CosmosAlpha improves multi-objective factor search with graph exploration

CosmosAlpha: Multi-Relational Graph Search with Pareto-Guided Exploration for LLM-Driven Alpha Mining

Artificial IntelligenceMachine Learning

Summary

Finding useful signals in noisy financial data is hard because it involves balancing many goals and avoiding repeated tries of similar ideas. The authors created CosmosAlpha, which builds a network of potential signals showing how they relate based on how they were made and how similar they are. This network helps the system learn from related signals instead of treating each one alone, improving the search for valuable financial formulas. It also balances multiple goals, like quality and diversity, to find better investment signals.

What this means in practice

  • For quantitative finance teams: Improve discovery of predictive financial factors by using graph-based search that balances multiple objectives beyond typical lineage tracking.$Commercial implications: Enables enhanced alpha factor generation tools for asset managers and hedge funds seeking interpretable investment signals.
  • For data science teams in finance: Implement relational graph structures to better share information between related candidate features and optimize factor selection processes.

Authors

Zhan-He Wang, Zi-Yang Lang, Yi-Yang Li, Qing-Yu Lan, Yun-Song Piao

Abstract

Discovering predictive alpha factors from noisy financial data is a fundamental yet challenging problem in quantitative finance. LLM-driven formulaic alpha mining offers an attractive paradigm by combining interpretable symbolic expressions with iterative generation and refinement guided by quantitative feedback. However, existing generation-centric and tree-based search strategies primarily organize factors through generation lineage, leaving structural and statistical relationships across search branches underutilized. Consequently, information remains localized along parent--child trajectories, which can confine exploration to redundant regions. Moreover, effective factor discovery must account for multiple distinct objectives that cannot be adequately captured by a single search criterion. We introduce CosmosAlpha, a graph-based LLM factor-mining framework that represents discovered factors as a multi-relational graph encoding generation lineage, expression-structure similarity, and cross-sectional exposure correlation. A Graph Flow mechanism propagates search utilities across the graph, allowing each factor to be assessed using relational evidence from nearby factors rather than in isolation. Pareto-guided selection jointly considers quality, fertility and diversity, explicitly modeling their trade-offs within a multi-objective search framework. Across CSI 300 and CSI 500, CosmosAlpha-Graph achieves the best mean rank over seven predictive and portfolio metrics and outperforms its controlled CosmosAlpha-MCTS counterpart on every reported metric, showing a consistent advantage for relational graph search in LLM-driven factor mining. The source code is available at https://github.com/wzhmtldswz/CosmosAlpha