TradExpert: Revolutionizing Trading with Mixture of Expert LLMs

TL;DR

TradExpert introduces a Mixture of Experts framework where four specialized LLMs (News, Market Data, Alpha Factors, Fundamental Data) operate in parallel to generate expert reports, which are then synthesized by a General Expert to predict stock movements or rank stocks for trading. A reprogramming mechanism converts time-series data into LLM-friendly embeddings, enabling multimodal reasoning across unstructured text and structured market signals; two modes—Prediction and Ranking—support stock movement forecasting and Top-K trading decisions, respectively. The authors release a large-scale multi-source financial dataset and demonstrate that TradExpert consistently outperforms state-of-the-art baselines on both stock movement prediction and stock trading simulation, with ablation analyses confirming the contributions of each data source and the ranking methodology. The work advances practical AI-assisted trading by integrating heterogeneous data through MoE LLMs and by proposing a robust ranking mechanism, while noting latency as a current limitation and pointing to future extensions into high-frequency and broader-market contexts.

Abstract

The integration of Artificial Intelligence (AI) in the financial domain has opened new avenues for quantitative trading, particularly through the use of Large Language Models (LLMs). However, the challenge of effectively synthesizing insights from diverse data sources and integrating both structured and unstructured data persists. This paper presents TradeExpert, a novel framework that employs a mix of experts (MoE) approach, using four specialized LLMs, each analyzing distinct sources of financial data, including news articles, market data, alpha factors, and fundamental data. The insights of these expert LLMs are further synthesized by a General Expert LLM to make a final prediction or decision. With specific prompts, TradeExpert can be switched between the prediction mode and the ranking mode for stock movement prediction and quantitative stock trading, respectively. In addition to existing benchmarks, we also release a large-scale financial dataset to comprehensively evaluate TradeExpert's effectiveness. Our experimental results demonstrate TradeExpert's superior performance across all trading scenarios.

Figures (17)

• Figure 1: Illustration of traditional, LLM-based, and MoE LLMs-based financial models with diverse financial data sources.
• Figure 2: TradExpert operates by processing distinct sources of financial data such as news texts, market data, alpha factors, and fundamental data through specialized expert LLMs. Then their reports are sumarized and sent to a General Expert which delivers the final outputs: (1) prediction of stock movement with prediction mode, (2) which of the two stocks is better or worse with ranking mode.
• Figure 3: Instruction and prompt for the News Analyst.
• Figure 4: Instruction and prompt for the Market Analyst.
• Figure 5: Instructions and prompts for the General Expert LLM: (Top) Prediction mode, (Bottom) Ranking mode.