Verified65% confidenceBenchmarkExact time
Research has found that large models perform best when key information is located at the beginning or end of the context, and performance drops significantly when it is in the middle, with accuracy declining by more than 20 percentage points on certain tasks.
3
Sources
65%
Confidence
Long-term
Relevance
8/23/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen the context approaches its limit, the model may forget critical early information or produce lower-quality output.77% similarVerifiedOn convergent tasks, there is a clear performance saturation phenomenon after models reach a certain scale, where continued increases in parameter count or training data yield rapidly diminishing marginal contributions to accuracy.76% similarUnverifiedLearning-based methods can achieve very strong performance when annotated data is sufficient, but their robustness degrades significantly when facing domain shift.75% similarUnverifiedDSA稀疏注意力不仅更快，在某些任务上甚至更准，因为跳过无关token让模型更专注于有效信息75% similarUnverified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74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/791386API
curl https://kongchang.com/api/v1/knowledge/claims/791386MCP
get_claim(id=791386)