Utility Difference Analysis of Single-Agent and Multi-Agent Large Language Models in Stock Investment Decision-Making

Yichuan Luo

doi:10.54097/vc9ccz57

Authors

Yichuan Luo College of Business, Shanghai University of Finance and Economics, Shanghai, 200433, China

DOI:

https://doi.org/10.54097/vc9ccz57

Keywords:

Large Language Models (LLMs), Multi-Agent Systems, Stock Investment Decision-Making, Quantitative Investment, Artificial Intelligence in Finance.

Abstract

The area of financial investment decision-making is undergoing significant changes with the help of artificial intelligence and especially large language models (LLMs). The current state of research has largely been dedicated to the single agent application paradigm, this paper proposes and analyzes an innovative multi-agent LLM architecture. The research has developed a two-level structure of the division of labor into a specialist and a centralized decision-making site. There are three expert agents in the analysis layer: Value, Growth, and Macro. One of the agents of this multi-agent system happens to be the Chief Investment Officer (CIO) agent who oversees dynamic integration and the results of the 24-year backtest of a Nasdaq 100 index fund prove that the multi-agent system is much superior to the single-agent system on all the key performance indicators. It has grown by 111.8 percent in total return to 1482.31, its Sharpe ratio has almost doubled, the biggest drawdown has diminished by 36.9 percent, and it has effectively turned a negative value of Alpha to a positive value. This suggests that the counterbalancing of cognitive bias with the help of the multi-agent architecture will prove more efficient in generating excess returns and managing the risk.

Downloads

Download data is not yet available.

References

[1] Cao, B., Wang, S., Lin, X., Wu, X., Zhang, H., Ni, L. M., & Guo, J. (2025). From deep learning to LLMs: A survey of AI in quantitative investment. arXiv preprintarXiv:2503.21422.

[2] Ding, H., Li, Y., Wang, J., & Chen, H. (2024). Large language model agent in financial trading: A survey. arXiv preprintarXiv:2408.06361.

[3] Grosu, F. (2025). How can multi-agents AI systems help reduce biases in trading algorithms? Review of International Comparative Management.

[4] Jiang, J., Kelly, B., & Xiu, D. (2023). (Re-)Imag(in)ing price trends. The Journal of Finance, 78(5), 2617–2663.

[5] Kelly, B. (2023). The virtue of complexity in return prediction. The Journal of Finance, 78(4), 1839–1886.

[6] Kou, Z., Yu, H., Luo, J., Peng, J., Li, X., Liu, C., Dai, J., Chen, L., Han, S., & Guo, Y. (2024). Automate strategy finding with LLM in quant investment. arXiv preprintarXiv:2409.06289.

[7] Lee, J., Kim, R., Yi, S.-W., & Kang, J. (2020). MAPS: Multi-agent reinforcement learning-based portfolio management system. In Proceedings of the 29th International Joint Conference on Artificial Intelligence (IJCAI)(pp. 4596–4602).

[8] Li, W. W., Kim, H., Cucuringu, M., & Ma, T. (2025). Can LLM-based financial investing strategies outperform the market in long run? arXiv preprintarXiv:2505.07078.

[9] Lussange, J., et al. (2021). Multi-agent reinforcement learning for market microstructure statistical inference. Journal of Financial Data Science, 3(4), 112–127.

[10] Malberg, S., Poletukhin, R., Schuster, C. M., & Groh, G. (2024). A comprehensive evaluation of cognitive biases in LLMs. arXiv preprint.

[11] Wall, L. F., & Leitner, S. (2021). Agent-based computational economics in management accounting research: Opportunities and difficulties. Journal of Management Accounting Research, 33(3), 189–212.

[12] Xiao, Y., Sun, E., Luo, D., & Wang, W. (2024). TradingAgents: Multi-agents LLM financial trading framework. arXiv preprint.

[13] Yu, Y., Yao, Z., Li, H., Deng, Z., Cao, Y., Chen, Z., et al. (2024). FINCON: A synthesized LLM multi-agent system with conceptual verbal reinforcement for enhanced financial decision making. In Advances in Neural Information Processing Systems (NeurIPS).