Can Multi-modal (reasoning) LLMs detect document manipulation?

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liang, Zisheng, Zewde, Kidus, Singh, Rudra Pratap, Patil, Disha, Chen, Zexi, Xue, Jiayu, Yao, Yao, Chen, Yifei, Liu, Qinzhe, Ren, Simiao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912538217676800
author Liang, Zisheng
Zewde, Kidus
Singh, Rudra Pratap
Patil, Disha
Chen, Zexi
Xue, Jiayu
Yao, Yao
Chen, Yifei
Liu, Qinzhe
Ren, Simiao
author_facet Liang, Zisheng
Zewde, Kidus
Singh, Rudra Pratap
Patil, Disha
Chen, Zexi
Xue, Jiayu
Yao, Yao
Chen, Yifei
Liu, Qinzhe
Ren, Simiao
contents Document fraud poses a significant threat to industries reliant on secure and verifiable documentation, necessitating robust detection mechanisms. This study investigates the efficacy of state-of-the-art multi-modal large language models (LLMs)-including OpenAI O1, OpenAI 4o, Gemini Flash (thinking), Deepseek Janus, Grok, Llama 3.2 and 4, Qwen 2 and 2.5 VL, Mistral Pixtral, and Claude 3.5 and 3.7 Sonnet-in detecting fraudulent documents. We benchmark these models against each other and prior work on document fraud detection techniques using a standard dataset with real transactional documents. Through prompt optimization and detailed analysis of the models' reasoning processes, we evaluate their ability to identify subtle indicators of fraud, such as tampered text, misaligned formatting, and inconsistent transactional sums. Our results reveal that top-performing multi-modal LLMs demonstrate superior zero-shot generalization, outperforming conventional methods on out-of-distribution datasets, while several vision LLMs exhibit inconsistent or subpar performance. Notably, model size and advanced reasoning capabilities show limited correlation with detection accuracy, suggesting task-specific fine-tuning is critical. This study underscores the potential of multi-modal LLMs in enhancing document fraud detection systems and provides a foundation for future research into interpretable and scalable fraud mitigation strategies.
format Preprint
id arxiv_https___arxiv_org_abs_2508_11021
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Can Multi-modal (reasoning) LLMs detect document manipulation?
Liang, Zisheng
Zewde, Kidus
Singh, Rudra Pratap
Patil, Disha
Chen, Zexi
Xue, Jiayu
Yao, Yao
Chen, Yifei
Liu, Qinzhe
Ren, Simiao
Computer Vision and Pattern Recognition
Computation and Language
Document fraud poses a significant threat to industries reliant on secure and verifiable documentation, necessitating robust detection mechanisms. This study investigates the efficacy of state-of-the-art multi-modal large language models (LLMs)-including OpenAI O1, OpenAI 4o, Gemini Flash (thinking), Deepseek Janus, Grok, Llama 3.2 and 4, Qwen 2 and 2.5 VL, Mistral Pixtral, and Claude 3.5 and 3.7 Sonnet-in detecting fraudulent documents. We benchmark these models against each other and prior work on document fraud detection techniques using a standard dataset with real transactional documents. Through prompt optimization and detailed analysis of the models' reasoning processes, we evaluate their ability to identify subtle indicators of fraud, such as tampered text, misaligned formatting, and inconsistent transactional sums. Our results reveal that top-performing multi-modal LLMs demonstrate superior zero-shot generalization, outperforming conventional methods on out-of-distribution datasets, while several vision LLMs exhibit inconsistent or subpar performance. Notably, model size and advanced reasoning capabilities show limited correlation with detection accuracy, suggesting task-specific fine-tuning is critical. This study underscores the potential of multi-modal LLMs in enhancing document fraud detection systems and provides a foundation for future research into interpretable and scalable fraud mitigation strategies.
title Can Multi-modal (reasoning) LLMs detect document manipulation?
topic Computer Vision and Pattern Recognition
Computation and Language
url https://arxiv.org/abs/2508.11021