UniFinEval: Towards Unified Evaluation of Financial Multimodal Models across Text, Images and Videos

Paper: arXiv 2601.22162

Authors: Zhi Yang, Lingfeng Zeng, Fangqi Lou, Qi Qi, Wei Zhang, Zhenyu Wu, Zhenxiong Yu, Jun Han, Zhiheng Jin, Lejie Zhang, Xiaoming Huang, Xiaolong Liang, Zheng Wei, Junbo Zou, Dongpo Cheng, Zhaowei Liu, Xin Guo, Rongjunchen Zhang, Liwen Zhang

Abstract

Multimodal large language models are playing an increasingly significant role in empowering the financial domain, however, the challenges they face, such as multimodal and high-density information and cross-modal multi-hop reasoning, go beyond the evaluation scope of existing multimodal benchmarks. To address this gap, we propose UniFinEval, the first unified multimodal benchmark designed for high-information-density financial environments, covering text, images, and videos. UniFinEval systematically constructs five core financial scenarios grounded in real-world financial systems: Financial Statement Auditing, Company Fundamental Reasoning, Industry Trend Insights, Financial Risk Sensing, and Asset Allocation Analysis. We manually construct a high-quality dataset consisting of 3,767 question-answer pairs in both chinese and english and systematically evaluate 10 mainstream MLLMs under Zero-Shot and CoT settings. Results show that Gemini-3-pro-preview achieves the best overall performance, yet still exhibits a substantial gap compared to financial experts. Further error analysis reveals systematic deficiencies in current models. UniFinEval aims to provide a systematic assessment of MLLMs’ capabilities in fine-grained, high-information-density financial environments, thereby enhancing the robustness of MLLMs applications in real-world financial scenarios. Data and code are available at https://github.com/aifinlab/UniFinEval.

Complexity vs Empirical Score

• Math Complexity: 2.0/10
• Empirical Rigor: 6.0/10 (capped from 8.0: no out-of-sample test; see penalty flags)
• Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper introduces UniFinEval, a novel multimodal benchmark for financial MLLMs spanning text, images, and videos with high information density. While mathematical complexity is minimal, the empirical rigor is strong, featuring 3,767 expert-curated bilingual QA pairs evaluated across 10 mainstream models with publicly available code and data.

Research Flowchart

  flowchart TD
    A[Research Goal: Evaluate MLLMs in High-Density Financial Multimodal Scenarios] --> B[Dataset Construction: 3,767 EN/ZH QA Pairs across Text, Images & Videos]
    B --> C[Grounding in 5 Core Scenarios: Auditing, Fundamentals, Trends, Risk & Asset Allocation]
    C --> D[Evaluation: 10 Mainstream MLLMs tested under Zero-Shot & CoT Settings]
    D --> E[Error Analysis: Systematic Deficiencies in Cross-Modal & Multi-Hop Reasoning]
    E --> F[Key Outcome: Gemini-3-pro-preview Leads but Significant Gap to Experts Remains]

Cited References (102)

1. The Theory of Parsing, Translation and Compiling (1972)
2. Publications Manual (1983)
3. Alternation (1981) — link
4. Scalable training of L1 -regularized log-linear models (2007)
5. Algorithms on Strings, Trees and Sequences (1997)
6. Yara Parser: A Fast and Accurate Dependency Parser (2015) — link
7. Finmmr: make financial numerical reasoning more multimodal, comprehensive, and challenging (2025)
8. Llm-based data science agents: A survey of capabilities, challenges, and future directions (2025) — link

13 of 102 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.