Updated
Updated · arxiv.org · Oct 7STOCK-JEPA: Prior-Anchored Latent Revision Representation Learning in Equity Markets
1 articles · Updated · arxiv.org · Oct 7Summary
- Researchers have introduced Stock-JEPA, a new framework that combines financial priors with deep learning for equity market representation.
- Stock-JEPA learns predictable revisions to financial model predictions, outperforming 13 strong baselines in both China and U.S. stock markets.
- This approach improves return forecasting and portfolio management, offering interpretable and robust representations in noisy financial environments.