LotusFilter: Fast Diverse Nearest Neighbor Search via a Learned Cutoff Table

TL;DR

LotusFilter is a fast post-processing method that diversifies approximate nearest neighbor search results by using a learned cutoff table to efficiently remove redundant vectors, improving diversity without sacrificing speed.

Contribution

It introduces LotusFilter, a novel, fast post-processing module that enhances diversity in ANNS results through a learned cutoff table, suitable for real-world applications.

Findings

Operates at 0.02 ms/query speed.

Effectively diversifies results in RAG-like settings.

Utilizes features like OpenAI embeddings.

Abstract

Approximate nearest neighbor search (ANNS) is an essential building block for applications like RAG but can sometimes yield results that are overly similar to each other. In certain scenarios, search results should be similar to the query and yet diverse. We propose LotusFilter, a post-processing module to diversify ANNS results. We precompute a cutoff table summarizing vectors that are close to each other. During the filtering, LotusFilter greedily looks up the table to delete redundant vectors from the candidates. We demonstrated that the LotusFilter operates fast (0.02 [ms/query]) in settings resembling real-world RAG applications, utilizing features such as OpenAI embeddings. Our code is publicly available at https://github.com/matsui528/lotf.