A dataset and benchmark for forecasting real-world time series with cause-driven irregularities and multimodal observations.

Contimask is the first post-hoc explainer for irregular time series models capable of explaining models that besides the value of observed data factor in missingness or time intensity.