The field of probabilistic forecasting is moving towards a more nuanced understanding of uncertainty quantification, with a focus on developing models that can accurately capture and represent uncertainty. This is being achieved through the development of new models and techniques, such as thermodynamically informed neural networks and probabilistic scenarios, which are designed to address the limitations of traditional sampling-based approaches. Notably, research is showing that models that incorporate both data and model uncertainty can improve both point and probabilistic forecasting. Additionally, the use of conformal prediction is being explored as a means of capturing uncertainty. Overall, the field is advancing towards more accurate and reliable probabilistic forecasting models. Noteworthy papers include:

• A study on uncertainty quantification in statistical and deep learning models, which found that incorporating both data and model uncertainty can improve forecasting performance.
• A paper introducing a new paradigm for probabilistic forecasting, which operates by directly producing a finite set of scenario-probability pairs, and achieves state-of-the-art results across multiple benchmark datasets.