A Study on Correlation Between WAT and CP Yields Based on Deep Learning Methods

Wafer Acceptance Testing (WAT) parameters reflect Chip Probing (CP) yields, prompting extensive discussions on their correlation. This paper proposes an effective process to explore this relationship. Based on theoretical analysis, a classifier is developed to accurately identify CP yields reflected by WAT, with precision exceeding 95%. In the model training phase, deep learning methods are used to address challenges like data imbalance and missing labels, reducing prediction errors by 71% compared to the baseline. Experiments show that this process provides more accurate and interpretable insights into the WAT-CP yield relationship for mass-produced ICs.