We forecast the U.S. Treasury yield curve using the Dynamic Nelson-Siegel model in a state-space framework. A Kalman Filter estimates latent factors, and a GARCH(1,1) extension captures time-varying volatility. The KF-GARCH model improves uncertainty quantification, especially for long-term yields.