Researchers introduce XAlpha, a memory-driven AI quant researcher designed to automate the end-to-end process of alpha discovery in financial markets. The system addresses the limitations of existing methods by functioning as a continuous loop that absorbs external knowledge and learns from accumulated feedback.
XAlpha utilizes a multi-source research memory system integrating report-grounded financial knowledge with discovery feedback. Its architecture includes three core components: a Macro Brain for planning research themes, a Micro Brain for transforming hypotheses into executable code, and a Cross Brain for consolidating empirical outcomes to guide future exploration.
Experiments conducted on the CSI300 index demonstrate that XAlpha achieves stronger overall alpha discovery performance compared to representative baselines.