Future-Aligned Soft Contrastive Learning for Asset Retrieval

FASCL is a representation learning framework that uses pairwise future return correlations as continuous supervision targets for asset retrieval. Tested on 5,631 US-listed securities, the method outperforms 14 baselines in predicting correlated future trajectories.