Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

Text-Attributed Graph Anomaly Detection via Multi-Scale Cross- and Uni-Modal Contrastive Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read The paper claims that training a text encoder and a graph encoder together with multi-scale cross-modal and uni-modal contrastive losses turns text-attributed graph anomaly detection into an inconsistency-scoring problem, and that the…

desk verdict New task and resources, but synthetic contextual anomalies are engineered to match the method's objective, so the SOTA margins are likely inflated. read the letter →

arxiv 2508.00513 v1 pith:OCYACUBV submitted 2025-08-01 cs.LG

classification cs.LG
keywords text-attributedgraphsgraphanomalydetectioncross-modalcontrastivelearningmulti-scaleself-supervisedscoreestimationneuralnetworkslanguagemodels
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that graph anomaly detection on text-attributed graphs – networks whose nodes carry raw text, such as papers with abstracts or products with reviews – is held back by a separation between text encoding and anomaly-detection training. It proposes CMUCL, an end-to-end method that trains a Transformer text encoder and a GNN graph encoder together, using contrastive losses that reward agreement between text and topology at both node and neighborhood scales. The central claim is that anomalies are precisely the nodes where these modalities disagree, so inconsistency scores computed from the trained contrasts can rank nodes by how anomalous they are. To back this, the paper releases eight text-attributed graph datasets with injected anomalies and reports an average 11.13% gain in average precision and a 4.68% gain in AUC over the runner-up across eleven baselines. It also claims to be the first to define the text-attributed graph anomaly detection task itself.

What carries the argument

The load-bearing mechanism is multi-scale cross-modal and uni-modal contrastive learning with an inconsistency-based anomaly score. The cross-modal part aligns text and graph representations of the same node, both at the node scale and between node and neighborhood-context scales; the uni-modal part aligns each node's representation with its own neighborhood context within the text domain and within the graph domain. Each contrastive view is an infoNCE-style loss where the positive pair is the same object in the other modality or scale and the negatives are other nodes in the batch. The anomaly score estimator then reads out inconsistencies: a normal node has high positive similarity, low negative similarities, and a small cross-entropy term, while an anomalous node deviates on one or more of those terms. The quadratic stability term across repeated sampling rounds is a second mechanism that exploits the instability of anomalous nodes under different negative samples.

What would settle it

Run CMUCL and a strong feature-only baseline on a text-attributed graph whose anomalies are human-verified rather than injected, for instance a citation network with known duplicate or fraudulent papers, and compare AUC: if the cross-modal advantage disappears, the synthetic anomaly protocol is the reason for the reported gains. A faster check is to measure per-type AUC separately on naturally occurring contextual and structural anomalies, since the paper's largest margins come from injected contextual anomalies.

Watch

Extended reading notes

Core claim

CMUCL treats a text-attributed graph as two views of the same underlying reality: a text view produced by a Transformer encoder reading each node's raw text, and a graph view produced by a GCN reading the node's neighborhood. Both views are computed at node level and at a coarser context level, the mean of neighboring node representations, and the training objective maximizes agreement across modalities and across scales while pushing apart representations of other nodes in the batch. Normal nodes are assumed to satisfy this multi-scale consistency; anomalous nodes violate it. At inference, the model scores each node by combining, over all contrastive views, the gap between negative and positive similarities plus the cross-entropy of the contrastive prediction, and it averages this across many sampling rounds while adding the variance across rounds as a stability penalty. The paper claims this yields state-of-the-art results on all eight released datasets, including a graph with over 1.1 million nodes, and that both contextual anomalies, where text is inconsistent with the neighborhood, and structural anomalies, where connection patterns are unusual, are detected better than by previous methods, with the largest margin on contextual anomalies. It presents this as the first formulation of text-attributed graph anomaly detection.

Load-bearing premise

The results stand on the assumption that anomalies made by transplanting text from a very dissimilar node and by randomly adding edges resemble the anomalies that actually occur in text-attributed graphs.

Editorial extensions

If this is right

  • Graph anomaly detection pipelines can use raw text directly, so text encoding no longer needs to be a frozen preprocessing step that ignores the detection objective.
  • Anomalies that are invisible to topology alone, such as a node whose text does not match its neighborhood, become detectable through cross-modal inconsistency.
  • Multi-scale contrasts matter: aligning node and context representations raises detection performance beyond single-scale alignment, as the ablation study shows.
  • The released eight datasets, spanning citation and e-commerce networks and including a 1.1-million-node graph, give the field a common test bed for text-attributed graph anomaly detection.
  • The framework scales to large graphs because it processes batches and does not need the full graph to be reconstructed, unlike reconstruction-based baselines that run out of memory.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If injected anomalies resemble real ones, the same cross-modal inconsistency signal could transfer to other text-plus-structure domains, such as detecting fraudulent product listings whose text and purchase graphs disagree.
  • The ablation finding that deeper text encoders help suggests a testable scaling hypothesis the paper does not pursue: detection of contextual anomalies should keep improving as the text encoder grows.
  • Because the framework treats text and graph as two views of one reality, the same contrastive objective could attach other node modalities, such as images or user activity logs, to the same graph encoder.
  • A frozen-text-encoder variant would isolate how much of the reported 11.13% AP gain comes from joint training itself rather than from the multi-scale contrastive architecture, an experiment the paper's ablations do not run.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes CMUCL, an unsupervised end-to-end framework for anomaly detection on text-attributed graphs. It jointly trains a Transformer text encoder and a GCN graph encoder using cross-modal and uni-modal multi-scale contrastive losses, and derives anomaly scores from positive/negative similarities and cross-entropy stability across multiple sampling rounds. The authors release eight text-attributed graph datasets with injected anomalies, compare CMUCL against eleven baselines, and report average gains of 4.68% AUC and 11.13% AP over the runner-up.

Significance. If the reported gains are robust, the paper makes a useful contribution: it identifies a real limitation of prior GAD pipelines that rely on frozen, shallow text features, proposes a coherent end-to-end training framework, and releases large-scale benchmarks that could support future work. The experiments are extensive, including eight datasets, eleven baselines, ablations, parameter studies, and runtime analyses, and code/data are promised. However, the main empirical claim depends critically on the validity of the anomaly-injection procedure, which is causally aligned with the method's objective, and the absence of a clearly described validation protocol weakens the confidence in the reported SOTA margins.

major comments (3)
  1. [Appendix A and Section 5.1] The benchmark construction is likely to inflate CMUCL's advantage. Contextual anomalies are generated by selecting, for each target node, the candidate with the lowest BGE cosine similarity and inserting or replacing its text; Section 5.1 then uses the same BGE encoder to featurize all methods. The labels are therefore not an independent sample of real-world anomalies: they are generated by a similarity notion that is directly encoded in the features every method receives, and they instantiate exactly the text-graph inconsistency that Eqs. (4)-(10) are designed to penalize. Table 4 is consistent with this concern: CMUCL's largest relative gain is on contextual anomalies (65.09 vs 59.30 AUC), the category manufactured by least-similar text transplant. This does not make the method internally inconsistent, but it makes the headline SOTA claim a benchmark-validity risk. To support the central claim, the authors should evaluate on anomaly labels that are not constructed by this least-similar BGE procedure (e.g., real or manually annotated anomalies, or injection based on random topical shifts independent of the encoders used by all methods), and report per-anomaly-type results on those benchmarks.
  2. [Section 5.1 and Appendix B] No validation procedure is described for hyperparameter selection. The paper frames the comparison as fully unsupervised, yet Appendix B lists per-dataset learning rates, gamma values, and epoch counts selected from specified ranges. If those values were chosen using the test labels, the comparison is not fully unsupervised and the reported gains may be optimistically biased. The authors should either describe a label-free validation protocol (e.g., model selection based on a pretext-score criterion or a held-out split of normal nodes) or explicitly state how the reported hyperparameters were selected.
  3. [Section 5.5] The complexity analysis appears inconsistent with the batched implementation. The paper states a total complexity of O(|V| d(|V| + L^2 + d) + |E| d), with the O(|V|^2 d) term attributed to similarity calculation over all comparison pairs. In the actual method, the contrastive losses in Eqs. (4)-(11) are computed within batches of size N, giving O(N^2 d) per batch and O(|V| N d) per epoch if N is constant, not O(|V|^2 d). Please clarify whether the reported complexity assumes full-batch computation and, if so, how the batched training/inference in the experiments is reconciled with that assumption.
minor comments (5)
  1. [Abstract] The phrase 'average accuracy (AP)' is imprecise; AP is average precision, not accuracy.
  2. [Section 5.1] The baseline name 'AENMONE' is a typo for ANEMONE.
  3. [Section 4.2] The sentence introducing Eq. (7) says 'The cross-modal inner-scale loss is expressed as L_inter'; this should read 'inter-scale' for consistency with the surrounding text.
  4. [Appendix A] The sentence 'we follow standard construction methods from prior research [6, 22, 8, 8]' contains a duplicated reference '[8, 8]'; one citation should be removed.
  5. [Section 4.1 and Figure 1] The term 'bi-model' should be 'bi-modal' throughout the methodology section and figure captions.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: CMUCL is a self-contained empirical method validated against external baselines; the main concern (synthetic anomaly labels are aligned with the method's inductive bias) is a benchmark-validity issue, not a circular derivation.

full rationale

The paper's contribution is empirical and architectural. CMUCL trains a text Transformer (Eq. 1) and a GCN (Eq. 2) with cross-modal (Eqs. 4-10) and uni-modal (Eq. 11) contrastive losses, then converts the trained contrastive signals into anomaly scores (Eqs. 13-15). No parameter is fitted to the anomaly labels, and the reported AUC/AP are computed against eleven external baselines on released benchmarks. The self-citations ([43]-[45] and related) appear in background/related-work contexts and are not load-bearing for the central derivation. The main substantive concern is Appendix A's label-generation protocol: contextual anomalies are created by transplanting text from the BGE-least-similar candidate node into a target node, which directly instantiates the text-vs-neighborhood inconsistency that CMUCL's cross-modal contrast is designed to penalize, and the same BGE encoder is used to featurize inputs for all methods. This makes the benchmark causally aligned with CMUCL's inductive bias and is a legitimate threat to external validity, but it is not circular: the anomaly scores are not computed from the BGE similarity used to generate labels, and the method still must learn the inconsistency signal through training and generalizes to structural anomalies (Table 4). Under the definition of circularity used here (a claim reducing to its inputs by construction or to a fitted parameter renamed as a prediction), the derivation is self-contained and the score is low.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The method uses standard neural network components and a benchmark construction procedure; no new physical or conceptual entities are postulated. The main load-bearing assumptions are the cross-modal consistency hypothesis, the adequacy of BGE features as graph input, and the representativeness of the synthetic anomaly injection.

free parameters (8)
  • learning_rate = 2e-4 (Citeseer), 2e-5 (Pubmed, History, Computers, CitationV8), 5e-5 (Photo, Children), 1e-5 (ogbn-Arxiv)
    Selected per dataset from {1e-5, 2e-5, 5e-5, 2e-4}; no validation split or selection protocol is described.
  • gamma (trade-off parameter) = 5e-3, 1e-3, 0.5, 1e-3, 1e-2, 0.5, 1e-2, 0.5 for the eight datasets
    Tuned per dataset from {0.001, 0.005, 0.01, 0.5, 1.0, 1.5, 2.0, 5.0, 10.0}; balances cross-modal and uni-modal losses.
  • epochs = 2, 2, 2, 3, 3, 2, 2, 2 for the eight datasets
    Selected from {1, 2, 3}; no validation protocol stated.
  • sampling rounds R = 256
    Set after parameter study showing gains plateau beyond R=64; affects the stability term in the anomaly score.
  • temperature tau = 0.07
    Fixed contrastive learning temperature, standard in InfoNCE-style objectives.
  • GCN hidden dimension = 128
    Architecture choice for the graph encoder and feature alignment.
  • Transformer layers, width, heads = 12 layers, width 512, 8 heads
    Selected after the layer study in Figure 4; larger text encoders generally improve performance.
  • anomaly injection candidate set size K = 50
    Chosen in dataset construction to make contextual perturbations large enough; affects benchmark difficulty.
assumptions (5)
  • domain assumption Normal nodes satisfy cross-modal consistency between text and graph structure, while anomalies violate this consistency.
    Section 1 states that normal nodes strive to recover a representation of reality that satisfies a consistency principle, while abnormal objects may not adhere to this alignment. This is the core inductive bias that justifies the contrastive objective.
  • domain assumption BGE embeddings of raw text are adequate initial node features for the graph encoder, while the text encoder consumes raw text.
    Section 5.1 encodes raw text with BGE for all baselines and for the graph domain in CMUCL. The method assumes these two views are complementary and can be aligned in a shared space.
  • domain assumption The InfoNCE contrastive loss with batch negatives produces useful representations without collapse.
    The training objective in Eqs. (4)-(11) relies on the standard contrastive learning assumption that in-batch negatives provide sufficient discrimination and that the encoder does not collapse to a trivial solution.
  • domain assumption Synthetic anomalies generated by text insertion/replacement and random edge addition are representative of real-world anomalies in text-attributed graphs.
    Appendix A describes the injection procedure. The evaluation of CMUCL and all baselines uses these synthetic labels, so the validity of the empirical claim depends on this representativeness.
  • domain assumption A 12-layer Transformer can be trained effectively with the proposed contrastive objective without requiring pre-training or a dedicated initialization scheme.
    Implementation details describe the Transformer architecture but do not state initialization or pre-training. The reported success implies this assumption holds, but it is not tested or discussed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Text-Attributed Graph Anomaly Detection via Multi-Scale Cross- and Uni-Modal Contrastive Learning." pith.science (2026). https://pith.science/paper/OCYACUBV

@misc{pith2026250800513,
  author       = {Pith},
  title        = {Pith review of: Text-Attributed Graph Anomaly Detection via Multi-Scale Cross- and Uni-Modal Contrastive Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OCYACUBV}},
  note         = {Machine review of arXiv:2508.00513}
}
read the original abstract

The widespread application of graph data in various high-risk scenarios has increased attention to graph anomaly detection (GAD). Faced with real-world graphs that often carry node descriptions in the form of raw text sequences, termed text-attributed graphs (TAGs), existing graph anomaly detection pipelines typically involve shallow embedding techniques to encode such textual information into features, and then rely on complex self-supervised tasks within the graph domain to detect anomalies. However, this text encoding process is separated from the anomaly detection training objective in the graph domain, making it difficult to ensure that the extracted textual features focus on GAD-relevant information, seriously constraining the detection capability. How to seamlessly integrate raw text and graph topology to unleash the vast potential of cross-modal data in TAGs for anomaly detection poses a challenging issue. This paper presents a novel end-to-end paradigm for text-attributed graph anomaly detection, named CMUCL. We simultaneously model data from both text and graph structures, and jointly train text and graph encoders by leveraging cross-modal and uni-modal multi-scale consistency to uncover potential anomaly-related information. Accordingly, we design an anomaly score estimator based on inconsistency mining to derive node-specific anomaly scores. Considering the lack of benchmark datasets tailored for anomaly detection on TAGs, we release 8 datasets to facilitate future research. Extensive evaluations show that CMUCL significantly advances in text-attributed graph anomaly detection, delivering an 11.13% increase in average accuracy (AP) over the suboptimal.

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. CAMERA: Adapting to Semantic Camouflage in Unsupervised Text-Attributed Graph Fraud Detection

    cs.LG 2026-05 unverdicted novelty 7.0 of 10

    CAMERA is an ego-decoupled mixture-of-experts model with context-informed gating and one-class objectives for unsupervised fraud detection in text-attributed graphs facing semantic camouflage.

  2. Node-to-Neighborhood Semantic Consistency: Text-Topology Alignment for TAGs Anomaly Detection

    cs.CL 2026-06 unverdicted novelty 5.0 of 10

    N2NSC framework detects anomalies in text-attributed graphs by enforcing node-to-neighborhood semantic consistency via two complementary fusion paths that align textual semantics with topology.

Reference graph

Works this paper leans on

51 extracted references · 47 canonical work pages · cited by 2 Pith papers

  1. [1]

    Bandyopadhyay, N

    S. Bandyopadhyay, N. Lokesh, and M. N. Murty. Outlier aware network embedding for attributed networks. In AAAI, 2019

  2. [2]

    M. M. Breunig, H.-P. Kriegel, R. T. Ng, and J. Sander. Lof: identifying density-based local outliers. In SIGMOD, pages 93–104, 2000

  3. [3]

    R. Chen, T. Zhao, A. Jaiswal, N. Shah, and Z. Wang. Llaga: Large language and graph assistant. arXiv preprint arXiv:2402.08170, 2024

  4. [4]

    Z. Chen, H. Mao, H. Li, W. Jin, H. Wen, X. Wei, S. Wang, D. Yin, W. Fan, H. Liu, et al. Exploring the potential of large language models (llms) in learning on graphs. SIGKDD, 2024

  5. [5]

    Chien, W.-C

    E. Chien, W.-C. Chang, C.-J. Hsieh, H.-F. Yu, J. Zhang, O. Milenkovic, and I. S. Dhillon. Node feature extraction by self-supervised multi-scale neighborhood prediction. ICLR, 2022

  6. [6]

    K. Ding, J. Li, R. Bhanushali, and H. Liu. Deep anomaly detection on attributed networks. In SDM, pages 594–602, 2019

  7. [7]

    K. Ding, J. Li, N. Agarwal, and H. Liu. Inductive anomaly detection on attributed networks. In IJCAI, pages 1288–1294, 2021

  8. [8]

    J. Duan, S. Wang, P. Zhang, E. Zhu, J. Hu, H. Jin, Y . Liu, and Z. Dong. Graph anomaly detection via multi-scale contrastive learning networks with augmented view. In AAAI, 2023

Show all 51 references
  1. [9]

    Gasteiger, S

    J. Gasteiger, S. Weißenberger, and S. Günnemann. Diffusion improves graph learning. NeurIPS, 32, 2019

  2. [10]

    Hassani and A

    K. Hassani and A. H. Khasahmadi. Contrastive multi-view representa- tion learning on graphs. In ICML, 2020

  3. [11]

    X. He, X. Bresson, T. Laurent, A. Perold, Y . LeCun, and B. Hooi. Har- nessing explanations: Llm-to-lm interpreter for enhanced text-attributed graph representation learning. ICLR, 2024

  4. [12]

    J. Hu, B. Xiao, H. Jin, J. Duan, S. Wang, Z. Lv, S. Wang, X. Liu, and E. Zhu. Samcl: Subgraph-aligned multiview contrastive learning for graph anomaly detection. TNNLS, 2023

  5. [13]

    M. Huh, B. Cheung, T. Wang, and P. Isola. The platonic representation hypothesis. arXiv preprint arXiv:2405.07987, 2024

  6. [14]

    B. Jin, G. Liu, C. Han, M. Jiang, H. Ji, and J. Han. Large lan- guage models on graphs: A comprehensive survey. arXiv preprint arXiv:2312.02783, 2023

  7. [15]

    M. Jin, Y . Liu, Y . Zheng, L. Chi, Y .-F. Li, and S. Pan. Anemone: Graph anomaly detection with multi-scale contrastive learning. In CIKM, pages 3122–3126, 2021

  8. [16]

    T. N. Kipf and M. Welling. Semi-supervised classification with graph convolutional networks. ICLR, 2017

  9. [17]

    J. Li, H. Dani, X. Hu, and H. Liu. Radar: Residual analysis for anomaly detection in attributed networks. In IJCAI, volume 17, pages 2152– 2158, 2017

  10. [18]

    P. Li, H. Yu, and X. Luo. Context-aware graph neural network for graph-based fraud detection with extremely limited labels. In AAAI, volume 39, pages 12112–12120, 2025

  11. [19]

    Y . Li, Z. Li, P. Wang, J. Li, X. Sun, H. Cheng, and J. X. Yu. A survey of graph meets large language model: Progress and future directions. arXiv preprint arXiv:2311.12399, 2023

  12. [20]

    Y . Lin, J. Tang, C. Zi, H. V . Zhao, Y . Yao, and J. Li. Unigad: Unifying multi-level graph anomaly detection. NeurIPS, 2024

  13. [21]

    K. Liu, Y . Dou, Y . Zhao, X. Ding, X. Hu, R. Zhang, K. Ding, C. Chen, H. Peng, K. Shu, et al. Bond: Benchmarking unsupervised outlier node detection on static attributed graphs. NeurIPS, 35:27021–27035, 2022

  14. [22]

    Y . Liu, Z. Li, S. Pan, C. Gong, C. Zhou, and G. Karypis. Anomaly de- tection on attributed networks via contrastive self-supervised learning. TNNLS, 33(6):2378–2392, 2021

  15. [23]

    Y . Liu, S. Li, Y . Zheng, Q. Chen, C. Zhang, and S. Pan. Arc: A generalist graph anomaly detector with in-context learning. NeurIPS, 2024

  16. [24]

    X. Ma, J. Wu, S. Xue, J. Yang, C. Zhou, Q. Z. Sheng, H. Xiong, and L. Akoglu. A comprehensive survey on graph anomaly detection with deep learning. TKDE, 35(12):12012–12038, 2021

  17. [25]

    J. Ni, J. Li, and J. McAuley. Justifying recommendations using distantly-labeled reviews and fine-grained aspects. InEMNLP-IJCNLP, pages 188–197, 2019

  18. [26]

    Pandit, D

    S. Pandit, D. H. Chau, S. Wang, and C. Faloutsos. Netprobe: a fast and scalable system for fraud detection in online auction networks. In WWW, pages 201–210, 2007

  19. [27]

    Z. Peng, W. Huang, M. Luo, Q. Zheng, Y . Rong, T. Xu, and J. Huang. Graph representation learning via graphical mutual information maxi- mization. In WWW, pages 259–270, 2020

  20. [28]

    Perozzi, L

    B. Perozzi, L. Akoglu, P. Iglesias Sánchez, and E. Müller. Focused clustering and outlier detection in large attributed graphs. In SIGKDD, 2014

  21. [29]

    A. Roy, J. Shu, J. Li, C. Yang, O. Elshocht, J. Smeets, and P. Li. Gad- nr: Graph anomaly detection via neighborhood reconstruction. WSDM, 2023

  22. [30]

    A. Roy, J. Shu, J. Li, C. Yang, O. Elshocht, J. Smeets, and P. Li. Gad-nr: Graph anomaly detection via neighborhood reconstruction. In WSDM, pages 576–585, 2024

  23. [31]

    Sakurada and T

    M. Sakurada and T. Yairi. Anomaly detection using autoencoders with nonlinear dimensionality reduction. In MLSDA workshop, pages 4–11, 2014

  24. [32]

    B. Shi, B. Dong, Y . Xu, J. Wang, Y . Wang, and Q. Zheng. An edge feature aware heterogeneous graph neural network model to support tax evasion detection. ESWA, 213:118903, 2023

  25. [33]

    K. Shin, B. Hooi, J. Kim, and C. Faloutsos. Densealert: Incremental dense-subtensor detection in tensor streams. In SIGKDD, 2017

  26. [34]

    D. B. Skillicorn. Detecting anomalies in graphs. In 2007 IEEE Intelli- gence and Security Informatics , pages 209–216. IEEE, 2007

  27. [35]

    X. Song, M. Wu, C. Jermaine, and S. Ranka. Conditional anomaly detection. TKDE, 19(5):631–645, 2007

  28. [36]

    J. Tang, F. Hua, Z. Gao, P. Zhao, and J. Li. Gadbench: Revisiting and benchmarking supervised graph anomaly detection. NeurIPS, 36, 2024

  29. [37]

    Van den Oord, Y

    A. Van den Oord, Y . Li, and O. Vinyals. Representation learning with contrastive predictive coding. arXiv e-prints, pages arXiv–1807, 2018

  30. [38]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin. Attention is all you need. NeurIPS, 30, 2017

  31. [39]

    Velickovic, W

    P. Velickovic, W. Fedus, W. L. Hamilton, P. Liò, Y . Bengio, and R. D. Hjelm. Deep graph infomax. ICLR, 2019

  32. [40]

    Wen and Y

    Z. Wen and Y . Fang. Augmenting low-resource text classification with graph-grounded pre-training and prompting. In SIGIR, pages 506–516, 2023

  33. [41]

    S. Xiao, Z. Liu, P. Zhang, and N. Muennighoff. C-pack: Packaged re- sources to advance general chinese embedding, 2023

  34. [42]

    X. Xu, N. Yuruk, Z. Feng, and T. A. Schweiger. Scan: a structural clustering algorithm for networks. In SIGKDD, 2007

  35. [43]

    Y . Xu, B. Shi, T. Ma, B. Dong, H. Zhou, and Q. Zheng. Cldg: Con- trastive learning on dynamic graphs. In ICDE, pages 696–707. IEEE, 2023

  36. [44]

    Y . Xu, Z. Peng, B. Shi, X. Hua, and B. Dong. Learning dynamic graph representations through timespan view contrasts.Neural Networks, 176: 106384, 2024

  37. [45]

    Y . Xu, Z. Peng, B. Shi, X. Hua, B. Dong, S. Wang, and C. Chen. Re- visiting graph contrastive learning on anomaly detection: A structural imbalance perspective. In AAAI, volume 39, pages 12972–12980, 2025

  38. [46]

    Y . Xu, B. Shi, B. Dong, J. Wang, H. Wei, and Q. Zheng. Ted: related party transaction guided tax evasion detection on heterogeneous graph. Data Mining and Knowledge Discovery , 39(2):15, 2025

  39. [47]

    H. Yan, C. Li, R. Long, C. Yan, J. Zhao, W. Zhuang, J. Yin, P. Zhang, W. Han, H. Sun, et al. A comprehensive study on text-attributed graphs: Benchmarking and rethinking. NeurIPS, 36:17238–17264, 2023

  40. [48]

    Zhang, S

    J. Zhang, S. Wang, and S. Chen. Reconstruction enhanced multi-view contrastive learning for anomaly detection on attributed networks. IJ- CAI, 2022

  41. [49]

    J. Zhao, M. Qu, C. Li, H. Yan, Q. Liu, R. Li, X. Xie, and J. Tang. Learn- ing on large-scale text-attributed graphs via variational inference.ICLR, 2023

  42. [50]

    Zheng, Y

    Q. Zheng, Y . Xu, H. Liu, B. Shi, J. Wang, and B. Dong. A survey of tax risk detection using data mining techniques. Engineering, 34:43–59, 2024

  43. [51]

    Zheng, M

    Y . Zheng, M. Jin, Y . Liu, L. Chi, K. T. Phan, and Y .-P. P. Chen. Gener- ative and contrastive self-supervised learning for graph anomaly detec- tion. TKDE, 35(12):12220–12233, 2021. Appendix A Datasets details The eight widely used benchmark text-attributed graph datasets i...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.