Unverified50% confidenceFactExact time
自主代理的规划-执行-反馈循环架构最早在ReAct（Reasoning + Acting）论文中被系统化描述
1
Sources
50%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
Claude Sonnet 5接入Devin：编程性能升级，配额消耗降低30%
rss6/30/2026
Related Claims
VerifiedReAct（Reasoning + Acting）论文最早系统化描述了AI代理的「感知-推理-行动」迭代闭环机制，后来成为LangChain、AutoGPT等框架的基础范式78% similarUnverified代理式工具引入ReAct（Reasoning + Acting）框架，形成完整的感知-规划-执行闭环77% similarUnverified自主Agent基于ReAct（Reasoning + Acting）框架进行'感知-规划-行动-观察'循环，对初始上下文质量要求更高77% similarVerifiedPlan-and-Execute和ReAct是当前Agent领域两种代表性的规划范式77% similarVerifiedReAct（Reasoning + Acting）架构中Agent在每一步先推理当前状态再决定下一步行动，形成感知-规划-执行的闭环76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/201812API
curl https://kongchang.com/api/v1/knowledge/claims/201812MCP
get_claim(id=201812)