The field of time series anomaly detection is rapidly evolving, with a focus on developing innovative methods to identify unusual patterns in complex data. Recent research has emphasized the importance of adapting to concept drift, leveraging pre-training strategies, and exploiting domain-specific characteristics to improve detection accuracy. Notably, the use of BERT-style pretraining and transformer-based architectures has shown promising results in various applications, including battery fault detection and stock price prediction. Furthermore, the development of novel clustering methods and similarity learning frameworks has enhanced the ability to differentiate between abnormal changes and varying normal behaviors. Overall, the field is moving towards more robust and adaptive anomaly detection approaches that can effectively handle complex temporal dependencies and non-stationary data. Noteworthy papers include: BatteryBERT, which proposes a novel framework for battery fault detection using BERT-style pretraining and achieves state-of-the-art results. AnDri, which introduces a system for anomaly detection in the presence of concept drift and develops a new clustering method called Adjacent Hierarchical Clustering. mTSBench, which provides a comprehensive benchmark for multivariate time series anomaly detection and model selection, highlighting the importance of adaptive model selection and robust detection methods.