# New private dense retrieval method adds 0.38 s to RAG pipelines

> Researchers show a hashing‑based filter can keep queries private while only adding 6.5% latency to large‑scale retrieval‑augmented generation.

Oossa · 2026-11-13 · https://oossa.com/en/new-private-dense-retrieval-method-adds-0-38-s-to-rag-pipelines

Peichun Hua and co‑authors released a technique for private dense retrieval that uses a learned binary hash to pick a short candidate list. The shortlist is then reranked with encryption, hiding both the query and the final documents. On a 2.68 million‑passage dataset, the method adds just 0.38 seconds to a Qwen3‑32B RAG run over a 10‑Gbps link, matching full‑corpus quality with only 200‑500 candidates.

## The facts

- Paper published on 2026-11-13
- Adds 0.38 seconds (6.5%) latency to a 128‑token Qwen3‑32B RAG pipeline

## Why it matters

Enterprises can deploy private search over large document stores without noticeable slowdown.

## Sources & references

1. [Pointing the Way, Hiding the Destination: Practical Private Dense Retrieval at Scale](https://arxiv.org/abs/2608.25735) – arXiv cs.IR, 2026-11-13

Last updated: 2026-11-13