Reproduce “Recast: Refinement-Centered Approximate Nearest Neighbor Search Optimized for Storage” 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
Zehai Yang, Fengyi Tao, Wenzhuo Dong, Shimin Chen
Recast moves the first pass of large-scale nearest-neighbor search into memory, storing compact RecastQ codes while keeping exact vectors on disk. Its QUIC method uses uncertainty intervals to select candidates for exact refinement, and its mutable IVF layout supports updates. At Recall@10 of 0.99, Recast achieved 1.56–2.01 times baseline QPS and read 1.73–4.10 times fewer pages across tested 100-million-vector datasets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.