No More K-means: Single-Stage Sparse Coding for Efficient Multi-Vector Retrieval

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Guo, Lixuan, Wang, Yifei, Wen, Tiansheng, Feng, Aosong, Jegelka, Stefanie, You, Chenyu
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913171960233984
author Guo, Lixuan
Wang, Yifei
Wen, Tiansheng
Feng, Aosong
Jegelka, Stefanie
You, Chenyu
author_facet Guo, Lixuan
Wang, Yifei
Wen, Tiansheng
Feng, Aosong
Jegelka, Stefanie
You, Chenyu
contents Multi-vector retrieval (MVR) models, exemplified by ColBERT, have established new benchmarks in retrieval accuracy by preserving fine-grained token-level interactions. However, this granularity imposes prohibitive storage and retrieval efficiency bottlenecks: to manage the immense memory footprint and computational overhead of billion-scale token vectors, state-of-the-art systems are forced to rely on aggressive dimension reduction and complex clustering (e.g., K-means). This compromise introduces two critical limitations: excessive indexing latency of clustering large-scale corpora and semantic information loss inherent to compression. In this paper, we propose Single-stage Sparse Retrieval (SSR}, a paradigm shift that replaces expensive clustering with efficient sparse coding. Instead of compressing features into low-dimensional dense vectors, we utilize Sparse Autoencoder (SAE) to project token embeddings into a high-dimensional but highly sparse representation. This transformation enables us to bypass vector clustering entirely and leverage inverted indexing for precise, high-throughput retrieval. Extensive experiments on the BEIR benchmark demonstrate that SSR achieves a "trifecta" of improvements: it reduces indexing time by 15x compared to ColBERTv2, halves retrieval latency, and simultaneously improves retrieval performance over leading baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2605_30120
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle No More K-means: Single-Stage Sparse Coding for Efficient Multi-Vector Retrieval
Guo, Lixuan
Wang, Yifei
Wen, Tiansheng
Feng, Aosong
Jegelka, Stefanie
You, Chenyu
Information Retrieval
Artificial Intelligence
Machine Learning
Multi-vector retrieval (MVR) models, exemplified by ColBERT, have established new benchmarks in retrieval accuracy by preserving fine-grained token-level interactions. However, this granularity imposes prohibitive storage and retrieval efficiency bottlenecks: to manage the immense memory footprint and computational overhead of billion-scale token vectors, state-of-the-art systems are forced to rely on aggressive dimension reduction and complex clustering (e.g., K-means). This compromise introduces two critical limitations: excessive indexing latency of clustering large-scale corpora and semantic information loss inherent to compression. In this paper, we propose Single-stage Sparse Retrieval (SSR}, a paradigm shift that replaces expensive clustering with efficient sparse coding. Instead of compressing features into low-dimensional dense vectors, we utilize Sparse Autoencoder (SAE) to project token embeddings into a high-dimensional but highly sparse representation. This transformation enables us to bypass vector clustering entirely and leverage inverted indexing for precise, high-throughput retrieval. Extensive experiments on the BEIR benchmark demonstrate that SSR achieves a "trifecta" of improvements: it reduces indexing time by 15x compared to ColBERTv2, halves retrieval latency, and simultaneously improves retrieval performance over leading baselines.
title No More K-means: Single-Stage Sparse Coding for Efficient Multi-Vector Retrieval
topic Information Retrieval
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2605.30120