Unverified50% confidenceBenchmarkExact time
LLMLingua通过语言模型识别并删除不重要的Token，可以在保持95%任务性能的情况下压缩50%的上下文长度
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/5/2026
First Seen
Valid until: 12/4/2026
Sources
Related Entities
Related Claims
UnverifiedLLMLingua等工具通过信息熵分析删除prompt冗余词汇，可将token消耗压缩30-50%77% similarUnverifiedLLMLingua等提示词压缩技术可在保留语义的同时将提示词长度压缩30%-50%77% similarUnverified上下文压缩主流实现方式包括基于LLMLingua等语言模型的语义压缩、滑动窗口截断、摘要化压缩以及基于重要性评分的选择性保留67% similarUnverifiedContext compression technology uses intelligent summarization, redundant information removal, and hierarchical caching to reduce actual Token count fed to models without losing critical information63% similarUnverified通过向量检索结合嵌入模型对代码库预处理索引，智能体只需加载语义相关的代码片段，Token消耗可降低20倍以上63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/860430API
curl https://kongchang.com/api/v1/knowledge/claims/860430MCP
get_claim(id=860430)