Understanding Financial Reasoning in AI: A Multimodal Benchmark and Error Learning Approach

TL;DR

This paper presents a multimodal benchmark for financial reasoning in AI, combining textual and visual data, and introduces an error-aware learning approach that improves model performance without fine-tuning.

Contribution

It introduces a new multimodal financial reasoning benchmark and proposes an error-aware learning framework that leverages model mistakes to enhance reasoning capabilities.

Findings

Multimodal inputs significantly improve reasoning performance.

Error feedback leads to consistent performance improvements.

Challenges remain in visual understanding and mathematical logic.

Abstract

Effective financial reasoning demands not only textual understanding but also the ability to interpret complex visual data such as charts, tables, and trend graphs. This paper introduces a new benchmark designed to evaluate how well AI models - especially large language and multimodal models - reason in finance-specific contexts. Covering 3,200 expert-level question-answer pairs across 15 core financial topics, the benchmark integrates both textual and visual modalities to reflect authentic analytical challenges in finance. To address limitations in current reasoning approaches, we propose an error-aware learning framework that leverages historical model mistakes and feedback to guide inference, without requiring fine-tuning. Our experiments across state-of-the-art models show that multimodal inputs significantly enhance performance and that incorporating error feedback leads to consistent and measurable improvements. The results highlight persistent challenges in visual understanding and mathematical logic, while also demonstrating the promise of self-reflective reasoning in financial AI systems. Our code and data can be found at https://anonymous/FinMR/CodeData.