Unverified50% confidenceFactExact time
RAG嵌入模型将文本片段转化为高维稠密向量，通常为768至1536维
1
Sources
50%
Confidence
Long-term
Relevance
7/17/2026
First Seen
Sources
Related Claims
Unverified嵌入模型通常只有几百MB大小，将文本转化为768维（或其他维度）的向量76% similarUnverified上下文窗口是大模型单次能够处理的最大信息量，以token为单位计算，现代顶级模型的上下文窗口已从早期的4K token扩展至200K token甚至更高63% similarVerifiedCurrent mainstream context window lengths across mainstream models range from 8K to 128K tokens62% similarVerifiedRAG实践中chunk_size通常设置在256至1024 token之间，并保留约10%至20%的重叠（Overlap）61% similarUnverified不同大模型在上下文窗口大小方面存在显著差异，从8K到128K tokens不等61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/544365API
curl https://kongchang.com/api/v1/knowledge/claims/544365MCP
get_claim(id=544365)