WildDoc: How Far Are We from Achieving Comprehensive and Robust Document Understanding in the Wild?

TL;DR

WildDoc introduces a new benchmark for evaluating document understanding in real-world conditions, revealing significant robustness gaps in current multimodal models when faced with natural environment challenges.

Contribution

This paper presents WildDoc, the first benchmark specifically designed to assess the robustness of document understanding models in natural, real-world scenarios.

Findings

State-of-the-art MLLMs show substantial performance drops on WildDoc.

Current models lack robustness to real-world document distortions.

WildDoc highlights the gap between digital and real-world document understanding.

Abstract

The rapid advancements in Multimodal Large Language Models (MLLMs) have significantly enhanced capabilities in Document Understanding. However, prevailing benchmarks like DocVQA and ChartQA predominantly comprise \textit{scanned or digital} documents, inadequately reflecting the intricate challenges posed by diverse real-world scenarios, such as variable illumination and physical distortions. This paper introduces WildDoc, the inaugural benchmark designed specifically for assessing document understanding in natural environments. WildDoc incorporates a diverse set of manually captured document images reflecting real-world conditions and leverages document sources from established benchmarks to facilitate comprehensive comparisons with digital or scanned documents. Further, to rigorously evaluate model robustness, each document is captured four times under different conditions. Evaluations of state-of-the-art MLLMs on WildDoc expose substantial performance declines and underscore the models' inadequate robustness compared to traditional benchmarks, highlighting the unique challenges posed by real-world document understanding. Our project homepage is available at https://bytedance.github.io/WildDoc.