Verified65% confidenceSolutionExact time
基于LLM的智能体框架通常采用ReAct（推理+行动）或Plan-and-Execute等设计模式
3
Sources
65%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
Related Claims
Unverified基于LLM的现代Agent以大语言模型为大脑，通过ReAct、Plan-and-Execute等范式自主规划步骤、调用外部工具并根据反馈动态调整策略84% similarUnverifiedReAct、AutoGen、LangGraph等Agent架构形式化了感知-推理-行动循环，使LLM专注于规划与编排81% similarUnverified任务规划（Planning）主流方案包括 ReAct（推理-行动交替）、Tree of Thoughts（思维树）和基于 LLM 的动态计划生成81% similarVerifiedReAct框架使LLM在回答问题时能够交替进行'推理'和'行动'步骤79% similarUnverifiedLLM作为调度中枢的设计模式被称为ReAct（Reasoning + Acting）框架78% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/504345API
curl https://kongchang.com/api/v1/knowledge/claims/504345MCP
get_claim(id=504345)