{"total":13,"items":[{"citing_arxiv_id":"2606.01660","ref_index":14,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Gate the Filter, Not the Message: Node-Channel Mixtures for Pre-Propagation GNNs","primary_cat":"cs.LG","submitted_at":"2026-06-01T04:14:13+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"FilterMoE uses joint node-channel routing of Chebyshev filter experts through a 3D gating tensor in pre-propagation GNNs and outperforms baselines on nine of eleven benchmarks while ranking first on all three large-scale ones with a 1.53-point average gain.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.31371","ref_index":38,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Softsign: Smooth Sign in Your Optimizer For Better Parameter Heterogeneity Handling","primary_cat":"cs.LG","submitted_at":"2026-05-29T14:41:36+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"SoftSignum replaces hard sign with soft-sign in optimizers via temperature control and quantile scheduling, extends to SoftMuon, provides a convergence proof for stochastic non-convex settings, and reports better performance than sign-based methods and AdamW on deep learning tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.26857","ref_index":1,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Generalist Graph Anomaly Detection via Prototype-Based Distillation","primary_cat":"cs.LG","submitted_at":"2026-05-26T11:16:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ProMoS introduces the first unsupervised generalist graph anomaly detection method via prototype-based distillation from a self-supervised GNN teacher to a mixture-of-students model for zero-shot cross-graph transfer.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.25429","ref_index":8,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Rethinking Feature Alignment in Generalist Graph Anomaly Detection: A Relational Fingerprint-based Approach","primary_cat":"cs.LG","submitted_at":"2026-05-25T05:12:26+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ReFi-GAD uses a semantics-aware relational fingerprint and transformer-based model with SNR refinement to align heterogeneous features for generalist graph anomaly detection across unseen graphs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.21435","ref_index":73,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Gaussian Sheaf Neural Networks","primary_cat":"cs.LG","submitted_at":"2026-05-20T17:26:32+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Gaussian Sheaf Neural Networks derive a sheaf Laplacian for Gaussian node features on graphs to preserve their geometric structure during message passing.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.21247","ref_index":65,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Graph Navier Stokes Networks","primary_cat":"cs.LG","submitted_at":"2026-05-20T14:36:52+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"GNSN adds convection governed by a dynamic velocity field to graph message passing, adaptively balancing it with diffusion to handle varying homophily levels and reduce oversmoothing while outperforming baselines on 12 datasets.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.20879","ref_index":24,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"NeighborDiv: Training-free Zero-shot Generalist Graph Anomaly Detection via Neighbor Diversity","primary_cat":"cs.LG","submitted_at":"2026-05-20T08:16:13+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"NeighborDiv detects graph anomalies via variance of inter-neighbor feature similarities under a new Neighbor-to-Neighbor Diversity Paradigm, achieving SOTA results with zero volatility in zero-shot cross-domain settings.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.20248","ref_index":34,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification","primary_cat":"cs.LG","submitted_at":"2026-05-18T06:47:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Transductive Sharpening adds an entropy-minimization term on unlabeled-node predictions to the training objective for graph node classification.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.12827","ref_index":18,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"GraphIP-Bench: How Hard Is It to Steal a Graph Neural Network, and Can We Stop It?","primary_cat":"cs.CR","submitted_at":"2026-05-12T23:49:45+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"GraphIP-Bench is a new unified benchmark showing GNN model extraction succeeds at moderate query budgets while most defenses fail to prevent it or retain verification signals on surrogates.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.11987","ref_index":25,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Random-Set Graph Neural Networks","primary_cat":"cs.AI","submitted_at":"2026-05-12T11:38:13+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"RS-GNNs predict random sets over classes using belief functions to jointly produce class probabilities and epistemic uncertainty estimates for graph nodes.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"InRS-GNN, the head outputs focal-set beliefs via a two-layer MLP with sigmoid outputs over a focal-set budgetF: ˆBelv(Ak) = sigmoid(g(zv))k, A k ∈ F.(23) 10 Masses over focal sets are recovered by a Möbius inversion matrixM(precomputed fromF): ˆmv = ˆBelvM,(24) and pignistic probabilities are computed by multiplying masses with the pignistic matrix P induced byF: BetPv = norm( ˆmvP),(25) where norm(·) denotes row-wise normalisation to ensure a proper probability vector. Final predictions use arg maxi BetPv(i), so RS-GNN differs from vanilla primarily in its uncertainty representation rather than in the decision rule. Focal-set selection (how F is constructed).To keep random-set outputs scalable, we do not enumerate all of 2Y. Instead, F contains: (i) all singleton sets over the full class space, and (ii) up to"},{"citing_arxiv_id":"2605.10975","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Hierarchical Multi-Scale Graph Neural Networks: Scalable Heterophilous Learning with Oversmoothing and Oversquashing Mitigation","primary_cat":"cs.LG","submitted_at":"2026-05-08T22:35:17+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"HMH builds soft hierarchies with orthonormal Haar bases and heterophily-aware encoders to apply learnable spectral filters while using skip unpooling to avoid oversmoothing and hub bias on heterophilous graphs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.27387","ref_index":3,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Robust Learning on Heterogeneous Graphs with Heterophily: A Graph Structure Learning Approach","primary_cat":"cs.AI","submitted_at":"2026-04-30T03:55:54+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"HGUL jointly recovers reliable neighborhoods via kNN, adaptively filters noisy edges, and models class relationships with a polynomial kernel affinity matrix to handle heterophily and structural noise in heterogeneous graphs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.09085","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Beyond Isolated Clients: Integrating Graph-Based Embeddings into Event Sequence Models","primary_cat":"cs.LG","submitted_at":"2026-04-10T08:11:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Three strategies for adding graph embeddings to event sequence SSL models improve AUC by up to 2.3% on four financial and e-commerce datasets, with graph density determining the best integration approach.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}