Tag
This paper introduces a lightweight pre-encoder gate mechanism for Transformer-based time-series forecasting models to regulate covariate admission, with experiments showing competitive performance against baselines on multiple datasets.