Visual Late Chunking: An Empirical Study of Contextual Chunking for Efficient Visual Document Retrieval
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908954957709312 |
|---|---|
| author | Yan, Yibo Ou, Mingdong Cao, Yi Huo, Jiahao Zou, Xin Liu, Shuliang Kwok, James Hu, Xuming |
| author_facet | Yan, Yibo Ou, Mingdong Cao, Yi Huo, Jiahao Zou, Xin Liu, Shuliang Kwok, James Hu, Xuming |
| contents | Multi-vector models dominate Visual Document Retrieval (VDR) due to their fine-grained matching capabilities, but their high storage and computational costs present a major barrier to practical deployment. In this paper, we propose ColChunk, a plug-and-play framework that introduces multimodal late chunking to construct efficient, contextualized multi-vectors. Unlike existing pruning or fixed-token approaches, ColChunk employs hierarchical clustering on patch-level embeddings, fused with a 2D position prior to ensure spatial-semantic coherence. This adaptive grouping allows for a content-aware representation that preserves global context while drastically reducing the vector count. Evaluations across 24 VDR datasets demonstrate ColChunk achieves over a 90% reduction in storage requirements while simultaneously delivering a 9-point average improvement in nDCG@5 across representative single-vector models. ColChunk provides a practical solution for balancing retrieval accuracy and efficiency in visual document systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_10167 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Visual Late Chunking: An Empirical Study of Contextual Chunking for Efficient Visual Document Retrieval Yan, Yibo Ou, Mingdong Cao, Yi Huo, Jiahao Zou, Xin Liu, Shuliang Kwok, James Hu, Xuming Computer Vision and Pattern Recognition Computation and Language Information Retrieval Multi-vector models dominate Visual Document Retrieval (VDR) due to their fine-grained matching capabilities, but their high storage and computational costs present a major barrier to practical deployment. In this paper, we propose ColChunk, a plug-and-play framework that introduces multimodal late chunking to construct efficient, contextualized multi-vectors. Unlike existing pruning or fixed-token approaches, ColChunk employs hierarchical clustering on patch-level embeddings, fused with a 2D position prior to ensure spatial-semantic coherence. This adaptive grouping allows for a content-aware representation that preserves global context while drastically reducing the vector count. Evaluations across 24 VDR datasets demonstrate ColChunk achieves over a 90% reduction in storage requirements while simultaneously delivering a 9-point average improvement in nDCG@5 across representative single-vector models. ColChunk provides a practical solution for balancing retrieval accuracy and efficiency in visual document systems. |
| title | Visual Late Chunking: An Empirical Study of Contextual Chunking for Efficient Visual Document Retrieval |
| topic | Computer Vision and Pattern Recognition Computation and Language Information Retrieval |
| url | https://arxiv.org/abs/2604.10167 |