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

TL;DR

This work tackles the difficulty of crypto portfolio management under limited asset history and high volatility by introducing an explainable, multi-modal, multi-agent framework that integrates data analysis, literature-informed knowledge, and real-time decision-making for the top 30 cryptocurrencies. It combines expert training (data, literature, explanation) with a three-agent investment module (market, crypto, trading) and leverages intrateam and interteam memory sharing to refine predictions and explanations. The approach demonstrates superior performance over single-agent baselines in classification accuracy, asset pricing, and portfolio returns, while also delivering richer, finance-grounded explanations; ablations reveal the value of collaboration. The framework advances practical crypto investment by increasing predictability, interpretability, and risk-adjusted returns in volatile markets, with empirical validation using data from mid-2023 to late-2024 and external benchmarks.

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, large language models (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.

Figures (7)

• Figure 1: Multi-modal data utilized by our multi-agent framework. The detailed data description can be found in \ref{['apx:data']}.
• Figure 2: Multi-agent framework for automated cryptocurrency portfolio management.
• Figure 3: Collaboration among expert agents within the same group and across different groups.
• Figure 4: Distribution and disagreement in the rise probability predictions of expert agents within the same group across various models.
• Figure 5: Performance comparisons in out-of-sample cumulative returns of our multi-agent model portfolio against baselines. The green span represents the boom period, while the red span indicates the bust period, as determined by the Nasdaq Crypto Index (Market).