The field of data assimilation is moving towards the development of innovative methods that combine machine learning techniques with traditional ensemble-based approaches. These new methods aim to improve the accuracy and efficiency of state estimation in complex systems, while also addressing the challenges of nonlinear dynamics and high-dimensional problems. Notable advancements include the use of neural networks to learn correction terms for ensemble-based methods, the application of score-based diffusion models for nonlinear filtering problems, and the development of structure-preserving sequential data assimilation frameworks. These innovations have the potential to significantly improve the performance of data assimilation systems in various applications, including wildfire spread prediction, sea ice velocity and concentration forecasting, and solar chimney modeling. Noteworthy papers include: Small Ensemble-based Data Assimilation, which proposes a novel machine learning-based approach that achieves higher accuracy than traditional ensemble Kalman filter methods with negligible additional computational cost. Fire-EnSF, which demonstrates the effectiveness of the Ensemble Score Filter for wildfire spread data assimilation, providing superior accuracy, stability, and computational efficiency. IEnSF, which develops an iterative ensemble score filter that substantially reduces the error in posterior score estimation in nonlinear settings, improving the accuracy of tracking high-dimensional dynamical systems.