返回首页
RadarAI··论文与技术

LEANN: 97% Storage Reduction for Vector Libraries via Graph Structure

中文摘要

LEANN利用图结构而非完整嵌入,将向量存储减少97%,支持在设备端对文件、邮件及代码进行本地索引与RAG应用。

English Summary

LEANN reduces vector storage by 97% using graph structures, enabling efficient on-device RAG with local indexing for files, emails, chats, and code.

原文节选

📌 一句话摘要 LEANN reduces vector storage by 97% using graph structures instead of storing all embeddings, enabling on-device RAG with local indexing for files, emails, chats, and code. 📝 详细摘要 LEANN solves the issue ...