Embeddings & Vector SearchVladimir Karpukhin et al. (FAIR, Princeton, University of Washington)FAIRPrinceton UniversityUniversity of Washington

Karpukhin et al. present Dense Passage Retrieval

Presented Dense Passage Retrieval — absolute gains of 9–19 percentage points over BM25 in top-20 retrieval accuracy

Why it matters

Flipped the default from sparse retrieval (BM25) to dense retrieval

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2004.04906

Added in Week 35, 2026 · See that week

Nearby on the same axis

Shen et al. present Towards Backward-Compatible Representation LearningPaperGoogle Research open-sources ScaNNOfficialMicrosoft Research India and CMU present DiskANNOfficialReimers and Gurevych present Sentence-BERTPaper