Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Multimodal Foundation Agent for Financial Trading: Tool-Augmented, Diversified, and Generalist

Wentao Zhang, Lingxuan Zhao, Haochong Xia, Shuo Sun, Jiaze Sun, Molei Qin, Xinyi Li, Yuqing Zhao, Yilei Zhao, Xinyu Cai, Longtao Zheng, Xinrun Wang, Bo An

TL;DR

FinAgent addresses the need for robust, multimodal decision making in financial trading by integrating textual, numerical, and visual market information through a Market Intelligence Module, aided by a dual-level Reflection mechanism and a diversified memory retrieval system. The framework combines memory, reasoning, and tool augmentation to ground actions in established trading principles while enabling rapid adaptation to market dynamics. Empirical evaluation on six real-world datasets (stocks and crypto) shows FinAgent significantly outperforms 12 baselines with over 36% average profit improvement and a 92.27% return on one dataset, demonstrating strong generalization and explainability. This work highlights a path toward explainable, generalizable autonomous trading agents that leverage multimodal data and domain tools to synthesize insights and justify decisions effectively.

Abstract

Financial trading is a crucial component of the markets, informed by a multimodal information landscape encompassing news, prices, and Kline charts, and encompasses diverse tasks such as quantitative trading and high-frequency trading with various assets. While advanced AI techniques like deep learning and reinforcement learning are extensively utilized in finance, their application in financial trading tasks often faces challenges due to inadequate handling of multimodal data and limited generalizability across various tasks. To address these challenges, we present FinAgent, a multimodal foundational agent with tool augmentation for financial trading. FinAgent's market intelligence module processes a diverse range of data-numerical, textual, and visual-to accurately analyze the financial market. Its unique dual-level reflection module not only enables rapid adaptation to market dynamics but also incorporates a diversified memory retrieval system, enhancing the agent's ability to learn from historical data and improve decision-making processes. The agent's emphasis on reasoning for actions fosters trust in its financial decisions. Moreover, FinAgent integrates established trading strategies and expert insights, ensuring that its trading approaches are both data-driven and rooted in sound financial principles. With comprehensive experiments on 6 financial datasets, including stocks and Crypto, FinAgent significantly outperforms 9 state-of-the-art baselines in terms of 6 financial metrics with over 36% average improvement on profit. Specifically, a 92.27% return (a 84.39% relative improvement) is achieved on one dataset. Notably, FinAgent is the first advanced multimodal foundation agent designed for financial trading tasks.

A Multimodal Foundation Agent for Financial Trading: Tool-Augmented, Diversified, and Generalist

TL;DR

FinAgent addresses the need for robust, multimodal decision making in financial trading by integrating textual, numerical, and visual market information through a Market Intelligence Module, aided by a dual-level Reflection mechanism and a diversified memory retrieval system. The framework combines memory, reasoning, and tool augmentation to ground actions in established trading principles while enabling rapid adaptation to market dynamics. Empirical evaluation on six real-world datasets (stocks and crypto) shows FinAgent significantly outperforms 12 baselines with over 36% average profit improvement and a 92.27% return on one dataset, demonstrating strong generalization and explainability. This work highlights a path toward explainable, generalizable autonomous trading agents that leverage multimodal data and domain tools to synthesize insights and justify decisions effectively.

Abstract

Financial trading is a crucial component of the markets, informed by a multimodal information landscape encompassing news, prices, and Kline charts, and encompasses diverse tasks such as quantitative trading and high-frequency trading with various assets. While advanced AI techniques like deep learning and reinforcement learning are extensively utilized in finance, their application in financial trading tasks often faces challenges due to inadequate handling of multimodal data and limited generalizability across various tasks. To address these challenges, we present FinAgent, a multimodal foundational agent with tool augmentation for financial trading. FinAgent's market intelligence module processes a diverse range of data-numerical, textual, and visual-to accurately analyze the financial market. Its unique dual-level reflection module not only enables rapid adaptation to market dynamics but also incorporates a diversified memory retrieval system, enhancing the agent's ability to learn from historical data and improve decision-making processes. The agent's emphasis on reasoning for actions fosters trust in its financial decisions. Moreover, FinAgent integrates established trading strategies and expert insights, ensuring that its trading approaches are both data-driven and rooted in sound financial principles. With comprehensive experiments on 6 financial datasets, including stocks and Crypto, FinAgent significantly outperforms 9 state-of-the-art baselines in terms of 6 financial metrics with over 36% average improvement on profit. Specifically, a 92.27% return (a 84.39% relative improvement) is achieved on one dataset. Notably, FinAgent is the first advanced multimodal foundation agent designed for financial trading tasks.
Paper Structure (92 sections, 4 equations, 36 figures, 8 tables)

This paper contains 92 sections, 4 equations, 36 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. LLM Agents for Decision Making
  4. AI for Financial Trading
  5. Problem Formulation
  6. Financial Trading as MDP
  7. Problem Formulation
  8. FinAgent Framework
  9. Market Intelligence Module
  10. Memory Module
  11. Reflection Module
  12. Tool-Augmented Decision-making Module
  13. Experiment Setup
  14. Datasets
  15. Evaluation Metrics
  16. ...and 77 more sections

Figures (36)

  • Figure 1: Overview of FinAgent.
  • Figure 2: The overall architecture of FinAgent. The ordinal numbers in the figure represent the order of execution, where augmented tools are implemented with the decision-making module.
  • Figure 3: Case studies of FinAgent. We only display the partial prompt for brevity.
  • Figure 4: Performance comparison over time between FinAgent and other benchmarks across all assets.
  • Figure 5: (a) Performance of FinAgent with/without diversified retrieval on AAPL. (b) Visualization of diversified retrieved market intelligence embedding by t-SNE on AAPL.
  • ...and 31 more figures