Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Kunihiro Miyazaki, Takanobu Kawahara, Stephen Roberts, Stefan Zohren

TL;DR

This work proposes a multi-agent LLM trading framework that explicitly decomposes investment analysis into fine-grained tasks, rather than providing coarse-grained instructions, and shows that fine-grained task decomposition significantly improves risk-adjusted returns compared to conventional coarse-grained designs.

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.

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

TL;DR

This work proposes a multi-agent LLM trading framework that explicitly decomposes investment analysis into fine-grained tasks, rather than providing coarse-grained instructions, and shows that fine-grained task decomposition significantly improves risk-adjusted returns compared to conventional coarse-grained designs.

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.
Paper Structure (54 sections, 6 equations, 3 figures, 5 tables)

This paper contains 54 sections, 6 equations, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Multi-Agent Trading Systems with LLMs
  4. Prompt Designs and Expert Task Settings
  5. Problem Setting
  6. Backtesting and Model Setup
  7. Evaluation Metrics
  8. Multi-Agent Framework Configuration
  9. Hierarchical Decision-Making Process
  10. Data Sources
  11. Methodology
  12. Tasks and Prompt Design
  13. Experimental Results
  14. Fine-Grained vs Coarse-Grained Tasks
  15. Comparison using All Agents
  16. ...and 39 more sections

Figures (3)

  • Figure 1: Overview of our multi-agent LLM trading system (see main text for details).
  • Figure 2: Sharpe ratios for fine-grained (pink) and coarse-grained (blue) settings across portfolio sizes. Box plot notches represent the 95% confidence interval of the median.
  • Figure 3: Sharpe ratio as a function of allocation between TOPIX 100 and the aggregated agent strategy in the test period. Gross performance (orange) and net performance after 10 bps one-way transaction cost (green) are shown.