Unverified50% confidenceOpinionExact time
初始token消耗越低，用户可用于实际任务的上下文空间越大；过长的系统提示词会引发提示词漂移干扰模型自主发挥
1
Sources
50%
Confidence
Long-term
Relevance
7/19/2026
First Seen
Sources
Related Claims
UnverifiedCompared to the polling model, an event-driven asynchronous architecture can achieve an order-of-magnitude reduction in token consumption.77% similarVerifiedEven models with a million-token context face an attention dilution problem when handling large codebases—when the context is too long, the model's attention to early content drops significantly.75% similarUnverifiedSub-agents with narrower task scopes require smaller context windows, directly reducing inference costs and lowering error rates caused by context contamination.74% similarUnverifiedToo much persistent context consumes window space and introduces noise, while too little transient context causes the agent to lose critical constraints, requiring a trade-off between the two.74% similarVerifiedIrrelevant information occupying context space (context pollution) can significantly degrade the model's performance on core tasks.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/561469API
curl https://kongchang.com/api/v1/knowledge/claims/561469MCP
get_claim(id=561469)