Unverified50% confidenceTradeoffExact time
多智能体架构中每增加一个Agent就多一轮LLM调用，导致响应延迟线性甚至更快地累加
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
Unverified在Agent工作流场景下，每一轮LLM调用的延迟会被串行叠加74% similarUnverified在Agent场景下由于存在多轮工具调用和状态传递，Token消耗会随迭代次数线性甚至非线性增长74% similarUnverified智能体系统常通过定时任务每分钟不间断运行，加上多轮模型调用、工具调用结果重新输入及反思步骤，令令牌成本呈指数级攀升72% similarUnverified相比LangChain原生的AgentExecutor，LangGraph能更精细控制多步骤任务的状态持久化、错误重试和人机交互介入点72% similarUnverified多智能体系统带来新复杂性：智能体间每次上下文传递消耗额外Token，错误会在智能体链中级联放大70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/527972API
curl https://kongchang.com/api/v1/knowledge/claims/527972MCP
get_claim(id=527972)