Princeton, Ant Group, and Stanford Researchers Introduce AQuA for Autonomous Factor Discovery in Quantitative Finance

Based on a report from MarkTechPost, researchers from Princeton University, Ant Group, and Stanford University have introduced AQuA, a two-component agentic framework for autonomous factor discovery and model development in quantitative finance. The framework separates research agents from a fixed evaluator to solve the problem of irreproducible backtesting results caused by methodological errors in traditional quantitative research. The first component focuses on discovering symbolic alpha factors on 5-minute cryptocurrency data, using a six-agent pipeline. After 20 research cycles, the Spearman IC was verified to reach approximately 0.190. The second component targets intraday U.S. stock data, predicting stock returns over the next 30 minutes. It employs a hybrid model architecture that includes a multi-scale 1D convolutional frontend and a configurable backbone network. The researchers describe this design as "asymmetric freedom," where agents explore freely within their domain-specific language, while the evaluator remains outside the adaptation surface.