Get Started
Home
Topics
Search
Library
Research questionHow can private outsourced vector retrieval scale to millions of documents without unacceptable latency or search-quality loss?Outsourcing a dense index can expose both proprietary documents and user queries, while privacy-preserving search must process corpus-scale state. At million-document scale, the resulting computation and communication can make retrieval too slow or reduce the quality of retrieved candidates.
AI
Information Retrieval
Machine Learning
Retrieval-Augmented Generation
Technology
Latest papersRecent research connected to this question, newest first.Spruce: Scalable Private Outsourced Retrieval Using Compact EmbeddingsThe source studies RAG-style dense retrieval with outsourced indexes, using two-server MPC, compact binary representations, private pruning, and an owner-operated dealer. Reported results cover four corpora ranging from 383K to 5.42M documents, with latency, throughput, communication, and retrieval-quality measurements under specified bandwidths.research paper · Sep 3, 2026
Related questions
How can vector databases protect document embeddings from inversion while preserving retrieval utility for RAG?How can federated learning personalize regional query understanding while preserving privacy and avoiding transformer training failures?How can dense retrievers specialize for biomedical search without sacrificing general-domain retrieval?How can visual-document RAG adapt page retrieval to each query without hurting answer accuracy?