Unverified80% confidenceFactExact time
典型的Agent框架如ReAct、AutoGPT、LangGraph采用感知-推理-行动的迭代模式，上下文随轮次累积导致后期每轮Token消耗呈线性甚至指数增长
1
Sources
80%
Confidence
Long-term
Relevance
4/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMost Agent frameworks adopt a full context replay strategy, requiring the entire conversation history to be passed in with each call, causing token consumption to grow quadratically with the number of loop iterations.79% similarUnverified在Agent应用中，每一轮推理-行动-观察循环都需要将完整历史上下文重新输入模型，导致Token消耗随任务复杂度呈非线性增长78% similarUnverifiedReAct强调单步推理-行动的紧密耦合，适合短链任务；AutoGPT类框架倾向先规划完整任务树再批量执行，适合长链任务78% similarUnverifiedFrameworks such as Reflexion, AutoGPT, and MetaGPT all follow the reasoning-action layered architecture of ReAct.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/10881API
curl https://kongchang.com/api/v1/knowledge/claims/10881MCP
get_claim(id=10881)