Paper: arXiv 2409.06289

Abstract

We present a novel three-stage framework leveraging Large Language Models (LLMs) within a risk-aware multi-agent system for automate strategy finding in quantitative finance. Our approach addresses the brittleness of traditional deep learning models in financial applications by: employing prompt-engineered LLMs to generate executable alpha factor candidates across diverse financial data, implementing multimodal agent-based evaluation that filters factors based on market status, predictive quality while maintaining category balance, and deploying dynamic weight optimization that adapts to market conditions. Experimental results demonstrate the robust performance of the strategy in Chinese & US market regimes compared to established benchmarks. Our work extends LLMs capabilities to quantitative trading, providing a scalable architecture for financial signal extraction and portfolio construction. The overall framework significantly outperforms all benchmarks with 53.17% cumulative return on SSE50 (Jan 2023 to Jan 2024), demonstrating superior risk-adjusted performance and downside protection on the market.

Complexity vs Empirical Score

  • Math Complexity: 4.5/10
  • Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs not modelled; see penalty flags)
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper demonstrates high empirical rigor with detailed backtesting on specific markets, code availability, and specific metrics, but its mathematical formalization is relatively accessible, focusing on conceptual alpha formulation rather than deep derivations.

Research Flowchart

  flowchart TD
  A["Research Goal: Automate Strategy Finding<br/>for Quant Investment using LLMs"] --> B["Stage 1: Alpha Factor Generation<br/>Prompt-engineered LLMs analyze financial data"]
  B --> C["Stage 2: Multi-Agent Evaluation<br/>Risk-aware filtering & market status assessment"]
  C --> D["Stage 3: Dynamic Weight Optimization<br/>Adaptive to market regimes"]
  D --> E["Key Outcomes: 53.17% Cumulative Return<br/>on SSE50 (Jan 2023-2024)"]
  E --> F["Superior Risk-Adjusted Performance<br/>with Downside Protection"]
  
  %% Data Inputs
  B -.->|Financial Data| G["Chinese & US Market Data"]
  C -.->|Multi-modal Agents| H["Predictive Quality & Category Balance"]
  D -.->|Market Conditions| I["Dynamic Portfolio Construction"]

Cited References (181)

  1. Alternative Data Market Size, Share, & Trends Analysis By Data Type (Credit & Debit Card Transactions, Mobile Application Usage, Social & Sentiment Data), By Industry, By End-user, By Region, And Segment Forecasts, 2024 - 2030 (2023) — link
  2. 3 Quantitative Strategies Based on Alternative Data (2024)
  3. Finding Alphas: A Quantitative Approach to Building Trading Strategies (2015)
  4. Efficient capital markets: A review of theory and empirical work (1970)
  5. Efficient market hypothesis and forecasting (2004)
  6. AlphaEvolve: A Learning Framework to Discover Novel Alphas in Quantitative Investment (2021)
  7. Using genetic algorithms to find technical trading rules (1999) — link
  8. Generating Synergistic Formulaic Alpha Collections via Reinforcement Learning — → 2306.12964
and 173 more
  1. Large Language Model Agent: A Survey on Methodology, Applications and Challenges (2025) — link
  2. Investment behaviors can tell what inside: Exploring stock intrinsic properties for stock trend prediction (2019)
  3. Fundamentally, Momentum is Fundamental Momentum (2015) — link
  4. Investlm: A Large Language Model for Investment Using Financial Domain Instruction Tuning (2023) — link
  5. Instruction tuning for large language models: A survey (2023) — link
  6. Large language models in finance: A survey (2023)
  7. GPT-4 for Financial Statements: Building an AI Analyst (2023)
  8. Llama: Open and Efficient Foundation Language Models (2023) — link
  9. A Comparative Analysis of Offline-Tuned LLMs and Few-Shot Learning of LLMs for Financial Sentiment Analysis (2023) — link
  10. Instruction Tuning for Sentiment Analysis of Finance Text (2023) — link
  11. Comprehensive Guidelines on Using LLMs for Textual Analysis (2023) — link
  12. LoRA: Low-Rank Adaptation of Large Language Models (2021)
  13. Is ChatGPT a Financial Expert? Evaluating Language Models on Financial Natural Language Processing (2023)
  14. Learning to Generate Explainable Stock Predictions Using Self-Reflective Large Language Models — → 2402.03659
  15. Exploration of Zero-Shot Learning Capabilities of LLMs (2021) — link
  16. Revisiting Stock Alpha Mining Based On Genetic Algorithm (2019) — link
  17. AutoAlpha: An Efficient Hierarchical Evolutionary Algorithm for Mining Alpha Factors in Quantitative Investment (2020) — link
  18. Alpha Go Everywhere: Machine Learning and International Stock Returns (2023) — link
  19. Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges Toward Responsible AI (2020)
  20. Stock Price Prediction via Discovering Multi-Frequency Trading Patterns (2017)
  21. Learning Equations for Extrapolation and Control (2018)
  22. Deep Symbolic Regression: Recovering Mathematical Expressions from Data via Risk-Seeking Policy Gradients (2021)
  23. SymbolicGPT: A Generative Transformer Model for Symbolic Regression (2021) — link
  24. CUMO: Scaling Multimodal LLM with Co-Upcycled Mixture-of-Experts (2024) — link
  25. Enhancing the Interpretability and Explainability of AI-Driven Risk Models Using LLM Capabilities (2024)
  26. LLM Multimodal Traffic Accident Forecasting (2023)
  27. Methods for Acquiring and Incorporating Knowledge into Stock Price Prediction: A Survey — → 2308.04947
  28. Mixture of experts: a literature survey (2014)
  29. Autogen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation Framework (2023) — link
  30. Quantitative equity investing: Techniques and strategies (2010)
  31. The sharpe ratio (1994)
  32. Chateval: Towards Better LLM-Based Evaluators Through Multi-Agent Debate (2023) — link
  33. Multi-Agent Collaboration: Harnessing the Power of Intelligent LLM Agents (2023) — link
  34. TradingGPT: Multi-Agent System with Layered Memory and Distinct Characters for Enhanced Financial Trading Performance (2023) — link
  35. FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design (2024)
  36. Symbolic Regression via Neural-Guided Genetic Programming Population Seeding (2021)
  37. A Review on Time Series Forecasting Techniques for Building Energy Consumption (2017)
  38. Online Detection of Stealthy False Data Injection Attacks in Power System State Estimation (2018)
  39. How Are You Feeling: A Personalized Methodology for Predicting Mental States from Temporally Observable Physical and Behavioral Information (2017)
  40. GPT-4 Technical Report (2023) — link
  41. Turkish stock market analysis using mixture of experts (2004)
  42. Predicting stock price from financial message boards with a mixture of experts framework (2006)
  43. Knowledge graph-based event embedding framework for financial quantitative investments (2020)
  44. Financial knowledge graph based financial report query system (2021)
  45. Hybrid dimension reduction by integrating feature selection with feature extraction method for text clustering (2015)
  46. Big data analysis of economic news: Hints to forecast macroeconomic indicators (2017)
  47. Trading algorithms with learning in latent alpha models (2019)
  48. Weighted gate layer autoencoders (2021)
  49. Universal features of price formation in financial markets: perspectives from deep learning (2021)
  50. The economics of financial markets (2005)
  51. Re-thinking data strategy and integration for artificial intelligence: concepts, opportunities, and challenges (2023)
  52. AlphaForge: A Framework to Mine and Dynamically Combine Formulaic Alpha Factors — → 2406.18394
  53. The Quest for Alpha: The Holy Grail of Investing (2010)
  54. Momentum in ESG indexes: A study on the passive capital flows effect on ESG stock prices (2021)
  55. Causal Factor Investing: Can Factor Investing Become Scientific? (2023)
  56. Financial time series prediction using mixture of experts (2003)
  57. Fingpt: Open-source financial large language models — → 2306.06031
  58. Alpha-gpt: Human-ai interactive alpha mining for quantitative investment — → 2308.00016
  59. Model-free implied dependence and the cross-section of returns (2023)
  60. Model-based deep reinforcement learning for dynamic portfolio optimization (2019) — link
  61. Can large language models beat wall street? unveiling the potential of ai in stock selection — → 2401.03737
  62. Mastering stock markets with efficient mixture of diversified trading experts (2023)
  63. End-to-end symbolic regression with transformers (2022)
  64. Financial regulation as interagency competition? The saga of venture capital rule-making in China (2020) — link
  65. Where do alphas come from?: A new measure of the value of active investment management (2007)
  66. Robo advisors, algorithmic trading and investment management: Wonders of fourth industrial revolution in financial markets (2021)
  67. Large language model based multi-agents: A survey of progress and challenges (2024) — link
  68. Contemporary symbolic regression methods and their relative performance (2021)
  69. A hybrid model integrating deep learning with investor sentiment analysis for stock price prediction (2021)
  70. Deep learning and symbolic regression for discovering parametric equations (2023)
  71. Integration of neural network-based symbolic regression in deep learning for scientific discovery (2020)
  72. Machine learning control by symbolic regression (2021)
  73. Interpretable scientific discovery with symbolic regression: a review (2024)
  74. Symbolic regression analysis (2002)
  75. Applications of genetic programming to finance and economics: past, present, future (2020)
  76. A genetic programming approach to binary classification problem (2021)
  77. Automated Formulaic Alpha Generation for Quantitative Investing using Evolutionary Algorithms (2022)
  78. Symbolic regression in materials science (2019)
  79. $$\backslash$text $\ $Alpha$\ $^ 2$: Discovering Logical Formulaic Alphas using Deep Reinforcement Learning — → 2406.16505
  80. Deep reinforcement learning in quantitative algorithmic trading: A review (2021) — link
  81. Financial numeral classification model based on BERT (2019)
  82. Role and challenges of ChatGPT and similar generative artificial intelligence in finance and accounting (2023)
  83. An empirical study of classification algorithm evaluation for financial risk prediction (2011)
  84. Ternary neural networks with fine-grained quantization (2017) — link
  85. How conventional measures overestimate electroweak fine-tuning in supersymmetric theory (2013)
  86. Few-shot parameter-efficient fine-tuning is better and cheaper than in-context learning (2022)
  87. Convolutional neural networks for medical image analysis: Full training or fine tuning? (2016)
  88. A review of generalized zero-shot learning methods (2022)
  89. A comprehensive survey of few-shot learning: Evolution, applications, challenges, and opportunities (2023)
  90. Few-shot learning with graph neural networks (2017) — link
  91. Zero-shot learning-the good, the bad and the ugly (2017)
  92. Fine-tuning and utilization methods of domain-specific llms (2024) — link
  93. Universal language model fine-tuning for text classification (2018) — link
  94. Designing heterogeneous llm agents for financial sentiment analysis (2024)
  95. Modern Generative AI with ChatGPT and OpenAI Models: Leverage the capabilities of OpenAI’s LLM for productivity and innovation with GPT3 and GPT4 (2023)
  96. Learning to compare: Relation network for few-shot learning (2018)
  97. Securing Large Language Models: Threats, Vulnerabilities and Responsible Practices (2024) — link
  98. Deepseek llm: Scaling open-source language models with longtermism (2024) — link
  99. Generalizing from a few examples: A survey on few-shot learning (2020)
  100. A review of two decades of correlations, hierarchies, networks and clustering in financial markets (2021)
  101. Domain adaptive multi-modality neural attention network for financial forecasting (2020)
  102. Multi-agent optimization of the intermodal terminal main parameters by using AnyLogic simulation platform: Case study on the Ningbo-Zhoushan Port (2021)
  103. Agentverse: Facilitating multi-agent collaboration and exploring emergent behaviors (2023)
  104. Dependence structure and dynamic connectedness between green bonds and financial markets: Fresh insights from time-frequency analysis before and during COVID-19 pandemic (2022)
  105. Reinforcement learning for quantitative trading (2023)
  106. Multi-omics data integration using ratio-based quantitative profiling with Quartet reference materials (2024)
  107. Prediction and portfolio optimization in quantitative trading using machine learning techniques (2018)
  108. Ai in finance: challenges, techniques, and opportunities (2022)
  109. FinBrain: when finance meets AI 2.0 (2019)
  110. Adapting financial rationality: Is a new paradigm emerging? (2014)
  111. Back-testing portfolio risk management strategies (2021)
  112. The ABCs of hedge funds: Alphas, betas, and costs (2011)
  113. Local pattern transformation based feature extraction techniques for classification of epileptic EEG signals (2017)
  114. The anatomy of the Korean general trading company (1984)
  115. Scaling vision with sparse mixture of experts (2021)
  116. Research on graph neural network in stock market (2022)
  117. Innovations in financial management: Recursive prediction model based on decision trees (2020)
  118. Uni-MoE: Scaling Unified Multimodal LLMs with Mixture of Experts (2024) — link
  119. Forecasting the Chinese stock volatility across global stock markets (2019)
  120. Multi-asset portfolio optimization and out-of-sample performance: an evaluation of Black–Litterman, mean-variance, and na "\i ve diversification approaches (2017)
  121. When MOE Meets LLMs: Parameter Efficient Fine-tuning for Multi-task Medical Applications (2024)
  122. On dynamic investment strategies (2000)
  123. Towards llm-guided causal explainability for black-box text classifiers (2024)
  124. FINANCIAL REPORTING TIMELINESS: A SCOPE REVIEW OF CURRENT LITERATURE (2023)
  125. Towards Greener LLMs: Bringing Energy-Efficiency to the Forefront of LLM Inference (2024) — link
  126. Price discovery among SSE 50 Index-based spot, futures, and options markets (2019)
  127. Next-gpt: Any-to-any multimodal llm (2023) — link
  128. Improving VWAP strategies: A dynamic volume approach (2008)
  129. Fintech Index Prediction Based on RF-GA-DNN Algorithm (2021)
  130. Mean squared error of empirical predictor (2004)
  131. Hedge fund due diligence: A source of alpha in a hedge fund portfolio strategy (2008)
  132. Predictive data mining in clinical medicine: current issues and guidelines (2008)
  133. ThinkPatterns-21k: A Systematic Study on the Impact of Thinking Patterns in LLMs (2025) — link
  134. FinMME: Benchmark Dataset for Financial Multi-Modal Reasoning Evaluation (2025) — link
  135. Large Language Model Agent: A Survey on Methodology, Applications and Challenges (2025) — link
  136. Sparse Causal Discovery with Generative Intervention for Unsupervised Graph Domain Adaptation (2025) — link
  137. Stock movement prediction from tweets and historical prices (2018)
  138. A survey of large language models (2023) — link
  139. Bloom: A 176b-parameter open-access multilingual language model (2023) — link
  140. Scaling instruction-finetuned language models (2022) — link
  141. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (2020)
  142. AlphaPortfolio: Direct Construction Through Deep Reinforcement Learning and Interpretable AI (2021) — link
  143. Stock Alpha Mining Based On Genetic Algorithm (2019) — link
  144. Long Short-Term Memory (1997)
  145. Attention is All You Need (2017)
  146. XGBoost: A Scalable Tree Boosting System (2016)
  147. LightGBM: A Highly Efficient Gradient Boosting Decision Tree (2017)
  148. REST: Relational Event-driven Stock Trend Forecasting (2021)
  149. HIST: A Graph-based Framework for Stock Trend Forecasting via Mining Concept-Oriented Shared Information (2021) — link
  150. Multimodal Gen-AI for Fundamental Investment Research — → 2401.06164
  151. Modal-Adaptive Knowledge-Enhanced Graph-Based Financial Prediction from Monetary Policy Conference Calls with LLM (2024) — link
  152. Revolutionizing Finance with LLMs: An Overview of Applications and Insights (2024) — link
  153. A Multimodal Foundation Agent for Financial Trading: Tool-Augmented, Diversified, and Generalist (2024)
  154. 101 formulaic alphas (2016)
  155. Asset returns in deep learning methods: An empirical analysis on SSE 50 and CSI 300 (2020)
  156. Rethinking the Bounds of LLM Reasoning: Are Multi-Agent Discussions the Key? (2024) — link
  157. Dynamic ReLU (2020)
  158. Information Coefficient as a Performance Measure of Stock Selection Models (2020) — link
  159. AlphaAgent: LLM-Driven Alpha Mining with Regularized Exploration to Counteract Alpha Decay (2025) — link
  160. Learning to Generate Explainable Stock Predictions using Self-Reflective Large Language Models — → 2402.03659
  161. The information ratio (1998)
  162. Language Models are Few-Shot Learners (2020) — link
  163. Large Language Models in Finance (FinLLMs) (2025) — link
  164. Large Language Models in Finance: A Survey (2024) — link
  165. BloombergGPT: A Large Language Model for Finance (2023) — link
  166. Integrating Stock Features and Global Information via Large Language Models for Enhanced Stock Return Prediction (2023) — link
  167. Can LLM-based Financial Investing Strategies Outperform the Market in Long Run? (2025) — link
  168. MAPS: Multi-Agent reinforcement learning-based Portfolio management System. (2020) — link
  169. Robust Market Making via Adversarial Reinforcement Learning (2020) — link
  170. When AI Meets Finance (StockAgent): Large Language Model-based Stock Trading in Simulated Real-world Environments (2024) — link
  171. FinCon: A Synthesized LLM Multi-Agent System with Conceptual Verbal Reinforcement for Enhanced Financial Decision Making (2024) — link
  172. FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design (2023) — link
  173. Proximal Policy Optimization Algorithms (2017) — link

10 of 181 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.