Advanced Defect Discovery addresses the throughput limitation of Review SEM tools by selecting a subset of defects detected by optical inspection that maximizes learning. Given a fixed sample budget, it applies machine learning to choose defects whose backgrounds resemble known high-risk patterns while also including dissimilar defects that may reveal new defect types.

The workflow draws on the Printed Pattern Database, the Design Decomposition Database, and machine learning models to balance capture of known defects of interest against discovery of new defect categories. It processes KLARF and optical patch images to produce a sample plan for SEM review, helping fabs learn more from each inspection cycle.