Unverified50% confidenceFactExact time
Agent系统通过ReAct、Chain-of-Thought等框架，让大语言模型在多轮循环中逐步分解和解决复杂任务
1
Sources
50%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
Related Claims
UnverifiedAgent能力的底层依赖于大语言模型的推理能力（Chain-of-Thought）以及ReAct（Reasoning + Acting）框架80% similarUnverified更复杂的Agent系统如ReAct框架和OpenAI的Assistants API在函数调用之上叠加了链式推理、记忆管理和多步规划能力77% similarUnverifiedThe ReAct (Reasoning + Acting) pattern refers to multi-step Agent tasks where each step dynamically decides the next action based on the previous step's output, with state continuously passed and accumulated throughout the entire pipeline.76% similarUnverifiedModern Agent systems can maintain a context window of up to tens of thousands of tokens within a single session and perform multi-step logical decomposition of complex experimental plans through chain-of-thought reasoning mechanisms76% similarUnverifiedReAct框架将推理与行动交织执行，AutoGen允许多Agent通过对话协作，LangGraph以有状态图结构管理复杂Agent执行流程76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/147822API
curl https://kongchang.com/api/v1/knowledge/claims/147822MCP
get_claim(id=147822)