Unverified50% confidenceBenchmarkExact time
斯坦福大学2023年研究显示当关键信息位于超长文档中间段落时模型检索准确率显著低于位于开头或结尾
1
Sources
50%
Confidence
Long-term
Relevance
7/24/2026
First Seen
Sources
Related Claims
VerifiedA 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.85% 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83% similarPartially VerifiedA 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).77% similarUnverifiedIn 2023, Nelson Liu and others from Stanford University argued in the paper "Lost in the Middle" that model performance is significantly lower when relevant information is located in the middle of a long context compared to when it is at the beginning or end, forming a U-shaped performance curve.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/605004API
curl https://kongchang.com/api/v1/knowledge/claims/605004MCP
get_claim(id=605004)