A Voronoi Cell Formulation for Principled Token Pruning in Late-Interaction Retrieval Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866917476292362240 |
|---|---|
| author | Kankanampati, Yash Zong, Yuxuan Tomeh, Nadi Piwowarski, Benjamin Roux, Joseph Le |
| author_facet | Kankanampati, Yash Zong, Yuxuan Tomeh, Nadi Piwowarski, Benjamin Roux, Joseph Le |
| contents | Late-interaction models such as ColBERT offer competitive performance across various retrieval tasks but require storing a dense embedding for each document token, leading to a substantial index storage overhead. Past works address this by attempting to prune low-importance token embeddings based on statistical and empirical measures, but they often either lack formal grounding or are ineffective. To address these shortcomings, we introduce a framework grounded in hyperspace geometry and cast token pruning as a Voronoi cell estimation problem in the embedding space. By interpreting each token's influence as a measure of its Voronoi region, our approach enables principled pruning that retains retrieval quality while reducing index size. Through our experiments, we demonstrate that this approach serves not only as a competitive pruning strategy but also as a valuable tool for improving and interpreting token-level behavior within dense retrieval systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_09933 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | A Voronoi Cell Formulation for Principled Token Pruning in Late-Interaction Retrieval Models Kankanampati, Yash Zong, Yuxuan Tomeh, Nadi Piwowarski, Benjamin Roux, Joseph Le Information Retrieval H.3.3 Late-interaction models such as ColBERT offer competitive performance across various retrieval tasks but require storing a dense embedding for each document token, leading to a substantial index storage overhead. Past works address this by attempting to prune low-importance token embeddings based on statistical and empirical measures, but they often either lack formal grounding or are ineffective. To address these shortcomings, we introduce a framework grounded in hyperspace geometry and cast token pruning as a Voronoi cell estimation problem in the embedding space. By interpreting each token's influence as a measure of its Voronoi region, our approach enables principled pruning that retains retrieval quality while reducing index size. Through our experiments, we demonstrate that this approach serves not only as a competitive pruning strategy but also as a valuable tool for improving and interpreting token-level behavior within dense retrieval systems. |
| title | A Voronoi Cell Formulation for Principled Token Pruning in Late-Interaction Retrieval Models |
| topic | Information Retrieval H.3.3 |
| url | https://arxiv.org/abs/2603.09933 |