Unverified60% confidenceFactTime unknown
Large language models suffer from 'attention decay' when generating long text, causing adherence to earlier instructions to decrease as context length increases
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
Related Claims
Verified大语言模型在长时间对话中会因上下文窗口限制而丢失早期信息79% similarUnverified级联式架构存在缺陷：每个环节引入延迟，中间的文字转换会丢失语调、情绪、语速等副语言信息77% similarVerified研究表明大语言模型在处理长文本时存在显著的「中间遗忘」（Lost in the Middle）现象，对输入文本开头和结尾的信息记忆较好，但对中间部分的信息关注度明显下降76% similarUnverifiedCharacter collapse occurs when conversations exceed the model's context length limit, causing early character setting information to be truncated or diluted75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/57496API
curl https://kongchang.com/api/v1/knowledge/claims/57496MCP
get_claim(id=57496)