Online Irregular Multivariate Time Series Forecasting via Uncertainty-Driven Dual-Expert Calibration

Researchers introduced Under-Cali for online irregular multivariate time series forecasting.

It uses uncertainty estimation, dual-expert calibration, and adaptive routing to split high- and low-uncertainty samples.

The base forecasting model stays frozen while a lightweight calibration module handles efficient online adaptation.

On IMTS benchmarks, the method delivered consistent accuracy gains with low computational cost.