Agent Trading Arena: A Study on Numerical Understanding in LLM-Based Agents
This addresses the gap in training LLM-based agents for real-world financial trading, though it is incremental as it builds on existing multi-agent and visualization methods.
The paper tackled the problem of LLMs' poor performance in dynamic financial environments by introducing the Agent Trading Arena, a virtual stock market for competitive multi-agent trading, and found that chart-based visualizations significantly enhance numerical reasoning and trading performance, with evaluations showing superiority under high volatility.
Large language models (LLMs) have demonstrated remarkable capabilities in natural language tasks, yet their performance in dynamic, real-world financial environments remains underexplored. Existing approaches are limited to historical backtesting, where trading actions cannot influence market prices and agents train only on static data. To address this limitation, we present the Agent Trading Arena, a virtual zero-sum stock market in which LLM-based agents engage in competitive multi-agent trading and directly impact price dynamics. By simulating realistic bid-ask interactions, our platform enables training in scenarios that closely mirror live markets, thereby narrowing the gap between training and evaluation. Experiments reveal that LLMs struggle with numerical reasoning when given plain-text data, often overfitting to local patterns and recent values. In contrast, chart-based visualizations significantly enhance both numerical reasoning and trading performance. Furthermore, incorporating a reflection module yields additional improvements, especially with visual inputs. Evaluations on NASDAQ and CSI datasets demonstrate the superiority of our method, particularly under high volatility. All code and data are available at https://github.com/wekjsdvnm/Agent-Trading-Arena.