System 1 agents: a method for local, reliable and fast agentic AI
arXiv:2611.21053v1 Announce Type: new Abstract: Agentic artificial intelligence could automate analyses of sensitive clinical data, but current agents depend on large language models (LLMs). Cloud LLMs expose patient data. Local LLMs hallucinate, and an invented value in a tool call propagates through every later step. Here we present a method to build System 1 agents: local agents that call tools without generating text. A deterministic compiler turns each tool's schema and the request into mul
阅读原始来源来源记录
项目分析
暂无五维分析,可先查看上方来源资料。