Graph-structured data plays a pivotal role in modeling complex relationships. However, real-world graphs are often incomplete due to data collection and observational constraints, severely limiting the effectiveness of modern graph learning pipelines. While existing Graph Data Augmentation (GDA) methods attempt to refine graph structures for improved downstream performance, they are typically label-dependent, computationally expensive, and inherently transductive, limiting their applicability in practical scenarios. In this work, we present a novel feature-centric graph data augmentation framework that bypasses explicit structure modeling by operating directly in the embedding space. Through a self-supervised inverse masking process, our method captures latent ties between observed and complete graphs, enabling recovery of unobserved structural signals through refined node representations. To enhance robustness under noisy and sparse supervision, we introduce a message regularizer and a bootstrap strategy for effective training and generalization. Evaluated on ten graph datasets spanning multiple domains, our approach, SelfAug, consistently outperforms state-of-the-art methods in both accuracy and efficiency across inductive and cold-start settings, highlighting its potential as a scalable and generalizable solution for real-world graph learning scenarios.

Self-supervised learning (SSL) has shown great promise in graph representation learning. However, most existing graph SSL methods are developed and evaluated under a single-dataset setting, leaving their cross-dataset transferability largely unexplored and limiting their ability to leverage knowledge transfer and large-scale pretraining, factors that are critical for developing generalized intelligence beyond fitting training data. To address this gap and advance foundation model research for graphs, we present GSTBench, the first systematic benchmark for evaluating the transferability of graph SSL methods. We conduct large-scale pretraining on ogbn-papers100M and evaluate five representative SSL methods across a diverse set of target graphs. Our standardized experimental setup decouples confounding factors such as model architecture, dataset characteristics, and adaptation protocols, enabling rigorous comparisons focused solely on pretraining objectives. Surprisingly, we observe that most graph SSL methods struggle to generalize, with some performing worse than random initialization. In contrast, GraphMAE, a masked autoencoder approach, consistently improves transfer performance. We analyze the underlying factors that drive these differences and offer insights to guide future research on transferable graph SSL, laying a solid foundation for the ''pretrain-then-transfer'' paradigm in graph learning. Our code is available at https://github.com/SongYYYY/GSTBench.

Link Prediction (LP) is a critical task in graph machine learning. While Graph Neural Networks (GNNs) have significantly advanced LP performance recently, existing methods face key challenges including limited supervision from sparse connectivity, sensitivity to initialization, and poor generalization under distribution shifts. We explore pretraining as a solution to address these challenges. Unlike node classification, LP is inherently a pairwise task, which requires the integration of both node- and edge-level information. In this work, we present the first systematic study on the transferability of these distinct modules and propose a late fusion strategy to effectively combine their outputs for improved performance. To handle the diversity of pretraining data and avoid negative transfer, we introduce a Mixture-of-Experts (MoE) framework that captures distinct patterns in separate experts, facilitating seamless application of the pretrained model on diverse downstream datasets. For fast adaptation, we develop a parameter-efficient tuning strategy that allows the pretrained model to adapt to unseen datasets with minimal computational overhead. Experiments on 16 datasets across two domains demonstrate the effectiveness of our approach, achieving state-of-the-art performance on low-resource link prediction while obtaining competitive results compared to end-to-end trained methods, with over 10,000x lower computational overhead.

A fundamental challenge in understanding graph neural networks (GNNs) lies in characterizing their optimization dynamics and loss landscape geometry, critical for improving interpretability and robustness. While mode connectivity-a lens for analyzing geometric properties of loss landscapes-has proven insightful for other deep learning architectures, its implications for GNNs remain unexplored. This work presents the first investigation of mode connectivity in GNNs. We uncover that GNNs exhibit distinct non-linear mode connectivity, diverging from patterns observed in fully-connected networks or CNNs. Crucially, we demonstrate that graph structure, rather than model architecture, dominates this behavior, with graph properties like homophily correlating with mode connectivity patterns. We further establish a link between mode connectivity and generalization, proposing a generalization bound based on loss barriers and revealing its utility as a diagnostic tool. Our findings further bridge theoretical insights with practical implications: they rationalize domain alignment strategies in graph learning and provide a foundation for refining GNN training paradigms.

Magnetic resonance imaging (MRI) is powerful in medical diagnostics, yet high-field MRI, despite offering superior image quality, incurs significant costs for procurement, installation, maintenance, and operation, restricting its availability and accessibility, especially in low- and middle-income countries. Addressing this, our study proposes an unsupervised learning algorithm based on cycle-consistent generative adversarial networks. This framework transforms 0.3T low-field MRI into higher-quality 3T-like images, bypassing the need for paired low/high-field training data. The proposed architecture integrates two novel modules to enhance reconstruction quality: (1) an attention block that dynamically balances high-field-like features with the original low-field input, and (2) an edge block that refines boundary details, providing more accurate structural reconstruction. The proposed generative model is trained on large-scale, unpaired, public datasets, and further validated on paired low/high-field acquisitions of three major clinical MRI sequences: T1-weighted, T2-weighted, and fluid-attenuated inversion recovery (FLAIR) imaging. It demonstrates notable improvements in tissue contrast and signal-to-noise ratio while preserving anatomical fidelity. This approach utilizes rich information from publicly available MRI resources, providing a data-efficient unsupervised alternative that complements supervised methods to enhance the utility of low-field MRI.

Given the ubiquity of graph data and its applications in diverse domains, building a Graph Foundation Model (GFM) that can work well across different graphs and tasks with a unified backbone has recently garnered significant interests. A major obstacle to achieving this goal stems from the fact that graphs from different domains often exhibit diverse node features. Inspired by multi-modal models that align different modalities with natural language, the text has recently been adopted to provide a unified feature space for diverse graphs. Despite the great potential of these text-space GFMs, current research in this field is hampered by two problems. First, the absence of a comprehensive benchmark with unified problem settings hinders a clear understanding of the comparative effectiveness and practical value of different text-space GFMs. Second, there is a lack of sufficient datasets to thoroughly explore the methods' full potential and verify their effectiveness across diverse settings. To address these issues, we conduct a comprehensive benchmark providing novel text-space datasets and comprehensive evaluation under unified problem settings. Empirical results provide new insights and inspire future research directions. Our code and data are publicly available from https://github.com/CurryTang/TSGFM.

Penalty prediction is one of the main tracks of legal judgement prediction (LJP) which is to apply artificial intelligence methods to predicting the court’s judgement based on case descriptions. It is still far from effective, as the elements affecting the penalty are numerous but sparsely distributed in case descriptions, which may make them covered by noise in contexts. Allocating reasonable attention to these elements related to the penalty is the key to improving the effect of the penalty prediction. To this end, we propose a novel model called MHA-AR which learns to focus on the key elements by a modulated hierarchical attention mechanism and a legal attribute recognition subtask. Besides benefiting the prediction accuracy, it also improves the reliability and credibility of predictions for users by providing comprehensible legal attributes related to the penalty. A series of experiments conducted on the real-world datasets demonstrate the correctness of our assumptions and the superiority of MHA-AR. The implementation of our proposed model will be available at https://github.com/realcatking/penaltyprediction.