个专业模型协作，能解决复杂的跨模态问题
90%UnverifiedThe theoretical foundations of tool-augmented large language models can be traced back to research such as Toolformer, the ReAct framework, and HuggingGPT from around 2023.
50%UnverifiedIn 2023, several important papers including Toolformer and HuggingGPT explored methods for enabling large models to autonomously select and call external tools; OpenAI's Function Calling mechanism standardized this capability.
50%