Design and Empirical Study of a Large Language Model-Based Multi-Agent Investment System for Chinese Public REITs

TL;DR

This paper presents a multi-agent LLM-based trading system for Chinese REITs, demonstrating improved risk-adjusted returns through collaborative analysis and prediction, with a comparison of large versus fine-tuned small models in backtests.

Contribution

It introduces a novel multi-agent framework utilizing LLMs for REITs trading and compares the effectiveness of general-purpose versus fine-tuned models within this system.

Findings

Both strategies outperform buy-and-hold benchmarks.

The multi-agent system enhances risk-adjusted returns.

Fine-tuned small models perform comparably or better than large models.

Abstract

This study addresses the low-volatility Chinese Public Real Estate Investment Trusts (REITs) market, proposing a large language model (LLM)-driven trading framework based on multi-agent collaboration. The system constructs four types of analytical agents-announcement, event, price momentum, and market-each conducting analysis from different dimensions; then the prediction agent integrates these multi-source signals to output directional probability distributions across multiple time horizons, then the decision agent generates discrete position adjustment signals based on the prediction results and risk control constraints, thereby forming a closed loop of analysis-prediction-decision-execution. This study further compares two prediction model pathways: for the prediction agent, directly calling the general-purpose large model DeepSeek-R1 versus using a specialized small model Qwen3-8B fine-tuned via supervised fine-tuning and reinforcement learning alignment. In the backtest from October 2024 to October 2025, both agent-based strategies significantly outperformed the buy-and-hold benchmark in terms of cumulative return, Sharpe ratio, and maximum drawdown. The results indicate that the multi-agent framework can effectively enhance the risk-adjusted return of REITs trading, and the fine-tuned small model performs close to or even better than the general-purpose large model in some scenarios.