Agent Trading Arena: A Study on Numerical Understanding in LLM-Based Agents
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916929149599744 |
|---|---|
| author | Ma, Tianmi Du, Jiawei Huang, Wenxin Wang, Wenjie Xie, Liang Zhong, Xian Zhou, Joey Tianyi |
| author_facet | Ma, Tianmi Du, Jiawei Huang, Wenxin Wang, Wenjie Xie, Liang Zhong, Xian Zhou, Joey Tianyi |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_17967 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Agent Trading Arena: A Study on Numerical Understanding in LLM-Based Agents Ma, Tianmi Du, Jiawei Huang, Wenxin Wang, Wenjie Xie, Liang Zhong, Xian Zhou, Joey Tianyi Machine Learning Artificial Intelligence Computation and Language Multiagent Systems Statistical Finance 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. |
| title | Agent Trading Arena: A Study on Numerical Understanding in LLM-Based Agents |
| topic | Machine Learning Artificial Intelligence Computation and Language Multiagent Systems Statistical Finance |
| url | https://arxiv.org/abs/2502.17967 |