Towards One-for-All Foundation Model for Attributed Graph Clustering

Yunhui Liu; Xudong Jin; Kang Zhang; Danshuo An; Yu Xing; Te Song; Jia Liu; Tieke He

cs.LGOct 6, 2026

Towards One-for-All Foundation Model for Attributed Graph Clustering

Authors: Yunhui Liu, Xudong Jin, Kang Zhang, Danshuo An, Yu Xing, Te Song, Jia Liu, Tieke He

Organizations: State Key Laboratory for Novel Software Technology Nanjing University Nanjing, China

Abstract

Attributed graph clustering aims to discover node groups by jointly exploiting node attributes and graph topology, yet its unsupervised nature makes model selection and adaptation inherently difficult. Existing methods typically train and tune a separate model for each input graph, leading to costly and fragile pipelines that often fail to transfer across graphs with different feature spaces, structural patterns, and attribute-structure correlations. In this paper, we study a one-for-all alternative: can a single model be trained once and directly applied to diverse attributed graphs without graph-specific training, fine-tuning, or hyperparameter search? We propose OFAG, a foundation model for attributed graph clustering. Building upon Prior-data Fitted Networks, OFAG learns a reusable clustering inference strategy from synthetic attributed graphs generated under broad priors over latent clusters, node attributes, and graph structures. To handle incompatible feature spaces across graphs, OFAG adopts a dimension-agnostic signal-wise graph encoder that treats each feature channel as a graph signal and models its response to shared graph filters. The model is trained with a hyperspherical clustering objective, producing clustering-friendly node representations in a single forward pass at inference time. On ten datasets, one frozen OFAG model achieves the best mean performance and average rank across NMI, ACC, ARI, and F1, while completing all ten datasets in 12.43 minutes total---over 6* faster than the second-fastest baseline and nearly 28* faster than the second-best on clustering quality. Our code and pretrained checkpoint are available at https://github.com/Cloudy1225/OFAG, allowing practitioners to directly apply OFAG to their own attributed graph datasets without additional training or tuning.

Figures & tables

Figure 1. Overview of OFAG . The model is pre-trained on diverse synthetic attributed graphs using the SwiFRT encoder and a hyperspherical clustering objective. After pre-training, the frozen model can be directly applied to unseen real-world graphs for zero-shot clustering without graph-specific fine-tuning.

Model

Metric

Homophilous Graphs

Heterophilous Graphs

Mean

Avg. Rank

Cora

Photo

ArXiv

Products

Flickr

MAG

Pokec

WebTopic

KMeans

NMI

13.89

{\scriptscriptstyle\pm 4.46}

31.70

{\scriptscriptstyle\pm 0.49}

22.59

{\scriptscriptstyle\pm 0.03}

11.08

{\scriptscriptstyle\pm 0.14}

29.24

{\scriptscriptstyle\pm 0.11}

10.17

{\scriptscriptstyle\pm 0.11}

1.21

{\scriptscriptstyle\pm 0.07}

28.32

{\scriptscriptstyle\pm 0.06}

1.41

{\scriptscriptstyle\pm 0.00}

2.45

{\scriptscriptstyle\pm 0.05}

15.21

16.30

ACC

33.09

{\scriptscriptstyle\pm 4.00}

42.54

{\scriptscriptstyle\pm 2.54}

16.95

{\scriptscriptstyle\pm 0.38}

8.59

{\scriptscriptstyle\pm 0.06}

21.23

{\scriptscriptstyle\pm 0.74}

14.12

{\scriptscriptstyle\pm 0.25}

26.02

{\scriptscriptstyle\pm 1.73}

7.09

{\scriptscriptstyle\pm 0.03}

1.91

{\scriptscriptstyle\pm 0.01}

9.63

{\scriptscriptstyle\pm 0.34}

18.12

15.90

Node2Vec

NMI

44.95

{\scriptscriptstyle\pm 1.49}

66.02

{\scriptscriptstyle\pm 1.05}

38.93

{\scriptscriptstyle\pm 0.15}

79.23

{\scriptscriptstyle\pm 0.38}

50.97

{\scriptscriptstyle\pm 0.20}

7.60

{\scriptscriptstyle\pm 0.13}

5.59

{\scriptscriptstyle\pm 0.00}

37.73

{\scriptscriptstyle\pm 0.03}

31.92

{\scriptscriptstyle\pm 0.02}

5.80

{\scriptscriptstyle\pm 0.09}

36.87

9.80

ACC

60.52

{\scriptscriptstyle\pm 2.81}

73.08

{\scriptscriptstyle\pm 1.06}

25.97

{\scriptscriptstyle\pm 0.57}

70.67

{\scriptscriptstyle\pm 1.68}

36.59

{\scriptscriptstyle\pm 0.74}

13.28

{\scriptscriptstyle\pm 0.41}

24.44

{\scriptscriptstyle\pm 0.01}

10.86

{\scriptscriptstyle\pm 0.09}

16.62

{\scriptscriptstyle\pm 0.19}

10.71

{\scriptscriptstyle\pm 0.12}

34.27

10.60

SSGC

NMI

51.85

{\scriptscriptstyle\pm 1.05}

70.75

{\scriptscriptstyle\pm 1.39}

46.12

{\scriptscriptstyle\pm 0.16}

51.61

{\scriptscriptstyle\pm 0.36}

52.04

{\scriptscriptstyle\pm 0.17}

12.15

{\scriptscriptstyle\pm 0.11}

4.48

{\scriptscriptstyle\pm 0.19}

41.52

{\scriptscriptstyle\pm 0.04}

4.33

{\scriptscriptstyle\pm 0.01}

3.92

{\scriptscriptstyle\pm 0.11}

33.88

7.60

Table 1. Clustering performance in NMI and ACC on homophilous and heterophilous attributed graphs. The best , second-best , and third-best results in each column are highlighted. For Avg. Rank, lower is better; for all other columns, higher is better.

Method	Cora (s)	Photo (s)	HM (m)	Flickr (m)	ArXiv (m)	Reddit (m)	MAG (m)	Pokec (m)	Products (m)	WebTopic (m)	Total (m)	Mean NMI
Node2Vec	24.57	27.96	1.50	20.97	5.05	22.67	77.58	236.53	64.94	75.92	506.04	36.87
DAEGC	35.80	156.84	1.10	4.87	9.06	14.28	27.52	14.28	37.28	11.11	122.71	25.36
DinkNet	13.31	36.07	0.14	1.40	3.72	4.49	5.46	6.18	85.27	4.27	111.75	31.54
MinCut	5.51	6.78	5.11	0.34	1.00	6.73	16.01	24.15	32.55	8.37	94.46	29.19
DMoN	3.58	6.99	5.08	0.34	1.37	1.22	15.46	22.91	22.88	6.75	76.19	30.05
Neuromap	4.73	12.89	5.17	0.47	1.62	6.73	14.56	23.07	23.00	4.97	79.88	30.22

Table 2. Wall-clock clustering time on benchmark datasets. Following PyAGC, all methods are evaluated on a single NVIDIA Tesla V100 GPU with 32GB memory. Cora and Photo are reported in seconds, while the remaining datasets and total time are reported in minutes. Lower is better. The

\dagger

mark indicates that OFAG is run on CPU due to GPU out-of-memory.

Variant	Homo Avg		Hetero Avg		All Avg
Variant	NMI	ACC	NMI	ACC	NMI	ACC
OFAG (full)	62.46	63.01	16.51	16.18	39.49	39.59
Continuous features only	62.02	62.35	16.12	15.83	39.07	39.09
Homophilous graphs only	61.46	59.24	15.65	15.22	38.56	37.23
$g_{\theta}$ : Transformer $\to$ MLP	60.24	59.41	13.80	14.32	37.02	36.86
w/o $\ell_{2}$ normalization	54.72	48.70	14.10	15.27	34.41	31.98

Table 3. Component ablation. Average NMI/ACC over the five homophilous (Homo), five heterophilous (Hetero), and all ten graphs (All).

Figure 2. Validation loss and NMI on the fixed synthetic validation set evaluated every 50 steps during pre-training.

Figure 3. 3D unit-sphere visualization of node representations on the Photo dataset, obtained by PCA projection and re-normalization. Left: raw node features are scattered uniformly with no discernible cluster structure. Right: OFAG embeddings form compact, well-separated clusters.

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Panel of Figure 4 — Figure 4. Representative synthetic graphs sampled from the CAGM prior, with varying numbers of nodes ( $N$ ), edges ( $E$ ), clusters ( $K$ ), and edge homophily ( $h$ ). Node colors indicate ground-truth cluster assignments. The four examples span the full structural spectrum: from strongly heterophilous ( $h=0.194$ , left) through near-neutral ( $h=0.491$ ) and mildly homophilous ( $h=0.589$ ) to strongly homophilous ( $h=0.863$ , right), illustrating the broad coverage of homophily regimes.

Model

Metric

Homophilous Graphs

Heterophilous Graphs

Mean

Avg. Rank

Cora

Photo

ArXiv

Products

Flickr

MAG

Pokec

WebTopic

KMeans

ARI

6.95

{\scriptscriptstyle\pm 2.61}

20.01

{\scriptscriptstyle\pm 1.01}

7.23

{\scriptscriptstyle\pm 0.15}

2.93

{\scriptscriptstyle\pm 0.05}

9.22

{\scriptscriptstyle\pm 0.33}

2.96

{\scriptscriptstyle\pm 0.09}

1.06

{\scriptscriptstyle\pm 0.15}

2.77

{\scriptscriptstyle\pm 0.03}

0.23

{\scriptscriptstyle\pm 0.00}

0.87

{\scriptscriptstyle\pm 0.04}

5.42

16.30

25.99

{\scriptscriptstyle\pm 2.91}

42.68

{\scriptscriptstyle\pm 3.32}

12.97

{\scriptscriptstyle\pm 0.18}

7.25

{\scriptscriptstyle\pm 0.08}

12.77

{\scriptscriptstyle\pm 0.41}

12.50

{\scriptscriptstyle\pm 0.20}

15.16

{\scriptscriptstyle\pm 0.85}

5.71

{\scriptscriptstyle\pm 0.07}

0.98

{\scriptscriptstyle\pm 0.01}

4.64

{\scriptscriptstyle\pm 0.04}

14.06

16.45

Node2Vec

ARI

33.61

{\scriptscriptstyle\pm 2.49}

57.11

{\scriptscriptstyle\pm 1.39}

16.69

{\scriptscriptstyle\pm 0.73}

64.34

{\scriptscriptstyle\pm 1.64}

18.53

{\scriptscriptstyle\pm 1.05}

1.84

{\scriptscriptstyle\pm 0.07}

1.52

{\scriptscriptstyle\pm 0.00}

4.75

{\scriptscriptstyle\pm 0.07}

9.27

{\scriptscriptstyle\pm 0.08}

1.53

{\scriptscriptstyle\pm 0.03}

20.92

10.90

61.66

{\scriptscriptstyle\pm 2.52}

69.33

{\scriptscriptstyle\pm 1.18}

20.96

{\scriptscriptstyle\pm 0.29}

68.61

{\scriptscriptstyle\pm 2.12}

25.41

{\scriptscriptstyle\pm 0.24}

11.65

{\scriptscriptstyle\pm 0.46}

19.68

{\scriptscriptstyle\pm 0.01}

8.90

{\scriptscriptstyle\pm 0.09}

12.66

{\scriptscriptstyle\pm 0.15}

7.40

{\scriptscriptstyle\pm 0.16}

30.63

8.70

SSGC

ARI

41.25

{\scriptscriptstyle\pm 1.69}

59.17

{\scriptscriptstyle\pm 3.81}

33.04

{\scriptscriptstyle\pm 0.29}

33.61

{\scriptscriptstyle\pm 1.29}

18.84

{\scriptscriptstyle\pm 0.40}

3.18

{\scriptscriptstyle\pm 0.02}

3.29

{\scriptscriptstyle\pm 0.11}

5.84

{\scriptscriptstyle\pm 0.09}

0.42

{\scriptscriptstyle\pm 0.01}

1.75

{\scriptscriptstyle\pm 0.04}

20.04

8.45

Appendix

Table 4. Clustering performance in ARI and F1 on homophilous and heterophilous attributed graphs. The best , second-best , and third-best results in each column are highlighted. For Avg. Rank, lower is better; for all other columns, higher is better.

Scale	Dataset	Domain	#Nodes	#Edges	Avg. Deg.	#Feat.	Feat. Type	#Clus.	$\mathcal{H}_{e}$	$\mathcal{H}_{n}$
Tiny	Cora	Citation	2,708	10,556	3.9	1,433	Textual	7	0.81	0.83
Tiny	Photo	Co-purchase	7,650	238,162	31.1	745	Textual	8	0.83	0.84
Small	HM	Co-purchase	46,563	21,461,990	460.9	120	Tabular	21	0.16	0.35
Small	Flickr	Social	89,250	899,756	10.1	500	Textual	7	0.32	0.32
Medium	ArXiv	Citation	169,343	1,166,243	6.9	128	Textual	40	0.65	0.64
	Reddit	Social	232,965	23,213,838	99.6	602	Textual	41	0.78	0.81

Appendix

Table 5. Summary of the benchmark datasets. The collection spans multiple graph scales, domains, and feature modalities.

\mathcal{H}_{e}

and

\mathcal{H}_{n}

denote edge homophily and node homophily, respectively.

Panel of Figure 5 — Figure 5. NMI of OFAG at different pre-training checkpoints on homophilous, heterophilous, and all benchmark graphs. Performance improves substantially during the early stage of pre-training and subsequently stabilizes, indicating that OFAG rapidly learns a transferable clustering inference strategy. The inset plots enlarge the early checkpoints to highlight the improvement from initialization to the first evaluated pre-training steps.

$L$	Homo Avg		Hetero Avg		All Avg
$L$	NMI	ACC	NMI	ACC	NMI	ACC
2	59.54	56.47	14.90	14.49	37.22	35.48
3	61.97	60.32	15.53	15.13	38.75	37.72
5	62.33	61.79	15.80	14.99	39.06	38.39
8	61.77	60.78	16.05	15.20	38.91	37.99
10	62.46	63.01	16.51	16.18	39.49	39.59

Appendix

Table 7. Sensitivity to filter order

L

in SwiFRT.

$\lambda$	Homo Avg		Hetero Avg		All Avg
$\lambda$	NMI	ACC	NMI	ACC	NMI	ACC
0	61.40	59.54	14.50	14.54	37.95	37.04
0.1	62.46	63.01	16.51	16.18	39.49	39.59
0.3	61.21	59.07	16.40	15.40	38.81	37.23
0.5	62.43	61.96	16.33	15.38	39.38	38.67
0.8	61.59	60.03	16.18	15.39	38.89	37.71

Appendix

Table 8. Sensitivity to dispersion weight

\lambda

Homophily range	Homo Avg		Hetero Avg		All Avg
Homophily range	NMI	ACC	NMI	ACC	NMI	ACC
$[0.1,\,0.9]$ (full, default)	62.46	63.01	16.51	16.18	39.49	39.59
$[0.7,\,0.9]$ (homo-only)	61.46	59.24	15.65	15.22	38.56	37.23
$[0.4,\,0.6]$ (near-neutral)	57.52	54.37	15.15	14.88	36.34	34.63
$[0.1,\,0.3]$ (hetero-only)	51.62	47.34	13.13	13.26	32.37	30.30

Appendix

Table 9. Ablation of the homophily range sampled during pre-training. Average NMI/ACC over the five homophilous (Homo), five heterophilous (Hetero), and all ten graphs (All).

Explore similar work

Jun 22, 2026cs.LG

Bridge the Gaps: Heterogeneous Attributed Graph Clustering via Quaternion Representation Learning

Attributed graph clustering partitions nodes by jointly exploiting node attributes and graph topology. It remains challenging due to attribute heterogeneity and representation degradation during graph learning. Real-world datasets often contain heterogeneous attributes, i.e., numerical and categorical attributes, complicating unified representation learning. This challenge becomes more complex in attributed graphs, where constructing a clustering-friendly graph structure from attributes and topology remains difficult. Under deep graph architectures, repeated graph propagation causes node embeddings to become overly similar, leading to the over-smoothing (OS) effect. Meanwhile, graph representation learning amplifies topological influence, making discriminative attribute information harder to exploit for clustering, an effect we refer to as over-dominating (OD). To bridge these gaps, an end-to-end framework, Any-type attributed Graph REpresentation lEarning (AGREE), is proposed. It unifies attributed graphs and any-type attributed data through multi-level alignment and similarity-based graph construction. Quaternion-based graph convolution strengthens attribute interaction to alleviate OD, while shallow graph architectures help relieve OS. The learned embeddings are jointly optimized for graph reconstruction and clustering, without requiring a predefined number of clusters during training. Experiments on diverse benchmarks show that AGREE achieves strong overall performance in accuracy, robustness, and adaptability.

Xinxi Chen, Junyang Chen, Yiqun Zhang +2

School of Computer Science and Technology, Guangdong University of Technology, Guangzhou 510006, China

Text-Attributed Graph Graph Representation Learning Clustering

Ask PDF

Jun 11, 2026cs.LG

Clustering Node Attributed Networks with Graph Neural Networks and Self Learning

Graph clustering - partitioning the node set of a graph into disjoint subsets that reflect some latent information - is a fundamental problem as it finds applications in a myriad of different scenarios. While this classic problem has been tackled for decades by different communities, a recent variation of the problem driven by real data considers the scenario where nodes have attributes that are also informative. This has triggered novel methods that simultaneously leverage network information (edges) and node information (attributed) in the design of novel clustering algorithms. This work proposes a novel framework that builds on prior works that have applied graph neural networks (GNN) to graph clustering. The proposed framework operates in rounds of self learning in a fully unsupervised setting. In each round, a GNN generates representations for nodes that are used to cluster the nodes. This clustering influences the graph used to generate the node representation in the next round. Moreover, a context graph built in each round using the original graph is used to generate the node representations. Empirical results show that the proposed methodology extracts information from both network edges and node attributes in synthetic data, outperforming algorithms focused solely on the network or attributes when neither are very informative. Multiple rounds of learning also improve the performance and always outperforms a long single round of training (i.e., classic GNN graph clustering). When considering real datasets, empirical results indicate that the proposed methodology is competitive to state-of-the-art methods when cluster sizes are balanced.

Rodrigo de Sapienza Luna, Daniel Ratton Figueiredo

Systems Engineering and Computer Science (PESC), Federal University of Rio de Janeiro (UFRJ), Technology Center, Block H, Room 319, 68511, Cidade Universitária, Rio de Janeiro, Brazil

Graph Neural Networks Community Detection Clustering Unsupervised Neural Network

Ask PDF

Sep 21, 2026cs.LG

G-NAC: Graph Neural Automata Clustering via Emergent Domain Formation

We introduce Graph Neural Automata Clustering (G-NAC), an unsupervised clustering method in which observations interact as cells on a fixed neighborhood graph. A shared recurrent graph-neural cellular rule evolves latent domain states through local interactions, which are converted into a rank-based spectral affinity for partitioning. Across 73 clustering tasks from 57 benchmark datasets, G-NAC achieved a mean adjusted Rand index (ARI) of 0.7951, comparable to Genie at 0.7941 and higher than the other evaluated baselines. Empirical training time and GPU memory scaled approximately linearly from 5,000 to 100,000 nodes. Learned transition rules also transferred from smaller source graphs to independent 100,000-node samples generated under matched conditions. These results demonstrate a recurrent graph-clustering formulation while identifying dependencies on graph quality, readout design, and source-target similarity.

Keith Miller, Tristan Crawford

Graph Neural Networks Emergence

Ask PDF