Alpha 建模（Alpha Modeling）傳統上具有較高的進入門檻，需要使用者同時具備深厚的金融理論、機器學習建模以及高效能運算等專業知識。雖然近年來基於大型語言模型（LLM）的輔助系統已在一定程度上降低了這些門檻，但由單一代理（Agent）負責處理所有子任務的架構，往往容易造成上下文過載（Context Overload），且系統仍高度依賴人類介入協調，使得完全自主化的能力尚未被充分探討。為了解決上述問題，本研究提出一個用於 Alpha 建模的階層式多代理協作框架（hierarchical multi-agent framework）。系統以主管代理（Supervisor Agent）作為核心協調者，在基於檔案系統的工作區中追蹤各類量化研究產出，並動態分派任務至四個專責工作代理（Worker Agents），分別負責機器學習模型訓練、策略回測、績效歸因分析以及研究報告生成。系統進一步結合 ReAct 與 Chain-of-Thought（CoT）推理機制，以實現自主化執行流程。為評估該系統在降低使用門檻方面的成效，本研究基於改編的科技接受模型（Technology Acceptance Model, TAM）設計以人為中心的實驗。共有 22 位來自不同背景的受試者參與評估，並從四個構面進行分析，包括知覺易用性（Perceived Ease of Use）、知覺自主性（Perceived Autonomy）、知覺有用性（Perceived Usefulness）以及使用態度（Attitude Toward Using）。本研究的主要貢獻包括：(1) 提出一個面向 Alpha 建模的專用階層式多代理實作架構；以及 (2) 建立一套基於 TAM 的人機互動評估框架。研究結果顯示，該系統能有效降低量化研究過程中的操作與技術門檻，且其自主能力獲得高度肯定。然而，對於非專業使用者而言，系統輸出的理解與詮釋仍存在認知落差。此結果凸顯了 Agentic AI 在推動量化投資研究普及化方面的潛力，同時也反映出其目前仍面臨的限制與挑戰。

Alpha modeling traditionally involves high entry barriers, requiring deep expertise in financial theory, machine learning modeling, and high-performance computing. While recent LLM-assisted systems lower these barriers, the reliance on a single agent to handle all sub-tasks often leads to context overload, and the dependence on human-in-the-loop coordination leaves the potential for fully autonomous dynamic routing largely unexamined. To address this, we propose a hierarchical multi-agent framework for alpha modeling. A Supervisor Agent dynamically routes tasks within a filesystem-based virtual workspace tracking quantitative artifacts, while four specialized worker agents handle ML training, backtesting, attribution analysis, and reporting. The system adopts ReAct and Chain-of-Thought (CoT) reasoning to enable autonomous execution. To evaluate its effectiveness in lowering entry barriers, we conduct a human-centric experiment based on a modified Technology Acceptance Model (TAM). A total of 22 participants from diverse backgrounds evaluated the system across four dimensions: Perceived Ease of Use, Perceived Autonomy, Perceived Usefulness, and Attitude Toward Using. This work contributes a domain-specific hierarchical multi-agent implementation and a TAM-based human evaluation framework. Results demonstrate that while the system successfully alleviates operational and technical barriers with highly recognized autonomy, cognitive gaps in output comprehensibility persist for non-expert users, highlighting both the potential and current limitations of agentic AI in democratizing quantitative investment research.

Bai, S., Kolter, J. Z., & Koltun, V. (2018, April 19). An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling. arXiv: 1803.01271 [cs.LG]. (Cit. p. 5).
Bergstra, J., & Bengio, Y. (2012). Random search for hyper-parameter optimization. Journal of machine learning research, 13(2), 281–305 (cit. p. 6).
Blount, A., Gulli, A., Saboo, S., Zimmermann, M., & Vuskovic, V. (2025, November). Introduction to Agents (White Paper). Google. (Cit. p. 6).
Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794 (cit. p. 5).
Cho, K., van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation. In A. Moschitti, B. Pang, & W. Daelemans (Eds.), Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) (pp. 1724–1734). Association for Computational Linguistics. (Cit. p. 5).
Davis, F. D. (1993). User acceptance of information technology: System characteristics, user perceptions and behavioral impacts. International Journal of Man-Machine Studies, 38(3), 475–487 (cit. pp. 3, 8, 9).
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User Acceptance of Computer Technology: A Comparison of Two Theoretical Models. Management Science, 35(8), 982–1003 (cit. pp. 8, 9).
Falkner, S., Klein, A., & Hutter, F. (2018). BOHB: Robust and efficient hyperparameter optimization at scale. International Conference on Machine Learning, 1437–1446 (cit. p. 6).
Fang, H., Han, B., Erickson, N., Zhang, X., Zhou, S., Dagar, A., Zhang, J., Turkmen, A. C., Hu, T., Rangwala, H., Wu, Y. N., Wang, Y., & Karypis, G. (2025). MLZero: A multi-agent system for end-to-end machine learning automation. In D. Belgrave, C. Zhang, H. Lin, R. Pascanu, P. Koniusz, M. Ghassemi, & N. Chen (Eds.), Advances in neural information processing systems (pp. 69001–69070, Vol. 38). Curran Associates, Inc. (Cit. p. 9).
Fishbein, M., & Ajzen, I. (1975, May 1). Belief, attitude, intention and behaviour: An introduction to theory and research (Vol. 27). Addison-Wesley. (Cit. p.8).
Gottweis, J., Weng, W.-H., Daryin, A., Tu, T., Palepu, A., Sirkovic, P., Myaskovsky, A., Weissenberger, F., Rong, K., Tanno, R., Saab, K., Popovici, D., Blum, J., Zhang, F., Chou, K., Hassidim, A., Gokturk, B., Vahdat, A., Kohli, P., … Natarajan, V. (2025, February 26). Towards an AI co-scientist. arXiv: 2502.18864 [cs]. (Cit. p. 7).
Gu, S., Kelly, B., & Xiu, D. (2020). Empirical Asset Pricing via Machine Learning. The Review of Financial Studies, 33(5), 2223–2273 (cit. p. 1).
Gulli, A., Nigam, L., Wiesinger, J., Vuskovic, V., Sigler, I., Nardini, I., Stroppa, N., Kartakis, S., Saribekyan, N., Nawalgaria, A., & Bount, A. (2025, February). Agents Companion (White Paper). Google. (Cit. p. 7).
Guo, J., Wang, S., Ni, L. M., & Shum, H.-Y. (2024). Quant 4.0: Engineering Quantitative Investment with Automated, Explainable, and Knowledge-Driven Artificial Intelligence. Frontiers of Information Technology & Electronic Engineering, 25(11), 1421–1445 (cit. pp. 1, 5).
He, X., Zhao, K., & Chu, X. (2021). AutoML: A survey of the state-of-the-art. Knowledge-Based Systems, 212, 106622 (cit. p. 5).
Hili, Y. A. E., Thomas, A., Tiomoko, M., Benechehab, A., Léger, C., Ancourt, C., & Kégl, B. (2026, February 13). LLMs as In-Context Meta-Learners for Model and Hyperparameter Selection. arXiv: 2510.26510 [cs]. (Cit. p. 9).
Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735–1780 (cit. p. 5).
Kakushadze, Z. (2016). 101 Formulaic Alphas: 101 Formulaic Alphas. Wilmott, 2016(84), 72–81 (cit. p. 5).
Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., & Liu, T.-Y. (2017). LightGBM: A Highly Efficient Gradient Boosting Decision Tree. Advances in Neural Information Processing Systems, 30 (cit. p. 5).
Koa, K. J., Ma, Y., Ng, R., & Chua, T.-S. (2024). Learning to Generate Explainable Stock Predictions using Self-Reflective Large Language Models. Proceedings of the ACM Web Conference 2024, 4304–4315 (cit. p. 5).
Krauss, C., Do, X. A., & Huck, N. (2017). Deep neural networks, gradient-boosted trees, random forests: Statistical arbitrage on the S&P 500. European Journal of Operational Research, 259(2), 689–702 (cit. p. 1).
Li, Y., Yang, X., Yang, X., Wang, X., Liu, W., & Bian, J. (2025). R&d-agent-quant: A multi-agent framework for data-centric factors and model joint optimization. In D. Belgrave, C. Zhang, H. Lin, R. Pascanu, P. Koniusz, M. Ghassemi, & N. Chen (Eds.), Advances in neural information processing systems (Vol. 38). Curran Associates, Inc. (Cit. pp. 1, 10).
Liu, S., Gao, C., & Li, Y. (2025). AgentHPO: Large Language Model Agent for Hyper-Parameter Optimization. Conference on Parsimony and Learning, 1146–1169 (cit. p. 9).
Mohammadi, M., Li, Y., Lo, J., & Yip, W. (2025). Evaluation and Benchmarking of LLM Agents: A Survey. Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2, 6129–6139 (cit. p. 7).
Nam, J., Yoon, J., Chen, J., Shin, J., Arik, S., & Pfister, T. (2025). MLE-STAR: Machine learning engineering agent via search and targeted refinement. In D. Belgrave, C. Zhang, H. Lin, R. Pascanu, P. Koniusz, M. Ghassemi, & N. Chen (Eds.), Advances in neural information processing systems (pp. 116692–116712, Vol. 38). Curran Associates, Inc. (Cit. p. 9).
Novikov, A., Vũ, N., Eisenberger, M., Dupont, E., Huang, P.-S., Wagner, A. Z., Shirobokov, S., Kozlovskii, B., Ruiz, F. J. R., Mehrabian, A., Kumar, M. P., See, A., Chaudhuri, S., Holland, G., Davies, A., Nowozin, S., Kohli, P., & Balog, M. (2025, June 16). AlphaEvolve: A coding agent for scientific and algorithmic discovery. arXiv: 2506.13131 [cs]. (Cit. p. 7).
Nunnally, J. C. (1978). Psychometric Theory. McGraw-Hill. (Cit. p. 33).
Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A. V., & Gulin, A. (2018). CatBoost: Unbiased boosting with categorical features. Advances in Neural Information Processing Systems, 31 (cit. p. 5).
Sapkota, R., Roumeliotis, K. I., & Karkee, M. (2026). AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challenges. Information Fusion, 126, 103599 (cit. pp. 1, 6).
Sezer, O. B., Gudelek, M. U., & Ozbayoglu, A. M. (2020). Financial time series forecasting with deep learning: A systematic literature review: 2005–2019. Applied Soft Computing, 90, 106181 (cit. p. 1).
Wang, L., Xu, W., Lan, Y., Hu, Z., Lan, Y., Lee, R. K.-W., & Lim, E.-P. (2023). Plan-and-solve prompting: Improving zero-shot chain-of-thought reasoning by large language models. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2609–2634 (cit. p. 6).
Wang, S., Kong, H., Guo, J., Hua, F., Qi, Y., Zhou, W., Zheng, J., Wang, X., Ni, L. M., & Guo, J. (2025a). QuantBench: Benchmarking AI methods for quantitative investment from a full pipeline perspective. Frontiers of Information Technology & Electronic Engineering, 26(11), 2282–2297 (cit. pp. 2, 5, 6).
Wang, S., Yuan, H., Zhou, L., Ni, L., Shum, H. Y., & Guo, J. (2025b). Alpha-gpt: Human-ai interactive alpha mining for quantitative investment. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, 196–206 (cit. p. 1).
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., Le, Q. V., & Zhou, D. (2022). Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35, 24824–24837 (cit. pp. 6, 14).
Wiesinger, J., Marlow, P., & Vuskovic, V. (2024, September). Agents (White Paper). Google. (Cit. p. 6).
Xiao, Y., Sun, E., Luo, D., & Wang, W. (2025, June 3). TradingAgents: Multi-Agents LLM Financial Trading Framework. arXiv: 2412.20138 [q-fin.TR]. (Cit. p. 10).
Yang, X., Liu, W., Zhou, D., Bian, J., & Liu, T.-Y. (2020, September 22). Qlib: An AI-oriented Quantitative Investment Platform. arXiv: 2009.11189 [q-fin.GN]. (Cit. pp. 2, 5, 23).
Yao, S., Yu, D., Zhao, J., Shafran, I., Griffiths, T., Cao, Y., & Narasimhan, K. (2023). Tree of Thoughts: Deliberate Problem Solving with Large Language Models. Advances in Neural Information Processing Systems, 36, 11809–11822 (cit. p. 6).
Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K. R., & Cao, Y. (2022). ReAct: Synergizing Reasoning and Acting in Language Models (cit. pp. 6, 14).
Yuan, H., Wang, S., & Guo, J. (2024, February 15). Alpha-GPT 2.0: Human-in-the-Loop AI for Quantitative Investment. arXiv: 2402.09746 [q-fin.CP]. (Cit. pp. 1, 2, 10, 11).
Zhang, M. R., Desai, N., Bae, J., Lorraine, J., & Ba, J. (2024, November 11). Using Large Language Models for Hyperparameter Optimization. arXiv: 2312.04528 [cs.LG]. (Cit. p. 9).