Paper: arXiv 2602.14670

Authors: Yanlong Wang, Jian Xu, Hongkang Zhang, Shao-Lun Huang, Danny Dongning Sun, Xiao-Ping Zhang

Abstract

Formulaic alpha factor mining is a critical yet challenging task in quantitative investment, characterized by a vast search space and the need for domain-informed, interpretable signals. However, finding novel signals becomes increasingly difficult as the library grows due to high redundancy. We propose FactorMiner, a lightweight and flexible self-evolving agent framework designed to navigate this complex landscape through continuous knowledge accumulation. FactorMiner combines a Modular Skill Architecture that encapsulates systematic financial evaluation into executable tools with a structured Experience Memory that distills historical mining trials into actionable insights (successful patterns and failure constraints). By instantiating the Ralph Loop paradigm – retrieve, generate, evaluate, and distill – FactorMiner iteratively uses memory priors to guide exploration, reducing redundant search while focusing on promising directions. Experiments on multiple datasets across different assets and Markets show that FactorMiner constructs a diverse library of high-quality factors with competitive performance, while maintaining low redundancy among factors as the library scales. Overall, FactorMiner provides a practical approach to scalable discovery of interpretable formulaic alpha factors under the “Correlation Red Sea” constraint.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper introduces FactorMiner, an innovative LLM-agent framework that uses experience memory and modular skills to solve the correlation decay problem in formulaic alpha discovery. While heavy on engineering design and empirical validation across multiple financial datasets, it relies less on deep mathematical derivations, making it a strong empirical contribution.

Research Flowchart

  flowchart TD
    A["Research Goal: Discover Diverse, Interpretable Alpha Factors under 'Correlation Red Sea'"] --> B["Data Inputs: Multi-Asset Financial Market Datasets"]
    B --> C["Ralph Loop Paradigm: Retrieve, Generate, Evaluate, Distill"]
    
    subgraph C ["Ralph Loop Paradigm"]
        C1["Modular Skill Architecture<br/>(Executable Financial Evaluation Tools)"]
        C2["Structured Experience Memory<br/>(Patterns & Constraints)"]
    end

    C --> D["Iterative Search & Knowledge Distillation"]
    D --> E["Outcomes: Low Redundancy, High-Quality Alpha Factor Library"]

Cited References (55)

  1. 101 Formulaic Alphas (2016) — link
  2. Genetic Programming: On the Programming of Computers by Means of Natural Selection (1992)
  3. Is Technical Analysis in the Foreign Exchange Market Profitable? A Genetic Programming Approach (1997) — link
  4. Using Genetic Algorithms to Find Technical Trading Rules (1999) — link
  5. Geometric Semantic Genetic Programming (2012) — link
  6. A Field Guide to Genetic Programming (2008) — link
  7. Empirical Asset Pricing via Machine Learning (2020) — link
  8. Taming the Factor Zoo: A Test of New Factors (2020) — link
and 47 more
  1. Shrinking the Cross-Section (2020) — link
  2. \ldots and the Cross-Section of Expected Returns (2016) — link
  3. Lucky Factors (2021) — link
  4. ReAct : Synergizing Reasoning and Acting in Language Models (2022) — link
  5. Reflexion: Language Agents with Verbal Reinforcement Learning (2023) — link
  6. Generative Agents: Interactive Simulacra of Human Behavior (2023) — link
  7. Common Risk Factors in the Returns on Stocks and Bonds (1993) — link
  8. On Persistence in Mutual Fund Performance (1997)
  9. Liquidity Sentiments (2019)
  10. Qlib : An AI -oriented Quantitative Investment Platform (2020) — link
  11. The Proof and Measurement of Association between Two Things (1904)
  12. AI Feynman : A Physics-Inspired Method for Symbolic Regression (2020) — link
  13. Symbolic Regression via Deep Reinforcement Learning Enhanced Genetic Programming Seeding (2021)
  14. AlphaAgent : LLM -Driven Alpha Mining with Regularized Exploration to Counteract Alpha Decay (2025) — link
  15. DreamCoder : Growing Generalizable, Interpretable Knowledge with Wake-Sleep Bayesian Program Learning (2021)
  16. LILO : Learning Interpretable Libraries by Compressing and Documenting Code (2023) — link
  17. Competition-Level Code Generation with AlphaCode (2022)
  18. CodeRL : Mastering Code Generation through Pretrained Models and Deep Reinforcement Learning (2022)
  19. Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks (2017) — link
  20. gplearn : Genetic Programming in Python — link
  21. On First-Order Meta-Learning Algorithms (2018) — link
  22. Meta-Learning Acquisition Functions for Transfer Learning in Bayesian Optimization (2020)
  23. Meta-Reinforcement Learning of Structured Exploration Strategies (2018) — link
  24. Continual Lifelong Learning with Neural Networks: A Review (2019) — link
  25. Model-Free Episodic Control (2016) — link
  26. Toolformer: Language Models Can Teach Themselves to Use Tools (2023) — link
  27. MRKL Systems: A Modular, Neuro-Symbolic Architecture that Combines Large Language Models, External Knowledge Sources and Discrete Reasoning (2022) — link
  28. HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face (2023) — link
  29. API -Bank: A Comprehensive Benchmark for Tool-Augmented LLM s (2023) — link
  30. Gorilla: Large Language Model Connected with Massive API s (2023) — link
  31. Tool LLM : Facilitating Large Language Models to Master 16000+ Real-world API s (2023) — link
  32. MemGPT: Towards LLM s as Operating Systems (2024) — link
  33. Voyager: An Open-Ended Embodied Agent with Large Language Models (2023) — link
  34. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead (2019) — link
  35. AlphaForge : A Framework to Mine and Dynamically Combine Formulaic Alpha Factors (2025) — link
  36. Generating Synergistic Formulaic Alpha Collections via Reinforcement Learning — → 2306.12964
  37. QuantFactor REINFORCE : Mining Steady Formulaic Alpha Factors with Variance-Bounded REINFORCE (2025) — link
  38. AlphaEvolve : A Learning Framework to Discover Novel Alphas in Quantitative Investment (2021)
  39. A Time Series is Worth 64 Words: Long-term Forecasting with Transformers (2023) — link
  40. Chronos : Learning the Language of Time Series (2024) — link
  41. LightGBM : A Highly Efficient Gradient Boosting Decision Tree (2017) — link
  42. Chronos-2 : From Univariate to Universal Forecasting (2025) — link
  43. Does Academic Research Destroy Stock Return Predictability? (2016) — link
  44. How and When Are High-Frequency Stock Returns Predictable? (2025) — link
  45. Common Risk Factors in Cryptocurrency (2022) — link
  46. Cross-sectional return dispersion and volatility prediction (2019) — link
  47. The information content of stock markets: why do emerging markets have synchronous stock price movements? (2000) — link

1 of 55 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.