Feed Entities Podcasts Deep Dives Developers About

[KongchangAI]

Real-time AI knowledge base. Every claim is traceable to its source, with data freshness under 24 hours.

Powered by 240+ first-hand AI sources

Knowledge

Feed
Hot Entities
Podcasts

Developers

CLI
MCP Server
REST API
About

© 2026 KongchangAI. The freshest, most accurate AI knowledge

Unverified50% confidenceFactExact time

即便是支持128K上下文的模型，在文档长度超过32K时，中段信息的检索准确率往往出现明显下滑

1

Sources

50%

Confidence

Long-term

Relevance

7/17/2026

First Seen

Sources

编程Agent记忆缺失？SSH同步开源方案让AI真正记住你

Related Claims

Unverified文档分块策略中过长则检索精度下降，过短则丢失上下文74% similar Unverified即使是支持128K token的大模型，在实际使用中也面临注意力衰减问题——模型对窗口中间位置的信息关注度明显低于首尾71% similar Unverified即使是128K Token的大模型在处理大量历史时也会出现注意力稀释问题69% similar Unverified即使支持128K或更长上下文的大语言模型，在一次性注入所有能力指令时也会出现注意力稀释（Attention Dilution）问题，导致特定任务执行质量下降69% similar Unverified人类工作记忆的容量约为7±2个信息块（Miller's Law），当系统复杂度超过这个阈值时，纯文本代码的可理解性急剧下降68% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/545095

API

curl https://kongchang.com/api/v1/knowledge/claims/545095

MCP

get_claim(id=545095)