Reproduce “Filtered Approximate Nearest Neighbor Search in Vector Databases: System Design and Performance Analysis” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Abylay Amanbayev, Brian Tsan, Tri Dang, Florin Rusu
The authors compare filtered approximate nearest-neighbor search across FAISS, Milvus, and pgvector, testing HNSW and IVFFlat with scalar inequality filters. Recall is measured against exact filtered neighbors. Milvus separates eligible results from navigation nodes and uses exact search when more than 93% fail; it reports 100% recall at 1–5 ms on typical segments. Results are bounded to in-memory, single-threaded tests.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.