Skip to main content
Search

Xiao Wang

1 related articles

Subscribe to updates
  1. arXiv

    WxFM-XL adapts univariate foundation models with band-specific error-correlation graphs and dynamic fusion, taking best MSE in all 16 multi-station weather settings

    WxFM-XL decomposes univariate time-series foundation model forecasts into low, mid, and high frequency components, builds a cross-station error-correlation prior graph per band, dynamically fuses it with a static spatial correlation graph via learnable gates, and applies a lightweight graph residual adapter to correct forecasts over a frozen backbone; across Hunan, French MeteoNet, and global hourly station datasets, a WxFM-XL variant attains the best MSE in all 16 dataset-variable-horizon settings.

You’re all caught up