Paper: arXiv 2603.20247

Authors: Zhangyuhua Weng, Shengli Zhang, Taotao Wang, Yihan Xia

Abstract

Factor investing is ultimately grounded in market logic - the latent mechanism behind observed alpha factors that explains why they should persist across assets and regimes. However, recent factor mining prioritizes factor discovery over logic discovery, producing complex alpha factors with unclear rationale, while market logic remains largely handcrafted and difficult to scale. To address this challenge, we propose AlphaLogics, a market logic-driven multi-agent system for factor mining. AlphaLogics consists of three key components: (i) Market Logic Mining: reverse-extracting market logic from historical factor libraries to construct an initial market logic library; (ii) Factor Generation and Optimization: using new market logics generated in (i) to guide factor generation, and optimizing factors with backtesting feedback; and (iii) Market Logic Generation and Optimization: generating new market logics conditioned on the initial market logic library, and refining each market logic by aggregating the backtest outcomes of its guided factors, continuously refreshing the library. Experiments on CSI 500 and S&P 500 show that AlphaLogics consistently improves predictive metrics and risk-adjusted returns over representative baselines, while producing a market logic library that remains empirically useful for guiding further factor discovery.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper introduces a novel LLM-based multi-agent framework (AlphaLogics) that closes the loop between factor generation and market logic extraction. It relies primarily on empirical agent workflows and backtesting across major equity indices rather than complex mathematical proofs, making it a strong data and methodology-driven contribution.

Research Flowchart

  flowchart TD
    A["Research Goal: Discover Scalable & Interpretable Alpha Factors Grounded in Market Logic"] --> B["Data Inputs: Historical Factor Libraries, CSI 500 & S&P 500 Equities Data"]
    
    subgraph C ["AlphaLogics Multi-Agent System"]
        C1["1. Market Logic Mining: Reverse-extract logic from historical factors"]
        C2["2. Market Logic Generation & Refinement: Generate new logics & update library via aggregated feedback"]
        C3["3. Factor Generation & Optimization: Mine factors guided by logic & optimize via backtest feedback"]
        C1 --> C2 --> C3
    end

    B --> C
    C --> D["Key Outcomes: Superior Predictive Metrics & Risk-Adjusted Returns"]
    C --> E["Output: Interpretable Market Logic Library for Ongoing Factor Discovery"]

Cited References (44)

  1. Empirical asset pricing via machine learning (2020)
  2. Autoalpha: an efficient hierarchical evolutionary algorithm for mining alpha factors in quantitative investment (2020) — link
  3. A five-factor asset pricing model (2015)
  4. Guotai Junan 191 Alpha Factor Library: Technical Report and Implementation Code (2025)
  5. Alphaevolve: A learning framework to discover novel alphas in quantitative investment (2021)
  6. 101 formulaic alphas (2016)
  7. Alpha Mining and Enhancing via Warm Start Genetic Programming for Quantitative Investment — → 2412.00896
  8. Alphaforge: A framework to mine and dynamically combine formulaic alpha factors (2025)
and 36 more
  1. Beyond the black box: Interpretability of llms in finance — → 2505.24650
  2. Why are we using black box models in AI when we don’t need to? A lesson from an explainable AI competition (2019)
  3. Ploutos: Towards interpretable stock movement prediction with financial large language model — → 2403.00782
  4. Alpha-gpt: Human-ai interactive alpha mining for quantitative investment (2025)
  5. FactorMAD: A Multi-Agent Debate Framework Based on Large Language Models for Interpretable Stock Alpha Factor Mining (2025)
  6. A survey of large language models for financial applications: Progress, prospects and challenges — → 2406.11903
  7. Integrating stock features and global information via large language models for enhanced stock return prediction — → 2310.05627
  8. Volatility spillover effect between Pakistan and Shanghai Stock Exchanges using copula and dynamic conditional correlation model (2023)
  9. Model Validation for Forecasting: Out-of-sample vs. Out-of-time (2025)
  10. Common risk factors in the returns on stocks and bonds (1993)
  11. Choosing factors (2018)
  12. From factor models to deep learning: Machine learning in reshaping empirical asset pricing — → 2403.06779
  13. Deep Learning for Conditional Asset Pricing Models — → 2509.04812
  14. Deep learning in characteristics-sorted factor models (2024)
  15. The Uncertainty of Machine Learning Predictions in Asset Pricing (2025) — link
  16. Automate strategy finding with llm in quant investment — → 2409.06289
  17. AlphaAgent: LLM-driven alpha mining with regularized exploration to counteract alpha decay (2025)
  18. Qlib: An ai-oriented quantitative investment platform (2020) — link
  19. BaoStock: A Tool for Obtaining Historical Data of China Stock Market (2024)
  20. yfinance: Download Market Data from Yahoo! Finance’s API (2024)
  21. Long short-term memory (2012)
  22. Empirical evaluation of gated recurrent neural networks on sequence modeling (2014) — link
  23. Multilayer perceptron (MLP) (2017)
  24. Lightgbm: A highly efficient gradient boosting decision tree (2017)
  25. XGBoost: A Scalable Tree Boosting System (2016)
  26. Attention is all you need (2017)
  27. Learning multiple stock trading patterns with temporal routing adaptor and optimal transport (2021)
  28. R&D-Agent-Quant: A Multi-Agent Framework for Data-Centric Factors and Model Joint Optimization — → 2505.15155
  29. Gemini 2.5: Advanced Reasoning, Multimodality, and Long Context Capabilities (2025) — link
  30. Training language models to follow instructions with human feedback (2022)
  31. DeepSeek-v3 Technical Report (2024) — link
  32. Alpha158 Factor Library: Data Handler in Microsoft Qlib (2025)
  33. Alpha360 Factor Library: High-Dimensional Factor Dataset in Microsoft Qlib (2025)
  34. o3-mini: Efficient Reasoning Model for Fast and Cost-Effective Inference (2025)
  35. DeepSeek-V3.1 Release (2025)
  36. GPT-5 mini Model (2025)

9 of 44 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.