FinDocMRE: A Benchmark for Document-Level Financial Multimodal Reasoning Evaluation

2026-05-18

AI Summary

This work addresses the limitations of existing large multimodal models in effectively integrating text, tables, and images for understanding complex financial documents, as well as the absence of a comprehensive evaluation benchmark tailored to this domain. The authors propose the first vision-centric, document-level multimodal reasoning benchmark spanning 12 financial sectors, comprising 2,878 reports and 12,207 samples, and supporting five distinct tasks. Constructed via a semi-automated pipeline with expert validation, the benchmark mitigates textual bias and ensures high annotation quality. Evaluations of 11 state-of-the-art multimodal large models reveal that none surpasses a 65% average score, with particularly poor performance in numerical estimation and cross-page visual grounding, highlighting critical bottlenecks in current models’ ability to reason over complex financial documents.

Abstract

While Large Multimodal Models (LMMs) excel in general visual tasks, their deployment in specialized financial contexts remains insufficient. Existing benchmarks prioritize isolated charts, often overlooking the need to integrate data from text, tables, and images within comprehensive financial documents. To address this limitation, we introduce FINDOCMRE, a multi-image document-level benchmark designed for financial multimodal reasoning. We construct the dataset via a semi-automated pipeline that combines Visual-Centric Generation with Expert Verification, thereby minimizing text bias and ensuring high annotation quality. Spanning twelve domains, the benchmark comprises 12,207 samples derived from 2,878 financial reports, designed to evaluate multi-image processing and document-level understanding across five distinct task types. Extensive experiments with eleven representative LMMs reveal that no model surpasses an overall score of 65, highlighting challenges in integrating visual grounding with logical reasoning within complex document environments. Specifically, we observe a significant performance divergence across tasks, where models exhibit proficiency in semantic narrative construction but struggle with numerical estimation and cross-page visual grounding. FINDOCMRE serves as a rigorous benchmark to guide the evolution of financial LMMs towards expert-level document analysis and reasoning.

Problem

• financial multimodal reasoning
• document-level understanding
• benchmark
• large multimodal models
• multi-image integration

Innovation

• document-level reasoning
• financial multimodal benchmark
• multi-image understanding
• visual grounding
• expert-verified dataset