Unverified50% confidenceFactExact time
现代嵌入模型（如sentence-transformers系列）通常输出768维或1536维的向量
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
Building an Offline RAG Application from Scratch: A Complete Guide to a Private PDF Knowledge Base
redditr/learnmachinelearning7/11/2026
Related Claims
Verified嵌入模型将文本转换为高维向量时，通常生成768维或1536维的浮点数数组77% similarUnverified嵌入模型将文本转化为高维稠密向量（通常为768至4096维），通过近似最近邻算法（ANN，如HNSW或IVF-PQ）快速检索语义相似内容63% similarUnverified主流嵌入模型如all-MiniLM-L6-v2、BGE系列、E5系列通常基于BERT架构，将文本编码为384至1536维的稠密浮点向量62% similarUnverified上下文窗口决定模型单次能处理的文本量：8K约6000词，32K约24000词，128K可处理整本书内容62% similarVerifiedEmbedding models (such as OpenAI's text-embedding-ada-002 or the open-source BGE series) convert text into high-dimensional vectors, typically 768 or 1536 dimensions.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/490389API
curl https://kongchang.com/api/v1/knowledge/claims/490389MCP
get_claim(id=490389)