ConceptDense Passage Retrieval / 稠密向量检索

DPR

A dense vector retrieval technique that encodes text into dense vectors and retrieves results by computing semantic similarity. DPR is a core component of Retrieval-Augmented Generation (RAG) systems, and differs from traditional sparse keyword-based retrieval methods such as BM25 and TF-IDF.

Timeline (last 90 days)

Sep 11

Representative models for dense vector retrieval include DPR, BGE, and text-embedding-ada.

Unverified50%

All Facts (2)

Unverified

RAG原始论文使用DPR（Dense Passage Retrieval）作为检索器、BART作为生成器

90%Unverified

Representative models for dense vector retrieval include DPR, BGE, and text-embedding-ada.

50%