Paper: arXiv 2507.18577

Authors: Liyuan Chen, Shuoling Liu, Jiangpeng Yan, Xiaoyu Wang, Henglin Liu, Chuang Li, Kecheng Jiao, Jixuan Ying, Yang Veronica Liu, Qiang Yang, Xiu Li

Abstract

The advent of foundation models (FMs), large-scale pre-trained models with strong generalization capabilities, has opened new frontiers for financial engineering. While general-purpose FMs such as GPT-4 and Gemini have demonstrated promising performance in tasks ranging from financial report summarization to sentiment-aware forecasting, many financial applications remain constrained by unique domain requirements such as multimodal reasoning, regulatory compliance, and data privacy. These challenges have spurred the emergence of financial foundation models (FFMs): a new class of models explicitly designed for finance. This survey presents a comprehensive overview of FFMs, with a taxonomy spanning three key modalities: financial language foundation models (FinLFMs), financial time-series foundation models (FinTSFMs), and financial visual-language foundation models (FinVLFMs). We review their architectures, training methodologies, datasets, and real-world applications. Furthermore, we identify critical challenges in data availability, algorithmic scalability, and infrastructure constraints and offer insights into future research opportunities. We hope this survey can serve as both a comprehensive reference for understanding FFMs and a practical roadmap for future innovation.

Complexity vs Empirical Score

  • Math Complexity: 4.0/10
  • Empirical Rigor: 3.0/10
  • Quadrant: Philosophers — conceptual discussion, lower math and empirical scores

Why this score: The paper is a survey that categorizes financial foundation models and discusses high-level architectures and challenges, but contains no mathematical derivations, formulas, or statistical metrics. It lacks specific implementation details, backtesting results, or dataset specifics, focusing instead on a conceptual overview and taxonomy.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>&quot;Advancing Financial Engineering with<br>Foundation Models: Progress, Applications, and Challenges&quot;"] --> B
  subgraph B ["Methodology: Comprehensive Survey & Taxonomy"]
      direction LR
      B1["Financial Language<br>Foundation Models"] --> B2["Financial Time-Series<br>Foundation Models"] --> B3["Financial Visual-Language<br>Foundation Models"]
  end
  B --> C["Computational Processes &<br>Real-World Applications"]
  C --> D["Key Findings & Outcomes:<br>1. Scalability & Generalization achieved via FFMs<br>2. Unique Challenges: Data scarcity, Regulation, Privacy<br>3. Future Roadmap: Multimodal reasoning & Infrastructure"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

We propose a novel taxonomy of financial foundation models and conduct a comprehensive review of existing methods across language, time-series, and visual-language modalities.

Source sentence
We propose a novel taxonomy of FFMs and conduct a comprehensive review of existing methods across the three major sub-fields: language (Section 2), time-series (Section 3), and visual-language (Section 4).

Assumes

No explicit assumptions found in the text.

Data

No dataset named in the text.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • Critical challenges are identified in data availability, algorithmic scalability, and infrastructure constraints. other
    Source sentence
    Furthermore, we identify critical challenges in data availability, algorithmic scalability, and infrastructure constraints and offer insights into future research opportunities.
  • Financial foundation models remain in an early yet rapidly evolving stage, with most resources limited in scale or realism, hindering assessment of long-horizon and multimodal capabilities. other
    Source sentence
    Summary FinTSFMs are still in an early yet rapidly evolving stage. While new datasets such as FNSPID [94] and FinTSB [95] broaden the evaluation scope, most resources remain limited in scale or realism, hindering the assessment of long-horizon and multimodal capabilities.
  • New datasets such as FNSPID and FinTSB broaden evaluation scope but most resources remain limited in scale or realism. other
    Source sentence
    While new datasets such as FNSPID [94] and FinTSB [95] broaden the evaluation scope, most resources remain limited in scale or realism, hindering the assessment of long-horizon and multimodal capabilities.

Checked by others

No other paper in the archive has been found to support or contradict these findings. Coverage is partial, so this is neither replication nor refutation.

Code

https://github.com/chengjunyan1/SocioDojo — the authors' own repository

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.