Unverified50% confidenceFactExact time
Lost in the Middle 效应指模型对位于极长输入序列中间位置的信息关注度会系统性下降，在超过 20K token 时开始显著，100K token 级别尤为突出
1
Sources
50%
Confidence
Long-term
Relevance
7/8/2026
First Seen
Sources
Claude Code桌面版完整指南：无账号接入DeepSeek国产模型
bilibili程序员比巴布7/6/2026
Related Claims
UnverifiedThe "Lost in the Middle" effect refers to the phenomenon where models pay less attention to content injected into the middle positions of the context than to content at the beginning and end.77% similarUnverifiedAn "lost in the middle" effect exists in ultra-long contexts, meaning the model has stronger memory for content at the beginning and end of a sequence while attention to middle passages decreases; this phenomenon was named by Stanford researchers.73% similarUnverified研究发现位于上下文首尾的信息被模型召回的概率远高于中间位置，衰减在上下文长度超过4K token后急剧加速71% similarVerified研究表明当上下文过长时，模型对中间部分信息的注意力会显著下降，即"Lost in the Middle"现象70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/279138API
curl https://kongchang.com/api/v1/knowledge/claims/279138MCP
get_claim(id=279138)