NCPNET is a novel conformal prediction framework designed for temporal graph neural networks that addresses the limitations of existing methods focused on static graphs. By introducing a diffusion-based non-conformity score and an efficiency-aware optimization algorithm, NCPNET effectively captures temporal uncertainties and enhances computational efficiency, achieving significant improvements in prediction set sizes across various real-world temporal graph datasets.

Large Language Models (LLMs) can significantly enhance data annotation but often produce incorrect labels due to uncertainty. This work proposes a candidate annotation paradigm that encourages LLMs to provide multiple possible labels, utilizing a teacher-student framework called CanDist to distill these annotations into unique labels for downstream tasks. Experiments demonstrate the effectiveness of this method across various text classification challenges.