Efficient Multivector Retrieval with Token-Aware Clustering and Hierarchical Indexing
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918476401082368 |
|---|---|
| author | Martinico, Silvio Nardini, Franco Maria Rulli, Cosimo Venturini, Rossano |
| author_facet | Martinico, Silvio Nardini, Franco Maria Rulli, Cosimo Venturini, Rossano |
| contents | Multivector retrieval models achieve state-of-the-art effectiveness through fine-grained token-level representations, but their deployment incurs substantial computational and memory costs. Current solutions, based on the well-known k-means clustering algorithm, group similar vectors together to enable both effective compression and efficient retrieval. However, standard k-means scales poorly with the number of clusters and dataset size, and favours frequent tokens during training while underrepresenting rare, discriminative ones. In this work, we introduce TACHIOM, a multivector retrieval system that exploits token-level structure to significantly accelerate both clustering and retrieval. By accounting for tokens' distribution during centroid allocation, TACHIOM easily scales to millions of centroids, enabling highly accurate document scoring using only centroids, avoiding expensive token-level computation. TACHIOM combines a graph-based index over centroids with an optimized Product Quantization layout for efficient final scoring. Experiments on MS-MARCOv1 and LoTTE show that TACHIOM achieves up to $247\times$ faster clustering than k-means and up to $9.8\times$ retrieval speedup over state-of-the-art systems while maintaining comparable or superior effectiveness. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_28142 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Efficient Multivector Retrieval with Token-Aware Clustering and Hierarchical Indexing Martinico, Silvio Nardini, Franco Maria Rulli, Cosimo Venturini, Rossano Information Retrieval Machine Learning Multivector retrieval models achieve state-of-the-art effectiveness through fine-grained token-level representations, but their deployment incurs substantial computational and memory costs. Current solutions, based on the well-known k-means clustering algorithm, group similar vectors together to enable both effective compression and efficient retrieval. However, standard k-means scales poorly with the number of clusters and dataset size, and favours frequent tokens during training while underrepresenting rare, discriminative ones. In this work, we introduce TACHIOM, a multivector retrieval system that exploits token-level structure to significantly accelerate both clustering and retrieval. By accounting for tokens' distribution during centroid allocation, TACHIOM easily scales to millions of centroids, enabling highly accurate document scoring using only centroids, avoiding expensive token-level computation. TACHIOM combines a graph-based index over centroids with an optimized Product Quantization layout for efficient final scoring. Experiments on MS-MARCOv1 and LoTTE show that TACHIOM achieves up to $247\times$ faster clustering than k-means and up to $9.8\times$ retrieval speedup over state-of-the-art systems while maintaining comparable or superior effectiveness. |
| title | Efficient Multivector Retrieval with Token-Aware Clustering and Hierarchical Indexing |
| topic | Information Retrieval Machine Learning |
| url | https://arxiv.org/abs/2604.28142 |