AlphaRLM trains language models to create market prediction formulas
AlphaRLM: Aligning Language Models for Literature-Grounded Symbolic Factor Discovery
Artificial IntelligenceComputation and LanguageMachine Learning
Summary
Finding formulas that predict financial markets is tough because the search space is huge and noisy. The authors present AlphaRLM, which teaches language models using reinforcement learning to build executable formulas based on ideas from financial literature. This approach aligns formula creation with real market performance by rewarding expressions that work well on market data. Their method outperforms other smart approaches in predicting stock market returns across different markets. This shows promise for combining financial texts and market signals to discover useful trading strategies.
What this means in practice
- •For quantitative finance teams: Create new trading formulas automatically by letting language models learn from financial research texts and market feedback.
- •For financial software developers: Develop tools that unify financial literature and market data to generate and test stock prediction expressions.
- •For algorithmic trading firms: Build market-aligned factor generation systems that improve excess return performance across multiple stock indices.$Commercial implications: The paper enables selling enhanced factor creation tools for trading firms looking for better automated market signals.
Authors
Jie Huang, Yuecheng Hong, Yao Chen, Hao Zhong, Linfeng Xiao, Xudong Liu, Hao Xia, Jianxuan Xin
Abstract
Formulaic alpha mining seeks to discover executable expressions that capture predictive market signals in a vast, noisy search space. Recent LLM-based approaches use prompting, iterative search, or seed-factor evolution, leaving direct factor construction through policy learning less explored. To address this, we introduce AlphaRLM, a reinforcement-learning framework that trains LLMs to construct executable symbolic factors from financial literature without input factor expressions. AlphaRLM incorporates Literature-Guided Operator Assembly (LEGO), which conditions factor construction on literature-derived ideas and executable operators, providing semantic priors for exploration. It further aligns the generation policy with market feedback through executable-validity and grouped-return-separation rewards that capture non-monotonic structures across factor quantiles. Extensive experiments on CSI 300, CSI 1000, and S&P 500 show that AlphaRLM outperforms strong factor-mining and LLM-based baselines in annualized excess return. Notably, the 8B model achieves strong cross-market performance, highlighting the promise of literature-conditioned, market-aligned factor generation.