# Towards Storage-Efficient Visual Document Retrieval: An Empirical Study on Reducing Patch-Level Embeddings

arXiv: https://arxiv.org/abs/2506.04997 · HTML: https://mlsanity.com/paper/2506.04997/

## Key facts

- Cited by 26 papers in the map (data through 14 Oct 2026).
- Placed in Visual token pruning in multimodal models, one of 12 directions in Efficient LLM inference and serving.
- Builds on 22 of its 32 references that are in the map.
- First posted on arXiv 5 Jun 2025.

## Placed in

- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

## Builds on (most cited first, 22 of 22)

- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Dense Passage Retrieval for Open-Domain Question Answering](https://mlsanity.com/paper/2004.04906/) (2020)
- [ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning](https://mlsanity.com/paper/2203.10244/) (2022)
- [DocVQA: A Dataset for VQA on Document Images](https://mlsanity.com/paper/2007.00398/) (2020)
- [ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT](https://mlsanity.com/paper/2004.12832/) (2020)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [PaliGemma: A versatile 3B VLM for transfer](https://mlsanity.com/paper/2407.07726/) (2024)
- [ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction](https://mlsanity.com/paper/2112.01488/) (2021)
- [InfographicVQA](https://mlsanity.com/paper/2104.12756/) (2021)
- [Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations](https://mlsanity.com/paper/2202.07800/) (2022)
- [PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction](https://mlsanity.com/paper/2410.17247/) (2024)
- [VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents](https://mlsanity.com/paper/2410.10594/) (2024)
- [MMLongBench-Doc: Benchmarking Long-context Document Understanding with Visualizations](https://mlsanity.com/paper/2407.01523/) (2024)
- [SlideVQA: A Dataset for Document Visual Question Answering on Multiple Images](https://mlsanity.com/paper/2301.04883/) (2023)
- [SPViT: Enabling Faster Vision Transformers via Soft Token Pruning](https://mlsanity.com/paper/2112.13890/) (2021)
- [M3DocRAG: Multi-modal Retrieval is What You Need for Multi-page Multi-document Understanding](https://mlsanity.com/paper/2411.04952/) (2024)
- [Unifying Multimodal Retrieval via Document Screenshot Embedding](https://mlsanity.com/paper/2406.11251/) (2024)
- [Towards Complex Document Understanding By Discrete Reasoning](https://mlsanity.com/paper/2207.11871/) (2022)
- [AntiLeakBench: Preventing Data Contamination by Automatically Constructing Benchmarks with Updated Real-World Knowledge](https://mlsanity.com/paper/2412.13670/) (2024)
- [AKEW: Assessing Knowledge Editing in the Wild](https://mlsanity.com/paper/2402.18909/) (2024)
- [Reducing the Footprint of Multi-Vector Retrieval with Minimal Performance Impact via Token Pooling](https://mlsanity.com/paper/2409.14683/) (2024)
- [Sailing by the Stars: A Survey on Reward Models and Learning Strategies for Learning from Rewards](https://mlsanity.com/paper/2505.02686/) (2025)

## Built on by (most cited first, 26 of 26)

- [DocPruner: A Storage-Efficient Framework for Multi-Vector Visual Document Retrieval via Adaptive Patch-Level Embedding Pruning](https://mlsanity.com/paper/2509.23883/) (2025)
- [Sculpting the Vector Space: Towards Efficient Multi-Vector Visual Document Retrieval via Prune-then-Merge Framework](https://mlsanity.com/paper/2602.19549/) (2026)
- [Scaling Beyond Context: A Survey of Multimodal Retrieval-Augmented Generation for Document Understanding](https://mlsanity.com/paper/2510.15253/) (2025)
- [Unlocking Multimodal Document Intelligence: From Current Triumphs to Future Frontiers of Visual Document Retrieval](https://mlsanity.com/paper/2602.19961/) (2026)
- [CausalEmbed: Auto-Regressive Multi-Vector Generation in Latent Space for Visual Document Embedding](https://mlsanity.com/paper/2601.21262/) (2026)
- [Structural Anchor Pruning: Training-Free Multi-Vector Compression for Visual Document Retrieval](https://mlsanity.com/paper/2601.20107/) (2026)
- [Hybrid-Vector Retrieval for Visually Rich Documents: Combining Single-Vector Efficiency and Multi-Vector Accuracy](https://mlsanity.com/paper/2510.22215/) (2025)
- [Beyond the Grid: Layout-Informed Multi-Vector Retrieval with Parsed Visual Document Representations](https://mlsanity.com/paper/2603.01666/) (2026)
- [HiKEY: Hierarchical Multimodal Retrieval for Open-Domain Document Question Answering](https://mlsanity.com/paper/2605.29606/) (2026)
- [NanoVDR: Distilling a 2B Vision-Language Retriever into a 70M Text-Only Encoder for Visual Document Retrieval](https://mlsanity.com/paper/2603.12824/) (2026)
- [MINER: Mining Multimodal Internal Representation for Efficient Retrieval](https://mlsanity.com/paper/2605.06460/) (2026)
- [Very Efficient Listwise Multimodal Reranking for Long Documents](https://mlsanity.com/paper/2605.11864/) (2026)
- [VikingMem: A Memory Base Management System for Stateful LLM-based Applications](https://mlsanity.com/paper/2605.29640/) (2026)
- [MM-Matryoshka: Towards Budget-Elastic Visual Document Retrieval via a 2D Multimodal Matryoshka Training Framework](https://mlsanity.com/paper/2606.07654/) (2026)
- [Do All Visual Tokens Matter Equally? Object-Evidence Preserving Token Merging for Vision-Language Retrieval](https://mlsanity.com/paper/2607.04605/) (2026)
- [LIR: The First Workshop on Late Interaction and Multi Vector Retrieval @ ECIR 2026](https://mlsanity.com/paper/2511.00444/) (2025)
- [AnchorFold: A Focus-Then-Fold Framework via Recursive Attention Propagation for Efficient Multi-Vector Visual Document Retrieval](https://mlsanity.com/paper/2608.08732/) (2026)
- [XL-DocBench: Benchmarking Evidence-Grounded Extra-Long Document Understanding](https://mlsanity.com/paper/2608.00036/) (2026)
- [KoVRE: Training an Efficient Embedding Model for Korean Visual Document Retrieval](https://mlsanity.com/paper/2608.01389/) (2026)
- [DistilVDR: A Compact End-to-End Visual Document Retriever via Dual-Student Distillation](https://mlsanity.com/paper/2608.10636/) (2026)
- [Spatial Matryoshka Training for Multi-Granularity Visual Document Retrieval](https://mlsanity.com/paper/2608.29951/) (2026)
- [MoEMB: Scaling Universal Multimodal Embeddings with Efficient Mixture-of-Experts Models](https://mlsanity.com/paper/2609.08663/) (2026)
- [MAGIC: Marginal-Guided Compression with Optimal Transport for Efficient Visual Document Retrieval](https://mlsanity.com/paper/2609.21018/) (2026)
- [AdaMerge: Tuning-Free Patch Compression for Multi-Vector Visual Document Retrieval](https://mlsanity.com/paper/2609.22562/) (2026)
- [ColNanoVDR: Document-Free Query Distillation for Multi-Vector Visual Document Retrieval via Optimal Transport](https://mlsanity.com/paper/2609.34899/) (2026)
- [ResComEmb: Effective and Efficient Multimodal Embedding via Residual Homogeneity Compression](https://mlsanity.com/paper/2609.37225/) (2026)

## Abstract (arXiv)

Despite the strong performance of ColPali/ColQwen2 in Visualized Document Retrieval (VDR), it encodes each page into multiple patch-level embeddings and leads to excessive memory usage. This empirical study investigates methods to reduce patch embeddings per page at minimum performance degradation. We evaluate two token-reduction strategies: token pruning and token merging. Regarding token pruning, we surprisingly observe that a simple random strategy outperforms other sophisticated pruning methods, though still far from satisfactory. Further analysis reveals that pruning is inherently unsuitable for VDR as it requires removing certain page embeddings without query-specific information. Turning to token merging (more suitable for VDR), we search for the optimal combinations of merging strategy across three dimensions and develop Light-ColPali/ColQwen2. It maintains 98.2% of retrieval performance with only 11.8% of original memory usage, and preserves 94.6% effectiveness at 2.8% memory footprint. We expect our empirical findings and resulting Light-ColPali/ColQwen2 offer valuable insights and establish a competitive baseline for future research towards efficient VDR.