A curated list of papers, code, and resources on graph-agnostic clustering — unsupervised clustering methods that work across graphs with arbitrary topology (homophilic or heterophilic), without assumptions baked into a single graph instance.
Classical deep graph clustering methods implicitly assume homophily — the property that connected nodes share labels. When this assumption breaks down (heterophilic graphs, mixed neighbourhoods, noisy connectivity), cluster-propagation pipelines degrade silently.
Graph-agnostic clustering seeks methods that remain robust across the full spectrum of graph structures: homophilic, heterophilic, and everything in between. This list curates work at and around that goal — including foundational GNN primitives, heterophilic GNNs, deep graph clustering methods, and related topics in propagation, diffusion, and uncertainty.
- Graph-Agnostic Clustering
- Heterophilic Graph Clustering
- Deep Graph Clustering
- Heterophilic GNNs (Supervised)
- Graph Structure Learning & Rewiring
- Graph Diffusion & ODEs
- Label Propagation & Graph Self-Training
- Calibration & Uncertainty on Graphs
- Multi-View Clustering
- Foundational GNN Primitives
- Surveys & Benchmarks
- Tools & Libraries
- Related Awesome Lists
- Contributing
Methods that explicitly aim to work across homophilic and heterophilic graphs.
- Beyond Homophily: Reconstructing Structure for Graph-agnostic Clustering · Pan, Kang, ICML 2023 · arxiv
- Provable Filter for Real-world Graph Clustering · Li et al., 2024 · arxiv
- Diffusion-based Graph-agnostic Clustering · Xie, Yang, Wang, WWW 2025 · doi
- Trustworthy Neighborhoods Mining: Homophily-Aware Neutral Contrastive Learning for Graph Clustering · Peng et al., 2025 · arxiv
Clustering methods designed for, or evaluated primarily on, heterophilic graphs.
- Homophily-enhanced Structure Learning for Graph Clustering (HoLe) · Zhao et al., 2023 · arxiv
- SiMilarity-Enhanced Homophily for Multi-View Heterophilous Graph Clustering (SMHGC) · Chen et al., 2024 · arxiv
- Robust Graph Structure Learning under Heterophily · Xie, Chen, Kang, Neural Networks 2025 · arxiv
- HeNCler: Node Clustering in Heterophilous Graphs via Learned Asymmetric Similarity · Achten et al., ICANN 2025 · arxiv
- Disentangling Homophily and Heterophily in Multimodal Graph Clustering (DMGC) · Guo et al., ACM MM 2025 · arxiv
General-purpose deep graph clustering methods that downstream graph-agnostic work builds on.
- Graph Clustering with Graph Neural Networks (DMoN) · Tsitsulin, Palowitch, Perozzi, Müller, JMLR 2023 · arxiv
- DGCluster: A Neural Framework for Attributed Graph Clustering via Modularity Maximization · Bhowmick, Kosan, Huang, Singh, Medya, AAAI 2024 · arxiv
Supervised GNN methods addressing heterophily — a key backbone area for graph-agnostic work.
- Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs (H2GCN) · Zhu et al., NeurIPS 2020 · arxiv
- Adaptive Universal Generalized PageRank Graph Neural Network (GPR-GNN) · Chien et al., ICLR 2021 · arxiv
- Revisiting Heterophily for Graph Neural Networks · Luan et al., NeurIPS 2022 · arxiv
- Understanding Heterophily for Graph Neural Networks · Wang, Guo, Yang, Wang, ICML 2024 · arxiv
- Resurrecting Label Propagation for Graphs with Heterophily and Label Noise (R²LP) · Cheng et al., KDD 2024 · arxiv
- Multi-Track Message Passing: Tackling Oversmoothing and Oversquashing via Preventing Heterophily Mixing · Pei et al., ICML 2024 · paper
- Mitigating Oversmoothing Through Reverse Process of GNNs for Heterophilic Graphs · Park, Heo, Kim, ICML 2024 · arxiv
Methods that modify graph topology to improve downstream clustering / learning.
- Understanding Over-Squashing and Bottlenecks on Graphs via Curvature · Topping et al., ICLR 2022 · arxiv
- DiffWire: Inductive Graph Rewiring via the Lovász Bound · Arnaiz-Rodríguez et al., LoG 2022 · arxiv
- Uncertainty-Aware Graph Structure Learning (UnGSL) · Zhao et al., WWW 2025 · arxiv
Continuous-time, diffusion, and ODE-based formulations of graph learning.
- A Generalized Neural Diffusion Framework on Graphs (HiD-Net) · Li, Wang, Liu, Shi, AAAI 2024 · arxiv
- Convection-Diffusion Equation: A Theoretically Certified Framework for Neural Networks (CDE) · Wang, Bao, Shi, IEEE TPAMI 2024 · arxiv
- Graph ODEs and Beyond: A Comprehensive Survey · Liu et al., 2025 · arxiv
Propagation-based and pseudo-label approaches closely related to cluster propagation.
- Learning with Local and Global Consistency · Zhou, Bousquet, Lal, Weston, Schölkopf, NeurIPS 2003 · paper
classic - Predict then Propagate: Graph Neural Networks meet Personalized PageRank (APPNP) · Klicpera, Bojchevski, Günnemann, ICLR 2018 · arxiv
- BANGS: Game-Theoretic Node Selection for Graph Self-Training · Wang, Liu, Medya, Yu, ICLR 2025 · arxiv
Calibration, uncertainty estimation, and conformal prediction on graph models.
- Confidence-based Graph Convolutional Networks for Semi-Supervised Learning · Vashishth, Yadav, Bhandari, Talukdar, AISTATS 2019 · arxiv
- Towards Calibrated Deep Clustering Network · Jia, Cheng, Liu, Hou, ICLR 2025 · arxiv
- Evidential Uncertainty Probes for Graph Neural Networks · Yu et al., AISTATS 2025 · arxiv
- Residual Reweighted Conformal Prediction for Graph Neural Networks · Zhang et al., UAI 2025 · arxiv
- Hierarchical Uncertainty-Aware Graph Neural Network · Hu et al., CIKM 2025 · arxiv
Related work on multi-view clustering methods relevant to multi-modal graph settings.
- COPER: Correlation-Based Permutations for Multi-View Clustering · Eisenberg, Svirsky, Lindenbaum, ICLR 2025 · arxiv
- ROLL: Robust Noisy Pseudo-label Learning for Multi-View Clustering with Noisy Correspondence · Sun et al., CVPR 2025 · paper
Core architectures that downstream graph-agnostic clustering work builds on.
- Variational Graph Auto-Encoders (VGAE) · Kipf, Welling, 2016 · arxiv
- Graph Attention Networks (GAT) · Veličković et al., ICLR 2018 · arxiv
- Deep Graph Infomax (DGI) · Veličković et al., ICLR 2019 · arxiv
- Simplifying Graph Convolutional Networks (SGC) · Wu et al., ICML 2019 · arxiv
- A Critical Look at the Evaluation of GNNs under Heterophily: Are We Really Making Progress? · Platonov, Kuznedelev, Diskin, Babenko, Prokhorenkova, ICLR 2023 · arxiv
benchmark critique - The Heterophilic Graph Learning Handbook · Luan et al., 2024 · arxiv
survey - Bridging Academia and Industry: A Comprehensive Benchmark for Attributed Graph Clustering (PyAGC) · Liu et al., 2026 · arxiv
benchmark - Beyond the Academic Monoculture: A Unified Framework and Industrial Perspective for Attributed Graph Clustering · Liu et al., 2026 · arxiv
survey
- PyTorch Geometric (PyG) — De-facto PyTorch GNN library.
- Deep Graph Library (DGL) — Multi-framework GNN library.
- NetworkX — Graph creation, manipulation, and analysis in pure Python.
- Spektral — GNN library in TensorFlow / Keras.
- Open Graph Benchmark (OGB) — Diverse graph datasets across scales and domains.
- TUDataset — Graph classification benchmark collection.
- GNNPapers — Must-read GNN papers, maintained by THUNLP.
- graph-based-deep-learning-literature — Extensive, venue-organised GNN bibliography.
- awesome-graph-classification
- awesome-self-supervised-gnn
Contributions are warmly welcomed. The list aims for depth over breadth — include a paper only if it meaningfully advances understanding of graph-agnostic clustering, or directly supports it as a prerequisite.
Entry format:
- **Paper Title** · *First Author et al., Venue Year* · [arxiv](https://arxiv.org/abs/XXXX.XXXXX)See CONTRIBUTING.md for full guidelines.
To the extent possible under law, the maintainers have waived all copyright and related rights to this list. See LICENSE.
Maintained by Haili Yuan · PhD in Computer Science @ University of Warwick