Abstract: This report synthesises findings from 1 peer-reviewed paper addressing the following research question: How does the computational overhead of decomposing independent user preferences in sequential recommendation models affect inference throughput relative to single-preference Transformer baselines. 7 claims were…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: To what extent does preference decomposition in sequential recommenders improve robustness against noisy user behavior data compared to standard GNN-based models when measured by Hit Rate and. 12 claims were…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: How does the training convergence speed and GPU memory efficiency of XSimGCL differ from SGL and SimGCL in large-scale collaborative filtering tasks. 8 claims were extracted from source literature; 8 were…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the inference throughput of XSimGCL compare to traditional graph contrastive learning baselines when scaled to billion-edge user-item graphs. 12 claims were extracted from source literature; 11 were…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: What is the impact of XSimGCL's augmentation-free strategy on recommendation accuracy metrics like NDCG and Recall compared to heavy augmentation baselines under extreme data sparsity. 10 claims were extracted…

Abstract: This report synthesises findings from 1 peer-reviewed paper addressing the following research question: How do the memory consumption patterns of LightGCL, SGL, and GCA differ during training on graphs scaling from 1 million to 20 million nodes. 13 claims were extracted from source literature; 11 were independently…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: What is the impact of neural network depth on the robustness of cross-domain recommendation systems when evaluated on NDCG@10 under varying levels of data sparsity. 9 claims were extracted from source literature;…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: What is the impact of LightGCL's spectral augmentation on training convergence speed compared to node-dropout based contrastive methods on billion-edge graphs. 5 claims were extracted from source literature; 5…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: How does the feature representation enhancement in improved PointPillars affect robustness metrics under sparse point cloud conditions relative to Dynamic Graph CNN. 5 claims were extracted from source…

Abstract: This report synthesises findings from 2 peer-reviewed papers addressing the following research question: How does LightGCL perform in cross-domain recommendation tasks relative to SGL and GCA when evaluated on Recall@20 under extreme data sparsity. 10 claims were extracted from source literature; 10 were…

Abstract: This report synthesises findings from 5 peer-reviewed papers addressing the following research question: How does the inference throughput of LightGCL compare to SGL and GCA when scaling to graphs with over 10 million nodes. 7 claims were extracted from source literature; 7 were independently verified against…

Abstract: This report synthesises findings from 8 peer-reviewed papers addressing the following research question: What is the impact of LightGCL's augmentation strategy on model robustness against adversarial edge perturbations compared to stochastic augmentation in SGL. 9 claims were extracted from source literature; 9 were…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How do inference latency and robust accuracy trade-offs differ between ViT and MLP-Mixer architectures under adaptive white-box attacks. 10 claims were extracted from source literature; 9 were independently…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does the adversarial robustness of Vision Transformers scale with model size compared to ConvNeXt under PGD attacks on ImageNet. 7 claims were extracted from source literature; 7 were independently verified…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: How does the latency overhead of ATLAS scale with increasing graph size for real-time recommendation tasks relative to distributed GNN inference frameworks. 8 claims were extracted from source literature; 8 were…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: What is the reduction in peak memory usage achieved by ATLAS during out-of-core inference on large-scale graph datasets compared to existing disk-backed GNN systems. 7 claims were extracted from source…

Abstract: This report synthesises findings from 8 peer-reviewed papers addressing the following research question: How does cross-domain adversarial transfer between image and text modalities affect the alignment stability of multimodal models under attack. 9 claims were extracted from source literature; 9 were independently…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: What is the impact of cross-dimensional transfer learning on the few-shot classification accuracy of deep learning models when evaluated on standardized medical datasets. 9 claims were extracted from source…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: What is the impact of spectral noise perturbations on the convergence rate of conditional graph diffusion models versus autoregressive graph generators. 10 claims were extracted from source literature; 10 were…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the impact of edge-level ego-network encodings on the robustness of GNNs against adversarial perturbations in node classification tasks. 9 claims were extracted from source literature; 9 were…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: How do multimodal transformers performing graph-based relational reasoning compare to dedicated GNNs in terms of alignment scores and computational efficiency on visual question answering tasks. 8 claims were…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: What is the impact of replacing self-attention with sparse message passing on LLM reasoning accuracy and inference throughput for long-context benchmarks. 10 claims were extracted from source literature; 10 were…

Abstract: This report synthesises findings from 5 peer-reviewed papers addressing the following research question: What is the computational efficiency trade-off when applying adaptive graph pooling in diffusion models like LapDDPM for code generation tasks, as measured by inference latency and throughput on the. 9 claims were…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: Does integrating sparse expert routing in alignment-focused multimodal models improve inference throughput without degrading performance on standard reasoning benchmarks. 9 claims were extracted from source…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does sparse mixture-of-experts routing in multimodal models affect alignment scores on RLHF benchmarks compared to dense baseline architectures. 10 claims were extracted from source literature; 9 were…