REVIEW 18 cited by
A critical look at the evaluation of GNNs under heterophily: Are we really making progress?
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
A critical look at the evaluation of GNNs under heterophily: Are we really making progress?
read the original abstract
Node classification is a classical graph machine learning task on which Graph Neural Networks (GNNs) have recently achieved strong results. However, it is often believed that standard GNNs only work well for homophilous graphs, i.e., graphs where edges tend to connect nodes of the same class. Graphs without this property are called heterophilous, and it is typically assumed that specialized methods are required to achieve strong performance on such graphs. In this work, we challenge this assumption. First, we show that the standard datasets used for evaluating heterophily-specific models have serious drawbacks, making results obtained by using them unreliable. The most significant of these drawbacks is the presence of a large number of duplicate nodes in the datasets Squirrel and Chameleon, which leads to train-test data leakage. We show that removing duplicate nodes strongly affects GNN performance on these datasets. Then, we propose a set of heterophilous graphs of varying properties that we believe can serve as a better benchmark for evaluating the performance of GNNs under heterophily. We show that standard GNNs achieve strong results on these heterophilous graphs, almost always outperforming specialized models. Our datasets and the code for reproducing our experiments are available at https://github.com/yandex-research/heterophilous-graphs
Forward citations
Cited by 18 Pith papers
-
GraphIP-Bench: How Hard Is It to Steal a Graph Neural Network, and Can We Stop It?
GraphIP-Bench shows stealing GNNs is easy at moderate query budgets, most defenses fail to block or reliably trace extraction, and watermarks lose verification power on surrogates while heterophilic graphs are harder ...
-
Gate the Filter, Not the Message: Node-Channel Mixtures for Pre-Propagation GNNs
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-sc...
-
Gaussian Sheaf Neural Networks
Gaussian Sheaf Neural Networks derive a sheaf Laplacian for Gaussian node features on graphs to preserve their geometric structure during message passing.
-
Graph Navier Stokes Networks
GNSN defines a dynamic velocity field on graphs to enable convection alongside diffusion, mitigating oversmoothing and outperforming baselines on twelve datasets with varying homophily.
-
NeighborDiv: Training-free Zero-shot Generalist Graph Anomaly Detection via Neighbor Diversity
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.
-
GraphIP-Bench: How Hard Is It to Steal a Graph Neural Network, and Can We Stop It?
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.
-
L2G-Net: Local to Global Spectral Graph Neural Networks via Cauchy Factorizations
The graph Fourier transform is exactly factored into a chain of local subgraph transforms stitched by Cauchy matrices, giving L2G-Net a spectral GNN with O(kn^2) setup cost and competitive long-range benchmarks.
-
HeAD-CP: Heterophily-Aware Diffused Conformal Prediction Sets for Graph Neural Networks
Graph conformal prediction with node-wise softmax-derived diffusion coefficients avoids DAPS's heterophily failure and beats it on 8/10 benchmarks in oracle selection.
-
Swarm-Inspired Generation of Collective Behaviors in Graph Dynamical Systems
SIES learns generalizable local coupling operators via signed source-target attention for controllable synchronization in graph dynamical systems and applies the principle to heterophilous graph representation learning.
-
Generalist Graph Anomaly Detection via Prototype-Based Distillation
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.
-
Rethinking Feature Alignment in Generalist Graph Anomaly Detection: A Relational Fingerprint-based Approach
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.
-
Graph Navier Stokes Networks
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 1...
-
Random-Set Graph Neural Networks
RS-GNNs predict random sets over classes using belief functions to jointly produce class probabilities and epistemic uncertainty estimates for graph nodes.
-
Hierarchical Multi-Scale Graph Neural Networks: Scalable Heterophilous Learning with Oversmoothing and Oversquashing Mitigation
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.
-
Softsign: Smooth Sign in Your Optimizer For Better Parameter Heterogeneity Handling
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 perf...
-
Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification
Transductive Sharpening adds an entropy-minimization term on unlabeled-node predictions to the training objective for graph node classification.
-
Robust Learning on Heterogeneous Graphs with Heterophily: A Graph Structure Learning Approach
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.
-
Beyond Isolated Clients: Integrating Graph-Based Embeddings into Event Sequence Models
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.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.