Embeddings & Vector SearchSuhas Jayaram Subramanya et al. (Microsoft Research India, CMU)Microsoft ResearchCarnegie Mellon University (CMU)

Microsoft Research India and CMU present DiskANN

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

Presented DiskANN — an index of 1 billion points on a single workstation with 64 GB of RAM using the Vamana graph, with 5,000+ QPS and under 3 ms latency on SIFT1B

Why it matters

An SSD-based ANN that broke HNSW's constraint of having to hold everything in memory

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

Paperneurips.cchttps://neurips.cc/virtual/2019/poster/14308

Added in Week 35, 2026 · See that week

On the map — ⑰ Embeddings line

  1. Previous stopSentence-BERT
  2. This recordDiskANN
  3. Next stopDPR
See it on the map

Nearby on the same axis

Reimers and Gurevych present Sentence-BERTPaperShen et al. present Towards Backward-Compatible Representation LearningPaperKarpukhin et al. present Dense Passage RetrievalPaperGoogle Research open-sources ScaNNOfficial