Unverified50% confidenceFactExact time
Overly long system instructions in LLMs lead to 'attention decay' on critical rules due to non-uniform attention distribution across the context window.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
Related Claims
VerifiedLLM存在'Instruction Drift'现象：随着对话轮次增加，早期系统提示在模型注意力权重中逐渐衰减80% similarUnverifiedLLM存在上下文窗口限制，在超长对话中面临注意力稀释问题，早期决策记录会被赋予更低注意力权重78% similarUnverified当LLM上下文窗口中的指令过多时，模型对每条指令的关注度会下降，导致遵循率降低76% similarUnverified过于冗长、矛盾或模糊的系统提示词会导致模型注意力分散，产生过度思考或指令遵循失败，精简 agents.md 可提升指令遵循的确定性74% similarUnverifiedLLM对上下文开头和结尾的信息保持较高敏感度，而对中间位置的信息存在'注意力衰减'现象73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/43296API
curl https://kongchang.com/api/v1/knowledge/claims/43296MCP
get_claim(id=43296)