Problem
Research questions and friction points this paper is trying to address.
Automated Factor Mining
Benchmark
Portfolio Construction
Financial Signals
Factor Evaluation
FactorBench: A Portfolio-Aware Benchmark for Automated Factor Mining
🤖 AI Summary
This study investigates whether automated factor mining paradigms can yield signals with superior generalizability and economic value. To this end, it constructs a portfolio-aware benchmark and introduces the first shared contract bridging heterogeneous algorithms with portfolio construction. Leveraging techniques such as genetic programming, reinforcement learning, generative models, and LLM-based agents, the work systematically evaluates approximately five thousand factors—derived from nine methodological categories across five markets—at multiple hierarchical levels. By unifying the assessment of factor effectiveness, uniqueness, and portfolio-level performance, this research demonstrates that no single paradigm consistently dominates. Ultimately, it establishes a standardized evaluation framework for automated factor mining in quantitative finance.
📝 Abstract
Factor mining seeks to discover signals from financial data that predict future asset returns and guide portfolio construction. Automated factor mining now spans genetic programming, reinforcement learning, generative models, and large language model agents. Yet it remains unclear whether advances across these paradigms yield more generalizable, distinct, and economically useful financial signals. We introduce FactorBench, a portfolio-aware benchmark comparing roughly five thousand mined factors from nine automated mining methods across five equity markets. A shared data and evaluation contract supports both symbolic expressions and executable Python factors, connecting heterogeneous discovery algorithms to common signal combination and portfolio construction procedures. FactorBench traces the outputs of mining systems across three levels: factor validity, temporal generalization, and predictiveness beyond measured risk and style exposures; within- and across-method pool distinctness, including similarity to the benchmark Alpha101; and composite-signal quality and after-cost long-only and long--short portfolio performance. After systematically assessing whether advances in factor mining translate into signal quality and portfolio performance, FactorBench finds that no paradigm consistently dominates.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Automated Factor Mining
Portfolio-Aware Benchmark
Heterogeneous Evaluation
Temporal Generalization
Signal Distinctness
🔎 Similar Papers
No similar papers found.
