Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: To what extent does the attention mechanism in GATs contribute to vulnerability against node injection attacks compared to the latent structure learning in Graph Inference Learning for dynamic graph. Graph neural…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: What is the robustness of Mul-GAD against adversarial attacks on graph structure or node features compared to other GNN-based anomaly detection methods, as measured by the drop in F1 score under. Anomaly…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: What is the difference in inference efficiency and robustness degradation between Graph Attention Networks and Graph Inference Learning models under iterative gradient-based attacks on traffic. Real-time traffic…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: How does the accuracy degradation of large language models under few-shot prompting compare to full fine-tuning when evaluated on reasoning benchmarks with varying levels of label noise. Reinforcement learning…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: How do Graph Attention Networks compare to Graph Inference Learning in maintaining accuracy on spatio-temporal forecasting tasks when subjected to targeted structural adversarial perturbations. Real-time traffic…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the performance of Mul-GAD compare to state-of-the-art semi-supervised GNN frameworks like GAT or GIN on heterogeneous graphs with varying sparsity levels in terms of precision-recall. In order to…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: What is the accuracy trade-off between dynamic GNNs and static GNNs when evaluated on node classification tasks using the T-GNN benchmark with varying graph sizes (10K–100K nodes), measured by. Graph neural…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the inference throughput of semi-supervised GNN anomaly detectors like Mul-GAD compare to unsupervised methods (e.g., DOMINANT) when evaluated on temporal graph benchmarks with node sizes. Graph neural…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: How do semi-supervised dynamic GNNs perform compared to fully unsupervised static GNNs in terms of memory efficiency (peak GPU RAM usage) when processing graphs with 100K nodes, using the DGN-Bench. Graph Neural…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: Does joint structure-label estimation in Graph Neural Networks demonstrate superior robustness to batch size variations compared to semi-supervised learning in terms of convergence speed and final. Heterogeneous…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: In what ways do advanced graph data augmentation techniques influence the inference efficiency and robustness of multimodal models trained on sparse, large-scale knowledge graphs compared to. The rise of…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: What is the impact of incorporating self-supervised contrastive learning (e.g., SimCLR) on the recommendation robustness of XSimGCL when evaluated on out-of-domain datasets like Goodreads or Steam. Contrastive…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: Does applying brute-force text augmentation during pre-training improve the code generation pass@1 scores of LLMs on the HumanEval dataset. Remote sensing vision tasks require extensive labeled data across…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the trade-off between memory efficiency and prediction performance for different batch sizes in Graph Neural Networks trained with joint structure-label objectives on the large-scale OGB-mag.…

Abstract: This report synthesises findings from 16 peer-reviewed papers addressing the following research question: Can CoATA's co-augmentation approach maintain consistent performance improvements across diverse graph domains while preserving computational efficiency. Graph Anomaly Detection (GAD) has demonstrated great…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: How does the inference efficiency of latent factor models scale with dataset size in music recommendation compared to collaborative filtering methods when evaluated on metrics like throughput and. Music…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: How does the integration of synthetic node feature generation and edge perturbation in graph augmentation frameworks compare to single-dimensional methods regarding convergence speed and final. The standard…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: To what extent does co-augmentation training improve the alignment stability of code generation models against syntax-preserving adversarial attacks as measured by pass@k scores on the HumanEval. This paper…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: How do multimodal extensions of graph contrastive learning models (e.g., combining visual and textual features) perform on cross-domain recommendation tasks compared to pure graph-based approaches. Multimedia…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the impact of contrastive learning objectives on the robustness of hybrid graph neural networks against adversarial attacks in few-shot node classification tasks, evaluated on accuracy and. We present…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does the computational efficiency of hybrid graph neural networks combining synthetic graph augmentation and contrastive learning scale with graph size compared to variational inference. We present a novel…

Abstract: This report synthesises findings from 16 peer-reviewed papers addressing the following research question: How does the integration of dynamic graph convolutional networks with transformer-based architectures improve node classification accuracy on heterogeneous graphs compared to pure GCN baselines, as. Graph…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: What is the impact of few-shot learning strategies on the reasoning capabilities of large language models when evaluated on the GSM8K dataset with varying numbers of demonstration examples. Reinforcement learning…

Abstract: This report synthesises findings from 2 peer-reviewed papers addressing the following research question: Can multimodal contrastive learning enhance the robustness of few-shot node classification models against label noise, and how does its performance compare to unimodal approaches on standard. The main task of…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: Does simplified noise injection maintain robustness in cross-domain graph contrastive learning when evaluated on benchmark datasets such as Cora and Citeseer using the normalized discounted. Acquiring reviewers…