Probing the fracture toughness associated with microstructural interfacial features such as grain boundaries can be a challenging task, as the sampled volume and the experimental complexity limit the statistical validity of the obtained results. In this talk, a recently developed computational framework aimed at extracting the grain-boundary toughness from meso-scale experiments will be described. The proposed framework relies on the ability of a graph neural network to perform high accuracy predictions of the micro-scale material toughness, utilizing a limited size dataset. The merit of the proposed framework arises from the capacity to enhance its performance in different material systems with limited additional training on data obtained from experiments that do not require complex measurements. While initially developed for crack growth along grain boundaries, the proposed method can be extended to any kind of interface. The method's efficiency is demonstrated by introducing new crack growth rules with limited datasets (200-300 interfaces) and exploring the obtained prediction accuracy for both computational and experimental model materials.