DocPruner: A Storage-Efficient Framework for Multi-Vector Visual Document Retrieval via Adaptive Patch-Level Embedding Pruning

ICLR 2026 Conference Submission15588 Authors

Keywords: Visual Document Retrieval, Multi-Vector Retrieval, Multimodal LLMs

TL;DR: DocPruner is a framework that adaptively prunes redundant patch-level embeddings, slashing storage overhead by 50-60% for large-scale visual document retrieval systems while maintaining accuracy.

Abstract:

Visual Document Retrieval (VDR), the task of retrieving visually-rich document pages using queries that combine visual and textual cues, is crucial for numerous real-world applications.
Recent state-of-the-art methods leverage Large Vision-Language Models (LVLMs) in a multi-vector paradigm, representing each document as patch-level embeddings to capture fine-grained details.
While highly effective, this approach introduces a critical challenge: prohibitive storage overhead, as storing hundreds of vectors per page makes large-scale deployment costly and impractical.
To address this, we introduce **DocPruner, the first framework to employ adaptive patch-level embedding pruning for VDR to effectively reduce the storage overhead**.
DocPruner leverages the intra-document patch attention distribution to dynamically identify and discard redundant embeddings for each document.
This adaptive mechanism enables a significant 50-60% reduction in storage for leading multi-vector VDR models with negligible degradation in retrieval performance.
Extensive experiments across more than ten benchmark datasets validate that DocPruner offers a robust, flexible, and effective solution for building storage-efficient, large-scale VDR systems.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 15588