Ask or search anything...

AlphaSAGE introduces a framework combining Relational Graph Convolutional Networks with Generative Flow Networks to discover diverse and predictive formulaic alphas. It addresses reward sparsity, structural underrepresentation, and limited diversity in automated alpha mining, demonstrating superior performance across multiple financial markets.

MASS introduces a framework for direct, end-to-end financial portfolio construction by simulating a market of heterogeneous investor agents and optimizing their collective decisions. The system consistently achieves higher excess returns and lower drawdowns compared to state-of-the-art baselines, demonstrating a multi-agent scaling effect where more agents lead to progressively better investment performance.