CroSel: Cross Selection of Confident Pseudo Labels for Partial-Label Learning
CroSel proposes a method leveraging historical predictions to select confident pseudo labels for training examples, outperforming state-of-the-art methods in benchmark datasets.