Abstract: This report synthesises findings from 7 peer-reviewed papers addressing the following research question: What is the impact of varying contamination rates on the AVPR and F1 score in LLM-based anomaly detection models, and how does this compare to traditional machine learning models. Achieving high accuracy in energy…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: How does the scalability of graph attention mechanisms in GNN-based anomaly detection models influence inference efficiency and F1 score stability under increasing graph sizes. Human knowledge provides a formal…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the choice of train-test split protocol in graph anomaly detection models affect the robustness of the F1 score and AUC metrics across different graph densities. Deep convolutional neural networks have…

Abstract: This report synthesises findings from 2 peer-reviewed papers addressing the following research question: Can the simplified augmentation strategy in LightGCL be adapted to multimodal recommendation systems, and how does it affect downstream task performance (e.g., accuracy, NDCG) compared to traditional. The…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does the performance of LightGCL compare to other GNN-based contrastive learning methods (e.g., SGL, GCA) on standard recommendation benchmarks (e.g., MovieLens, Amazon) when evaluated under. Graph neural…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: How effectively do prototype-based embeddings in federated GNNs transfer across domains, measured by cross-domain accuracy when applying models trained on financial graphs to healthcare networks. Deep learning…

Abstract: This report synthesises findings from 9 peer-reviewed papers addressing the following research question: What is the comparative robustness of GNN models trained with OptimES versus traditional federated learning approaches against adversarial attacks, quantified by accuracy degradation on perturbed. Abstract The…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: Does the strategic exploration in RLHF with KL constraints improve robustness in cross-domain code generation tasks, as evaluated by accuracy on multimodal benchmarks like MBPP and HumanEval across. Abstract The…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does the strategic exploration component in KL-regularized RLHF compare to offline PPO and DPO in terms of code generation accuracy on adversarial benchmarks like AdvBench, when measured using. As Large…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the robustness of XSimGCL and other simple graph contrastive learning methods to adversarial perturbations in the user-item interaction graph, as measured by the change in NDCG scores. Multilayer neural…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: How does the inference efficiency (measured in queries per second) of extremely simple graph contrastive learning models compare to complex augmentation-based models when deployed in large-scale. In the last few…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: Can graph contrastive learning methods with simplified augmentation pipelines maintain their performance on sparse interaction graphs when evaluated using Hit Ratio (HR) and Mean Average Precision. In the last…

Abstract: This report synthesises findings from 6 peer-reviewed papers addressing the following research question: How does the robustness of federated graph learning models with prototype-based embeddings compare to traditional embeddings under non-IID data distributions, measured by accuracy degradation and. Abstract The…

Abstract: This report synthesises findings from 2 peer-reviewed papers addressing the following research question: What is the effect of varying the number of federated clients on the convergence speed and inference latency of federated graph learning models using prototype-based embeddings versus traditional. Federated graph…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the integration of prototype-based embeddings impact the communication efficiency and model accuracy in federated graph learning when compared to traditional embeddings, as measured by. This paper…

Abstract: 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…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: How do different graph augmentation strategies in multi-scale contrastive learning affect the efficiency of GNN inference on large-scale datasets like OGBN-arXiv, as measured by throughput and memory. We trained…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: What are the accuracy trade-offs when applying adversarially-trained graph convolutional networks for missing data imputation in code generation dependency graphs, as measured by BLEU scores on. Traditional…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the integration of tensorial imputation techniques in multi-view graph neural networks compare to traditional missing indicator matrix approaches in terms of inference efficiency (measured. Deep learning…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: What is the impact of view dropout rates on the robustness of multi-view GNN anomaly detection models against adversarial edge perturbations. Deep convolutional neural networks have performed remarkably well on…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the sample efficiency of semi-supervised multi-view graph anomaly detection frameworks compare to single-view methods when trained with less than 5\% labeled anomalous nodes. Deep convolutional neural…

Abstract: This report synthesises findings from 6 peer-reviewed papers addressing the following research question: Does the performance gain of GRPO over PPO in LLaVA-UHD fine-tuning diminish as the model parameter count exceeds 7B when measured on SEED-Bench-R1 accuracy. Abstract The rapid evolution of large language models…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: To what extent does layer aggregation depth reduce inference accuracy in graph neural networks evaluated on large-scale graph benchmark suites. Molecular machine learning has been maturing rapidly over the last…

Abstract: This report synthesises findings from 9 peer-reviewed papers addressing the following research question: What is the trade-off between sampling rate and accuracy in deep graph neural networks when evaluated on node classification tasks across heterogeneous graph datasets. In the last few years, the deep learning (DL)…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: What is the correlation between the number of message-passing layers and performance degradation in semi-supervised graph representation learning on Cora and Citeseer. Graph Neural Networks (GNNs) have achieved…