Expired65% confidenceFactExact time
A 2023 study by Stanford University showed that when input text exceeds a certain length, the model's attention to information in the middle of the context window drops significantly (the "Lost in the Middle" phenomenon).
3
Sources
65%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Related Claims
UnverifiedA 2023 Stanford University paper titled 'Lost in the Middle' found that models pay significantly less attention to information in the middle of the context window than at the beginning and end, exhibiting a U-shaped attention distribution.82% similarVerified斯坦福大学2023年发表了研究论文《Lost in the Middle》，实验证明当关键信息被放置在长上下文的中间位置时，模型的检索准确率会大幅下降80% similarVerifiedA Stanford 2023 research experiment demonstrated that models have a significantly higher forgetting rate for information located in the middle of the context than for content at the beginning and end.79% 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.78% 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77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/45067API
curl https://kongchang.com/api/v1/knowledge/claims/45067MCP
get_claim(id=45067)