The field of artificial intelligence is moving towards increased interpretability, with a focus on understanding the decision-making processes of deep neural networks. Recent developments have introduced new methods for interpreting matching-based few-shot semantic segmentation models, mechanistic interpretability for algorithmic understanding of neural networks, and explainable visual anomaly detection via concept bottleneck models. These advances have the potential to support a more scientific understanding of machine learning systems and improve trust in AI applications. Notable papers include: Matching-Based Few-Shot Semantic Segmentation Models Are Interpretable by Design, which introduces the Affinity Explainer approach for interpreting few-shot segmentation models. Unboxing the Black Box: Mechanistic Interpretability for Algorithmic Understanding of Neural Networks proposes a unified taxonomy of mechanistic interpretability approaches and provides a detailed analysis of key techniques. Explainable Visual Anomaly Detection via Concept Bottleneck Models extends concept bottleneck models to the visual anomaly detection setting, providing human-interpretable descriptions of anomalies.