AgenticAD: Learning to Route, Observe, and Detect for Sample-Adaptive Industrial Anomaly Detection

Multi-class industrial anomaly detection faces strong sample-wise variation in detector preference, defect scale, and appearance. We introduce AgenticAD, a sample-adaptive route–observe–detect framework for supervised control of frozen anomaly detectors. Supervision is used only to train the lightweight routing and observation controllers (using pixel masks for pixel-level control); the anomaly detectors remain frozen. Stage I routes each image to one of four complementary experts using task-specific utilities derived from all expert outcomes. Stage II performs a bounded observe–detect loop conditioned on the initial score, heatmap, and observation history. At each step, the controller either stops or applies a heatmap-guided crop or photometric transformation; a generic local detector analyzes the resulting view and aligns its evidence to the original image. Fixed aggregation rules convert detector outputs into final scores. On IND-COM, comprising five industrial datasets and 78 object categories, Stage I improves image-level AUROC/AP from $79.82/91.36$ for category-best routing to $84.75/93.51$, and Stage II further reaches $91.34/96.34$. For pixel localization, Stage I reaches $92.08/51.22$ AUROC/AP, and Stage II reaches $92.45/51.53$, compared with $91.24/49.94$ for category-best routing. Overall, AgenticAD offers a practical way to adapt detector selection and inspection effort to the diversity of real-world industrial products and anomalies.