# How does Watermarking Affect Visual Language Models in Document Understanding?

arXiv: https://arxiv.org/abs/2504.01048 · HTML: https://mlsanity.com/paper/2504.01048/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 17 of its 25 references that are in the map.
- First posted on arXiv 1 Apr 2025.

## Placed in

- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)
- LLM safety and security / [Watermarking generated content and models](https://mlsanity.com/safety/watermarking-generated-content-and-models/)

## Builds on (most cited first, 17 of 17)

- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling](https://mlsanity.com/paper/2412.05271/) (2024)
- [mPLUG-Owl: Modularization Empowers Large Language Models with Multimodality](https://mlsanity.com/paper/2304.14178/) (2023)
- [On Evaluating Adversarial Robustness of Large Vision-Language Models](https://mlsanity.com/paper/2305.16934/) (2023)
- [PromptRobust: Towards Evaluating the Robustness of Large Language Models on Adversarial Prompts](https://mlsanity.com/paper/2306.04528/) (2023)
- [TextMonkey: An OCR-Free Large Multimodal Model for Understanding Document](https://mlsanity.com/paper/2403.04473/) (2024)
- [mPLUG-DocOwl: Modularized Multimodal Large Language Model for Document Understanding](https://mlsanity.com/paper/2307.02499/) (2023)
- [mPLUG-DocOwl2: High-resolution Compressing for OCR-free Multi-page Document Understanding](https://mlsanity.com/paper/2409.03420/) (2024)
- [BROS: A Pre-trained Language Model Focusing on Text and Layout for Better Key Information Extraction from Documents](https://mlsanity.com/paper/2108.04539/) (2021)
- [Spatial Dependency Parsing for Semi-Structured Document Information Extraction](https://mlsanity.com/paper/2005.00642/) (2020)
- [Evaluating the Instruction-Following Robustness of Large Language Models to Prompt Injection](https://mlsanity.com/paper/2308.10819/) (2023)
- [VisDoM: Multi-Document QA with Visually Rich Elements Using Multimodal Retrieval-Augmented Generation](https://mlsanity.com/paper/2412.10704/) (2024)
- [Enhancing Visual Document Understanding with Contrastive Learning in Large Visual-Language Models](https://mlsanity.com/paper/2402.19014/) (2024)
- ["Not Aligned" is Not "Malicious": Being Careful about Hallucinations of Large Language Models' Jailbreak](https://mlsanity.com/paper/2406.11668/) (2024)
- [Benchmarking Robustness of Multimodal Image-Text Models under Distribution Shift](https://mlsanity.com/paper/2212.08044/) (2022)
- [Mitigating Lost-in-Retrieval Problems in Retrieval Augmented Multi-Hop Question Answering](https://mlsanity.com/paper/2502.14245/) (2025)
- [MVTamperBench: Evaluating Robustness of Vision-Language Models](https://mlsanity.com/paper/2412.19794/) (2024)

## Built on by (most cited first, 2 of 2)

- [Gated Differentiable Working Memory for Long-Context Language Modeling](https://mlsanity.com/paper/2601.12906/) (2026)
- [Visually Legible but Adversarial: Watermark-Constrained Attacks on End-to-End VLMs for Document Understanding](https://mlsanity.com/paper/2610.15728/) (2026)

## Abstract (arXiv)

Visual Language Models (VLMs) have become foundational models for document understanding tasks, widely used in the processing of complex multimodal documents across domains such as finance, law, and academia. However, documents often contain noise-like information, such as watermarks, which inevitably leads us to inquire: Do watermarks degrade the performance of VLMs in document understanding? To address this, we propose a novel evaluation framework to investigate the effect of visible watermarks on VLMs performance. We takes into account various factors, including different types of document data, the positions of watermarks within documents and variations in watermark content. Our experimental results reveal that VLMs performance can be significantly compromised by watermarks, with performance drop rates reaching up to 36%. We discover that scattered watermarks cause stronger interference than centralized ones, and that semantic contents in watermarks creates greater disruption than simple visual occlusion. Through attention mechanism analysis and embedding similarity examination, we find that the performance drops are mainly attributed to that watermarks 1) force widespread attention redistribution, and 2) alter semantic representation in the embedding space. Our research not only highlights significant challenges in deploying VLMs for document understanding, but also provides insights towards developing robust inference mechanisms on watermarked documents.