FinMMDocR: A New Benchmark for Financial Multimodal Reasoning

Published:Dec 31, 2025 15:00
1 min read
ArXiv

Analysis

This paper introduces FinMMDocR, a new benchmark designed to evaluate multimodal large language models (MLLMs) on complex financial reasoning tasks. The benchmark's key contributions are its focus on scenario awareness, document understanding (with extensive document breadth and depth), and multi-step computation, making it more challenging and realistic than existing benchmarks. The low accuracy of the best-performing MLLM (58.0%) highlights the difficulty of the task and the potential for future research.

Reference

The best-performing MLLM achieves only 58.0% accuracy.