This book constitutes the refereed proceedings of the Third International Workshop on Automated Spatial and Temporal Anomaly Detection, ASTAD@AAAI 2026, held in conjunction with AAAI 2026, Singapore, on January 26, 2026.

The 9 full papers included in this book were carefully reviewed and selected from 18 submissions. They cover diverse areas, including vehicle engine diagnostics, thermal fault detection for energy storage systems, spatially conditioned diffusion models for anomaly generation, and domain-adaptive video
captioning for surveillance.

Inhalt

.- Unveiling the Flaws: A Critical Analysis of Initialization Effect on Time Series Anomaly Detection.
.- EngineAD: A Real-World Vehicle Engine Anomaly Detection Dataset.
.- Advancing Thermal Fault Diagnostics for Lithium-Ion Energy Storage Systems: An Autonomous Vision Foundation-Based Approach.
.- Leveraging Physics Anomaly Knowledge and Contrastive Learning for Region-Agnostic Landslide Prediction.
.- Domain-Adaptive Video Captioning for Surveillance Videos.
.- Anomaly-driven Reinforcement Learning.
.- Benchmarking IoT Time-Series AD with Event-Level Augmentations.
.- Bridging the Data Gap: Spatially Conditioned Diffusion Model for Anomaly Generation in Photovoltaic Electroluminescence Images.
.- Modular Segmentation-Agnostic Framework for Object-Based Change Detection in Multi-Temporal VHR Imagery.