> In recent years, transformer-based models have gained prominence in multivariate long-term time series forecasting<p>Prominence, yes. But are they generally better than non-deep learning models? My understanding was that this is not the case, but I don't follow this field closely.