LLM-Powered Multi-Agent System for Automated Crypto Portfolio Management (2501.00826v2)

Abstract: Cryptocurrency investment is inherently difficult due to its shorter history compared to traditional assets, the need to integrate vast amounts of data from various modalities, and the requirement for complex reasoning. While deep learning approaches have been applied to address these challenges, their black-box nature raises concerns about trust and explainability. Recently, LLMs have shown promise in financial applications due to their ability to understand multi-modal data and generate explainable decisions. However, single LLM faces limitations in complex, comprehensive tasks such as asset investment. These limitations are even more pronounced in cryptocurrency investment, where LLMs have less domain-specific knowledge in their training corpora. To overcome these challenges, we propose an explainable, multi-modal, multi-agent framework for cryptocurrency investment. Our framework uses specialized agents that collaborate within and across teams to handle subtasks such as data analysis, literature integration, and investment decision-making for the top 30 cryptocurrencies by market capitalization. The expert training module fine-tunes agents using multi-modal historical data and professional investment literature, while the multi-agent investment module employs real-time data to make informed cryptocurrency investment decisions. Unique intrateam and interteam collaboration mechanisms enhance prediction accuracy by adjusting final predictions based on confidence levels within agent teams and facilitating information sharing between teams. Empirical evaluation using data from November 2023 to September 2024 demonstrates that our framework outperforms single-agent models and market benchmarks in classification, asset pricing, portfolio, and explainability performance.

• The paper proposes an LLM-powered multi-agent system for automated crypto portfolio management, demonstrating superior performance against baselines in empirical tests.
• The framework utilizes a two-module architecture with specialized agents that collaborate and integrate multi-modal data for robust investment decisions.
• The framework enhances explainability via literature-trained agents, addressing AI black-box issues to build trust in AI-driven crypto management strategies.

Overview of LLM-Powered Multi-Agent System for Automated Crypto Portfolio Management

The paper presented offers a comprehensive exploration of a novel multi-agent framework for cryptocurrency portfolio management driven by LLMs. Given the inherent complexities and volatility in cryptocurrency markets, marked by their relatively short history and the need for multifaceted data integration, this research seeks to address the challenges of trust and explainability in financial decision-making processes.

Framework Architecture

The authors propose a sophisticated multi-modal, multi-agent system aimed at effective cryptocurrency investment in high-capital cryptocurrencies. This framework is organized into two principal modules: the expert training module and the multi-agent investment module. The modular design allows specialized agents to focus on task-specific responsibilities, thereby decomposing complex investment decisions into manageable subtasks.

1. Expert Training Module: This component employs collaboration amongst agents dedicated to data fetching, literature analysis, and explanation, utilizing historical data and investment literature to fine-tune the agents. The literature integration enriches the agents' capabilities to provide reasoned, explainable outputs.
2. Multi-Agent Investment Module: Utilizing real-time data, this module operationalizes an integrated investment strategy. Agents in this section analyze market-specific and crypto-specific risk factors, processing data through both market and crypto teams to inform investment decisions.

Technical Contributions

The authors highlight several technical advancements within the framework:

• Inter-agent Collaboration: Agents leverage sophisticated collaboration mechanisms both within and across teams, adjusting predictions and facilitating improved information synthesis. This aspect is fundamental to enhancing prediction accuracy and offering a more nuanced approach to cryptocurrency investment.
• Explainability Mechanisms: The inclusion of fine-tuned agents for specific subtasks, trained with literature-backed knowledge, boosts the framework's capacity to generate understandable and contextually grounded investment explanations.
• Multi-modal Data Utilization: The framework's capability to integrate diverse data modalities, including textual, visual, and risk factor data, fosters a robust foundation for decision-making processes.

Empirical Evaluation

The framework was empirically tested using data spanning from June 2023 to September 2024, evaluating its performance against both single-agent baselines and established market benchmarks. The results demonstrate that the proposed multi-agent system surpasses both fine-tuned and unfine-tuned single-agent models in terms of classification accuracy and asset pricing efficacy. Additionally, the framework's portfolio performance excelled against market benchmarks, highlighting its potential in delivering superior investment outcomes.

Implications and Future Directions

The implications of this research extend both practically and theoretically. Practically, the framework provides a viable solution for investors seeking explainable AI-driven cryptocurrency management strategies. The ability to decompose complex tasks into specialized subcomponents represents a significant stride towards mitigating black-box issues prevalent in financial applications of AI.

Theoretically, the framework embodies potential avenues for advancing the state-of-the-art in LLM-powered financial technologies. Future research may focus on refining collaboration mechanisms between agents and exploring applications in a broader array of financial domains. Additionally, investigating the integration of novel data types could further enhance predictive capabilities and portfolio outcomes.

In conclusion, this research presents a substantive development in LLM-powered cryptocurrency investment, offering a methodologically sound approach with significant implications for the wider acceptance and trust in AI-driven financial solutions. The comprehensive evaluation underscores this framework's capacity to function as a foundational model for future innovations in digital asset management.