Toward Expert Investment Teams:A Multi-Agent LLM System with Fine-Grained Trading Tasks

TL;DR

This paper introduces a multi-agent LLM trading system that decomposes investment analysis into detailed tasks, significantly improving risk-adjusted returns and transparency in financial trading applications.

Contribution

It proposes a novel multi-agent framework with fine-grained task decomposition, enhancing performance and interpretability over traditional coarse-grained approaches.

Findings

Fine-grained task decomposition improves risk-adjusted returns.

Alignment of analytical outputs with decision preferences is crucial.

Portfolio optimization exploits low correlation and output variance for better results.

Abstract

The advancement of large language models (LLMs) has accelerated the development of autonomous financial trading systems. While mainstream approaches deploy multi-agent systems mimicking analyst and manager roles, they often rely on abstract instructions that overlook the intricacies of real-world workflows, which can lead to degraded inference performance and less transparent decision-making. Therefore, we propose a multi-agent LLM trading framework that explicitly decomposes investment analysis into fine-grained tasks, rather than providing coarse-grained instructions. We evaluate the proposed framework using Japanese stock data, including prices, financial statements, news, and macro information, under a leakage-controlled backtesting setting. Experimental results show that fine-grained task decomposition significantly improves risk-adjusted returns compared to conventional coarse-grained designs. Crucially, further analysis of intermediate agent outputs suggests that alignment between analytical outputs and downstream decision preferences is a critical driver of system performance. Moreover, we conduct standard portfolio optimization, exploiting low correlation with the stock index and the variance of each system's output. This approach achieves superior performance. These findings contribute to the design of agent structure and task configuration when applying LLM agents to trading systems in practical settings.