Tag
The paper proposes the Slicing-Graphing-Alignment (SGA) method to quantify uncertainty in multi-step forecasting for time series foundation models, demonstrating superior performance and revealing an empirical scaling law.