Document Haystack: A Long Context Multimodal Image/Document Understanding Vision LLM Benchmark

TL;DR

Document Haystack is a new benchmark that evaluates vision language models on long, complex documents with multimodal content, addressing a key gap in multimodal AI research.

Contribution

It introduces a comprehensive, automated benchmark with diverse long documents and embedded multimodal elements to assess VLMs' retrieval and understanding capabilities.

Findings

Prominent VLMs show limited performance on long documents

Benchmark reveals challenges in multimodal retrieval at scale

Provides a standardized platform for future research

Abstract

The proliferation of multimodal Large Language Models has significantly advanced the ability to analyze and understand complex data inputs from different modalities. However, the processing of long documents remains under-explored, largely due to a lack of suitable benchmarks. To address this, we introduce Document Haystack, a comprehensive benchmark designed to evaluate the performance of Vision Language Models (VLMs) on long, visually complex documents. Document Haystack features documents ranging from 5 to 200 pages and strategically inserts pure text or multimodal text+image "needles" at various depths within the documents to challenge VLMs' retrieval capabilities. Comprising 400 document variants and a total of 8,250 questions, it is supported by an objective, automated evaluation framework. We detail the construction and characteristics of the Document Haystack dataset, present results from prominent VLMs and discuss potential research avenues in this area.