This report synthesises findings from 15 peer-reviewed papers addressing the following research question: How does the robustness of multi-scale contrastive learning in GNNs compare to traditional supervised learning methods when evaluated on the adversarially perturbed Reddit and Amazon datasets using. Abstract Semi-supervised learning is the branch of machine learning concerned with using labelled as well as unlabelled data to perform certain learning tasks. Conceptually situated between supervised and unsupervised learning, it permits harnessing the large amounts of. 7 claims were extracted from source literature; 7 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 9.2/10. This report is a machine-generated literature synthesis and does not constitute original research.Research goal: How does the robustness of multi-scale contrastive learning in GNNs compare to traditional supervised learning methods when evaluated on the adversarially perturbed Reddit and Amazon datasets using F1 score metrics?Autonomous literature synthesis. Automated review score: 9.2/10. Full text and citation available at Assignee Research.