Unverified50% confidenceFactExact time
In agent scenarios, the context window accumulates continuously with each conversation turn, causing the input tokens for each subsequent turn to grow persistently.
1
Sources
50%
Confidence
Long-term
Relevance
10/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAgentExecutor每次工具调用前会重新拼接完整对话历史进Prompt，导致Token消耗随任务步骤线性增长78% similarUnverified多代理系统中每个代理都需要携带系统提示、历史对话和工具定义进入上下文窗口，Token消耗呈乘数级增长74% similarUnverifiedAgent框架慢于直接对话主要因为注入大量系统提示词工具定义、ReAct多轮推理和上下文累积导致输入token数增长74% similarUnverifiedAgents repeatedly execute the perception-planning-action loop in cycles, and each round requires packing the complete history into the context window, causing input tokens to grow linearly or even exponentially.73% similarUnverifiedIn multi-agent collaboration, each subsequent agent must carry the full output of all preceding agents as context, causing the context window to continuously expand as the workflow progresses and token consumption at terminal nodes to increase accordingly.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/973259API
curl https://kongchang.com/api/v1/knowledge/claims/973259MCP
get_claim(id=973259)