The field of unsupervised anomaly detection is witnessing significant developments, with a focus on improving the robustness and accuracy of detection methods. Researchers are exploring new approaches to address the limitations of traditional methods, such as the assumption of normality and the presence of contaminated data. Novel techniques, including contrastive learning and diffusion models, are being proposed to enhance the detection of anomalies in complex data. These methods aim to preserve the normal regions and selectively alter anomalous areas, leading to more effective detection and localization of anomalies. Notably, some papers are proposing innovative architectures, such as two-branch decoder blocks and fuzzy cluster-aware contrastive clustering frameworks, to learn normal patterns and reconstruct them progressively. Noteworthy papers include:

• RoCA, which achieves state-of-the-art performance on both univariate and multivariate datasets by fusing one-class classification and contrastive learning.
• Deviation correction diffusion, which introduces a reformulation of the standard diffusion model to preserve normal regions and encourage transformations exclusively on anomalous areas.
• Omni-AD, which proposes a two-branch decoder block to learn global and local features for multi-class anomaly detection.
• Fuzzy Cluster-Aware Contrastive Clustering, which jointly optimizes representation learning and clustering for time series data.