Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Relation-Aware Graph Foundation Model

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Typed relations, not nodes or datasets, can serve as the shared vocabulary of a graph foundation model, enabling cross-domain pretraining and transfer.

desk verdict Good architecture, but the load-bearing transfer claim isn't supported: unseen relations fail, and the 'out-of-domain' experiment is mislabeled. read the letter →

arxiv 2505.12027 v1 pith:IFFGY5BY submitted 2025-05-17 cs.LG

Jianxiang Yu , Jiapeng Zhu , Hao Qian , Ziqi Liu , Zhiqiang Zhang , Xiang Li This is my paper · ORCID
classification cs.LG
keywords graphfoundationmodelsrelationtokenshypernetworkscross-domainpretrainingfew-shottransferlinkpredictionSentence-BERTembeddingsdataaugmentation
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 claims that typed relations between nodes—citation, hyperlink, co-purchase—are the right basic units for a graph foundation model, playing the role that tokens play in large language models. It proposes REEF, which converts textual descriptions of relation types into language-model embeddings, collects those embeddings into a shared relation vocabulary, and uses hypernetworks to generate relation-specific GNN aggregators and classifiers from each token. Because the same relation type recurs across datasets in a domain, pretraining on several graphs transfers to an unseen graph of that domain. The paper reports that this design beats GNN, contrastive, and existing graph-foundation baselines on pretraining and few-shot transfer, and that adding more pretraining datasets with a richer relation vocabulary improves average accuracy. The core claim is that relational semantics, not node identity or dataset identity, is what carries transferable graph knowledge.

What carries the argument

The central machinery is the relation vocabulary plus the hypernetworks that consume it. A relation token is the text description of an edge or task type embedded by Sentence-BERT and stored in one shared vocabulary over all pretraining datasets. From each token embedding, the aggregator hypernetwork $F_{\mathrm{Agg}}$ generates layerwise message-passing parameters $\Phi_r^{(l)}$, and the classifier hypernetwork $F_{\mathrm{Cls}}$ generates relation-specific classifier weights $\Psi_r$, so the GNN's behavior is conditioned on the semantics of the relation rather than on the dataset it came from. A third hypernetwork $F_{\mathrm{Proj}}$ maps the dataset description embedding to a projector $\Phi_G$, and a feature bias $h_G$ is added to each node's initial representation, allowing relation-level sharing while adapting to dataset-specific feature distributions. The pretraining objective is binary relation prediction on subgraph pairs, trained with mixed-dataset batching and edge-drop augmentation.

What would settle it

Take a citation graph whose edges are structurally identical but whose relation description is replaced with a semantically different phrase, and test transfer to Cora; if accuracy does not change, the language-model semantics are not load-bearing, and if it collapses, the consistency assumption is doing the work. A second check: compare PubMed-only versus Citeseer-only pretraining on Cora, where the paper itself predicts a gap because the two citation networks come from different fields.

Watch

Extended reading notes

Core claim

REEF's central claim is that a graph foundation model can be built on a vocabulary of relation tokens instead of node or dataset tokens. Each relation type is described in words, embedded with Sentence-BERT, and the embedding is passed to two hypernetworks that produce the weight matrices of that relation's message-passing aggregator and its task classifier; a third hypernetwork produces a dataset-specific feature projector and a feature bias from the dataset's text description. Pretraining is formulated as binary relation prediction between two subgraphs—does relation $r$ hold between $s_i$ and $s_j$?—with seven pretraining datasets spanning citation, WebKB, Amazon, and knowledge-graph domains and a shared vocabulary of 254 relation tokens. The same aggregator and classifier tokens are reused when fine-tuning on target datasets from the same domain. Reported results include a pretraining average accuracy of 79.70% on five benchmark datasets, accuracy gains over the best transfer baseline of 10.07–24.05% on three $C$-way-1-shot target graphs, and link-prediction accuracies of 91.04% on FB15K237 and 94.70% on WN18RR.

Load-bearing premise

The load-bearing premise is that the language-model embedding of a textual relation description is semantically consistent enough across datasets and domains that one relation token can generate useful aggregator and classifier weights for all of them, as when a citation token pretrained on PubMed and Citeseer is reused for Cora.

Editorial extensions

If this is right

  • Pretraining on more datasets with a richer relation vocabulary raises average downstream accuracy, a scaling trend the paper documents in Section 4.4.
  • A relation token learned on one dataset can be reused at fine-tuning time on an unseen dataset in the same domain, so citation and hyperlink knowledge transfers without retraining.
  • Relation-token pretraining extends beyond node classification to link prediction, outperforming the tree-vocabulary graph foundation baseline on both FB15K237 and WN18RR.
  • The ablation study shows that removing any of the three components—LM-initialized relation embeddings, dataset projector and feature bias, or edge-drop augmentation—lowers average performance, indicating each carries part of the transfer.
  • Out-of-domain transfer improves with relation diversity in pretraining: dropping WebKB and Amazon relations from the pretraining mix sharply degrades transfer to Cornell and Computers.

Reading between the lines

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

  • Editorial inference: if relation embeddings are the reusable unit, the vocabulary is open-ended, so a new relation type could be added by embedding its description without retraining; the paper does not run this zero-shot relation-extension experiment.
  • Editorial inference: the architecture splits transfer cleanly—relation tokens control aggregator and classifier behavior while dataset tokens control feature projection—so node-level tokenization could be added to the same backbone to combine both transfer mechanisms.
  • Editorial inference: the scaling results imply that relation vocabulary size, not node count or parameter count, may be the natural capacity measure for a graph foundation model, which a dedicated vocabulary-scaling study could test directly.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The manuscript proposes REEF, a graph foundation model built on relation tokens. Each edge or task type is mapped to a textual description, embedded with Sentence-BERT, and two hypernetworks generate relation-specific aggregator and classifier parameters; a third hypernetwork generates dataset-specific projectors and feature biases, and pretraining combines mixed datasets with edge-drop augmentation. The experiments evaluate pretraining accuracy on five datasets, C-way 1-shot transfer to Cora, Cornell, and Computers, link prediction on FB15K237 and WN18RR, ablations, and scaling analyses.

Significance. The relation-token design is original and the framework is coherent; the ablation study isolates the contributions of LM initialization, feature projector/bias, and augmentation, and the paper provides implementation details and code. If the claimed transferability held, this would be an important step toward graph foundation models with a shared relational vocabulary. However, as detailed below, the evidence for generalization to unseen relation types is negative, and the transfer experiments use relations already seen during pretraining, so the paper's strongest claims are currently not supported.

major comments (4)
  1. [Appendix F.3, Table 6] The out-of-domain experiment is mislabeled and, in the only genuinely unseen-relation condition, REEF performs at chance level. The first row uses the full seven-dataset pretraining set, which includes Wisconsin, Texas, and Photo, so the hyperlinks, webpage classification, co-purchase, and product classification tokens are all present in the relation vocabulary; this row is therefore an in-vocabulary transfer test. The second row (Pubmed, Citeseer, FB15K237, WN18RR) excludes those domains, but Cornell accuracy is 34.58±12.87 (chance is 20%) and Computers accuracy is 26.40±11.19 with AUC 47.65 and F1 6.10, which is chance-level. This directly contradicts the text's claim that REEF can still effectively transfer knowledge and adapt to new, unseen relations.
  2. [Section 4.3, Table 2] The main transfer results do not test vocabulary generalization. Cora, Cornell, and Computers reuse the exact relation tokens (citation/paper classification, hyperlink/webpage classification, co-purchase/product classification) that were pretrained on Pubmed/Citeseer, Wisconsin/Texas, and Photo respectively, as stated in Section 4.1.2. At best, these results show that reusing a known relation token across datasets in the same domain helps; they say nothing about whether the relation vocabulary can be extended to unseen relations.
  3. [Section 4.4, Figure 3] The scaling-law claim is not supported by the evidence. There are only six pretraining configurations, the trend is non-monotonic, and the citation-domain transfer results decrease when Pubmed is added to Citeseer (Cora accuracy drops from 49.69 to 44.97). The average curve is dominated by pretraining accuracy on the added datasets, so it does not demonstrate that larger pretraining sets improve transfer. A monotonic trend over more scales and repetitions, or a fitted model with error bars, would be needed to justify the term scaling law.
  4. [Section 4.2, Table 1] No variance or significance information is reported for the pretraining comparisons, and the experimental setup is not commensurable: REEF's per-relation binary classifiers are averaged over binary tasks, while the joint GCN/GAT/RGCN baselines are trained with a single multi-class head over the union of label spaces. The headline highest average accuracy of 79.70% therefore conflates different prediction regimes. Please report standard deviations, significance tests, and a matched evaluation protocol, and note that REEF does not beat GCN or GAT on Pubmed and Citeseer in Table 1.
minor comments (4)
  1. [Appendix F.3] The text says the pretraining configuration includes Pubmed, Citeseer, Wisconsin, Texas, Cornell, FB15K237, and WN18RR, but Cornell is a target dataset, not a pretraining dataset; this is inconsistent with Table 6 and should be corrected.
  2. [Throughout] There are several typos and formatting issues, including pre-training performace in Section 4.4, Imporve. in Table 2, differenet in Appendix F, and T exas in Figure 3; a careful proofread is needed.
  3. [Section 4.1.2] The relation vocabulary size of 254 is stated without a derivation; please clarify whether it counts both aggregator and classifier tokens for each dataset and all 237 plus 11 knowledge-graph relation types, since Table 7 suggests the count should be 248 plus 2 plus 2 plus 2 equals 254.
  4. [Section 6] The limitation discussion focuses only on homogeneous graphs and does not address the core assumption that a shared relation description has consistent semantics across domains; the paper should acknowledge that the LM-based relation representation was not directly validated for cross-field consistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: REEF's relation tokens are LM-initialized, hypernetwork-generated parameters are trained and evaluated on independent splits, and no reported prediction reduces to fitted inputs or self-citation chains.

full rationale

I examined REEF's derivation chain: relation tokens are initialized from frozen Sentence-BERT embeddings (Eq. 3), hypernetworks map these embeddings to aggregator, classifier, and projector parameters (Eqs. 4, 6, 8, 10), and pretraining is a binary relation-existence objective (Eq. 7) trained on held-out splits with mixed-dataset transfer to target datasets. None of the reported 'predictions' are fitted constants: pretraining accuracy (Table 1) is evaluated on test sets, transfer accuracy (Table 2) uses target datasets with 1:9 validation/test splits and multiple runs, and link prediction (Table 4) follows GFT's data splits. The ablation REEF-LM shows that LM initialization matters, which is an empirical check rather than a circular construction. The self-citations appearing in the references ([10], [47]) are used only for related-work context on graph prompting and are not load-bearing for the main claims; no uniqueness theorem is imported from the authors' prior work. The experimental concern that Table 6's 'out-of-domain' condition is weakened by the first row containing target-domain relations, and that the genuinely unseen-relation row performs near random, is a validity and overclaim issue, not a circularity of derivation: those numbers are genuine evaluations, not reductions of outputs to inputs. The Limitations section (Section 6) is brief and only notes the homogeneous-graph scope, but that is a completeness concern rather than circularity. Under the required standard of exhibiting a specific reduction of a claimed prediction to its own inputs, no circular step can be identified.

Assumptions & free parameters 8 free parameters · 5 assumptions · 1 invented entities

REEF rests on a large set of chosen hyperparameters (dimensions, layers, learning rate, augmentation rate) and on the domain assumption that textually described relations are reusable across datasets. No physically invented entities are introduced; the 'relation token' is a representational construct whose utility is tested empirically. The paper itself acknowledges the homogeneous-graph limitation in Section 6.

free parameters (8)
  • SVD feature dimension = 128
    All node features are reduced to 128 dimensions via SVD (Appendix C); this dimension is chosen by hand and affects alignment.
  • hidden dimension = 64
    Set to 64; Appendix F.4 shows performance varies with this value.
  • number of GNN layers L = 2
    Fixed to 2 for REEF and baselines (Appendix C).
  • edge mask rate = 0.2
    Random edge removal probability in graph data augmentation (Appendix C).
  • learning rate = 0.0002
    Adam optimizer hyperparameter (Appendix C).
  • dropout rate = 0.5
    Dropout for training (Appendix C).
  • batch size = 128
    Batch size for pre-training (Appendix C).
  • pre-training epochs = 100
    Number of epochs (Appendix C).
assumptions (5)
  • domain assumption Textual relation descriptions encoded by Sentence-BERT capture reusable relational semantics across datasets
    The whole framework maps LM(relation text) to GNN parameters through hypernetworks; ablation REEF-LM shows random relation vectors hurt performance, but cross-domain transfer assumes descriptions generalize.
  • domain assumption Relations are consistent across datasets in the same domain (citation, hyperlink, co-purchase)
    Used to share relation tokens between source and target (e.g., Cora uses citation token from Pubmed/Citeseer). Stated in Section 1 and used in Section 4.1.2.
  • domain assumption Homogeneous graphs only
    Section 6 states the study focuses on homogeneous graphs; heterogeneous extension is future work, so claims are scoped to homogeneous graphs.
  • standard math SVD is an appropriate feature alignment operator
    Used to project heterogeneous feature dimensions to 128 (Eq. 2, Section 3.1).
  • standard math GNN message passing with normalization 1/|N_i^r| is a valid aggregation
    Standard GNN assumption, Eq. (5).
invented entities (1)
  • Relation token vocabulary independent evidence
    purpose: A set of textual-description-derived tokens serving as the basic transferable unit for GFM pre-training and adaptation
    The framework's central novelty; its usefulness is empirically assessed via transfer to unseen datasets, but it is not an independently verifiable physical quantity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Relation-Aware Graph Foundation Model." pith.science (2026). https://pith.science/paper/IFFGY5BY

@misc{pith2026250512027,
  author       = {Pith},
  title        = {Pith review of: Relation-Aware Graph Foundation Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IFFGY5BY}},
  note         = {Machine review of arXiv:2505.12027}
}
read the original abstract

In recent years, large language models (LLMs) have demonstrated remarkable generalization capabilities across various natural language processing (NLP) tasks. Similarly, graph foundation models (GFMs) have emerged as a promising direction in graph learning, aiming to generalize across diverse datasets through large-scale pre-training. However, unlike language models that rely on explicit token representations, graphs lack a well-defined unit for generalization, making it challenging to design effective pre-training strategies. In this work, we propose REEF, a novel framework that leverages relation tokens as the basic units for GFMs. Inspired by the token vocabulary in LLMs, we construct a relation vocabulary of relation tokens to store relational information within graphs. To accommodate diverse relations, we introduce two hypernetworks that adaptively generate the parameters of aggregators and classifiers in graph neural networks based on relation tokens. In addition, we design another hypernetwork to construct dataset-specific projectors and incorporate a dataset-level feature bias into the initial node representations, enhancing flexibility across different datasets with the same relation. Further, we adopt graph data augmentation and a mixed-dataset pre-training strategy, allowing REEF to capture relational diversity more effectively and exhibit strong generalization capabilities. Extensive experiments show that REEF significantly outperforms existing methods on both pre-training and transfer learning tasks, underscoring its potential as a powerful foundation model for graph-based applications.

Figures

Figures reproduced from arXiv: 2505.12027 by the authors.

Figure 1
Figure 1. Relation vocabulary inspired by LLMs. graph data is inherently complex and diverse, with its basic units being implicit. This presents a significant challenge for GFMs in designing appropriate specialized units. Some existing methods primarily treat nodes or datasets as basic units for pre-training [13, 45, 16, 39]. They either directly learn distinct representations for each node or employ a mixture￾of-experts (MoE… view at source ↗
Figure 2
Figure 2. The overall framework of REEF. • We propose a novel framework for graph foundation models that leverages relation tokens as basic units of graph data, enabling effective pre-training and transfer across different datasets. • We design two hypernetworks to dynamically generate relation-specific aggregators and classifiers based on relational representations. Additionally, we introduce dataset-specific feature project… view at source ↗
Figure 3
Figure 3. Performance comparison across different dataset scales in the citation domain. To further explore the performance of REEF under different dataset scales, we focus on the citation domain and adopt six different pre-training dataset settings: (1) Pubmed; (2) Citeseer; (3) Pubmed and Citeseer; (4) Pubmed, Citeseer, Wisconsin, and Texas; (5) Pubmed, Citeseer, Wisconsin, Texas, and Photo; (6) Pubmed, Citeseer, Wisconsin,… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: T-SNE visualization of representations after pretraining. [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: Performance of different datasets with different hidden dimensions [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Test accuracy of different datasets during pre-training. [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Beyond Feature and Structure Alignment: Learning Transferable Propagation Knowledge for Graph Foundation Models

    cs.LG 2026-07 conditional novelty 6.0 of 10

    ProGFM transfers graph knowledge across domains by learning a prototype bank of per-edge, per-dimension propagation strengths and using them to modulate message passing on unseen graphs.

Reference graph

Works this paper leans on

64 extracted references · 35 canonical work pages · cited by 1 Pith paper

  1. [1]

    Achiam, S

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023

    arXiv 2023

  2. [2]

    Brown, B

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020

    1901

  3. [3]

    W. Cai, J. Jiang, F. Wang, J. Tang, S. Kim, and J. Huang. A survey on mixture of experts.arXiv preprint arXiv:2407.06204, 2024

    arXiv 2024

  4. [4]

    V . K. Chauhan, J. Zhou, P. Lu, S. Molaei, and D. A. Clifton. A brief review of hypernetworks in deep learning. Artificial Intelligence Review, 57(9):250, 2024

    work page 2024

  5. [5]

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

    arXiv 2024

  6. [6]

    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. arXiv preprint arXiv:2111.00064, 2021

    arXiv 2021

  7. [7]

    K. Dong, H. Mao, Z. Guo, and N. V . Chawla. Universal link predictor by in-context learning. arXiv preprint arXiv:2402.07738, 2024

    arXiv 2024

  8. [8]

    T. Fang, Y . Zhang, Y . Yang, C. Wang, and L. Chen. Universal prompt tuning for graph neural networks. Advances in Neural Information Processing Systems, 36, 2024

    work page 2024

Show all 64 references
  1. [9]

    Gao, T.-H

    Y . Gao, T.-H. Huang, and R. J. Passonneau. Abcd: A graph framework to convert complex sentences to a covering set of simple sentences. arXiv preprint arXiv:2106.12027, 2021

    2021 arXiv

  2. [10]

    C. Gong, X. Li, J. Yu, Y . Cheng, J. Tan, and C. Yu. Self-pro: A self-prompt and tuning framework for graph neural networks. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases, pages 197–215. Springer, 2024

    2024

  3. [11]

    D. Ha, A. Dai, and Q. V . Le. Hypernetworks. arXiv preprint arXiv:1609.09106, 2016

    2016 arXiv

  4. [12]

    Z. Hou, X. Liu, Y . Cen, Y . Dong, H. Yang, C. Wang, and J. Tang. Graphmae: Self-supervised masked graph autoencoders. In Proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining, pages 594–604, 2022

    2022

  5. [13]

    Huang, H

    Q. Huang, H. Ren, P. Chen, G. Kržmanc, D. Zeng, P. S. Liang, and J. Leskovec. Prodigy: Enabling in-context learning over graphs. Advances in Neural Information Processing Systems, 36, 2024

    2024

  6. [14]

    T. N. Kipf and M. Welling. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907, 2016

    2016 arXiv

  7. [15]

    H. Liu, J. Feng, L. Kong, N. Liang, D. Tao, Y . Chen, and M. Zhang. One for all: Towards training one graph model for all classification tasks. arXiv preprint arXiv:2310.00149, 2023

    2023 arXiv

  8. [16]

    J. Liu, H. Mao, Z. Chen, W. Fan, M. Ju, T. Zhao, N. Shah, and J. Tang. One model for one graph: A new perspective for pretraining with cross-domain graphs. arXiv preprint arXiv:2412.00315, 2024

    2024 arXiv

  9. [17]

    J. Liu, F. Xia, X. Feng, J. Ren, and H. Liu. Deep graph learning for anomalous citation detection. IEEE Transactions on Neural Networks and Learning Systems, 33(6):2543–2557, 2022

    2022

  10. [18]

    J. Liu, C. Yang, Z. Lu, J. Chen, Y . Li, M. Zhang, T. Bai, Y . Fang, L. Sun, P. S. Yu, et al. Towards graph foundation models: A survey and beyond. arXiv preprint arXiv:2310.11829, 2023

    2023 arXiv

  11. [19]

    Z. Liu, X. Yu, Y . Fang, and X. Zhang. Graphprompt: Unifying pre-training and downstream tasks for graph neural networks. In Proceedings of the ACM Web Conference 2023 , pages 417–428, 2023. 10

    2023

  12. [20]

    H. Mao, Z. Chen, W. Tang, J. Zhao, Y . Ma, T. Zhao, N. Shah, M. Galkin, and J. Tang. Position: Graph foundation models are already here. In Forty-first International Conference on Machine Learning

  13. [21]

    S. A. Myers, A. Sharma, P. Gupta, and J. Lin. Information network or social network? the structure of the twitter follow graph. In Proceedings of the 23rd international conference on world wide web, pages 493–498, 2014

    2014

  14. [22]

    H. Pei, B. Wei, K. C.-C. Chang, Y . Lei, and B. Yang. Geom-gcn: Geometric graph convolutional networks. arXiv preprint arXiv:2002.05287, 2020

    2002 arXiv

  15. [23]

    N. Reimers. Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv preprint arXiv:1908.10084, 2019

    1908 arXiv

  16. [24]

    Schlichtkrull, T

    M. Schlichtkrull, T. N. Kipf, P. Bloem, R. Van Den Berg, I. Titov, and M. Welling. Modeling relational data with graph convolutional networks. In The semantic web: 15th international conference, ESWC 2018, Heraklion, Crete, Greece, June 3–7, 2018, proceedings 15 , pages 593–60...

    2018

  17. [25]

    P. Sen, G. Namata, M. Bilgic, L. Getoor, B. Galligher, and T. Eliassi-Rad. Collective classifica- tion in network data. AI magazine, 29(3):93–93, 2008

    2008

  18. [26]

    Shchur, M

    O. Shchur, M. Mumme, A. Bojchevski, and S. Günnemann. Pitfalls of graph neural network evaluation. arXiv preprint arXiv:1811.05868, 2018

    2018 arXiv

  19. [27]

    X. Sun, H. Cheng, J. Li, B. Liu, and J. Guan. All in one: Multi-task prompting for graph neural networks. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 2120–2131, 2023

    2023

  20. [28]

    Y . Tan, H. Lv, X. Huang, J. Zhang, S. Wang, and C. Yang. Musegraph: Graph-oriented instruc- tion tuning of large language models for generic graph mining.arXiv preprint arXiv:2403.04780, 2024

    2024 arXiv

  21. [29]

    Thakoor, C

    S. Thakoor, C. Tallec, M. G. Azar, M. Azabou, E. L. Dyer, R. Munos, P. Veli ˇckovi´c, and M. Valko. Large-scale representation learning on graphs via bootstrapping. arXiv preprint arXiv:2102.06514, 2021

    2021 arXiv

  22. [30]

    Toutanova and D

    K. Toutanova and D. Chen. Observed versus latent features for knowledge base and text inference. In Proceedings of the 3rd workshop on continuous vector space models and their compositionality, pages 57–66, 2015

    2015

  23. [31]

    Touvron, T

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozière, N. Goyal, E. Hambro, F. Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023

    2023 arXiv

  24. [32]

    Veliˇckovi´c, G

    P. Veliˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Lio, and Y . Bengio. Graph attention networks. arXiv preprint arXiv:1710.10903, 2017

    2017 arXiv

  25. [33]

    Veliˇckovi´c, W

    P. Veliˇckovi´c, W. Fedus, W. L. Hamilton, P. Liò, Y . Bengio, and R. D. Hjelm. Deep graph infomax. arXiv preprint arXiv:1809.10341, 2018

    2018 arXiv

  26. [34]

    D. Wang, Y . Zuo, F. Li, and J. Wu. Llms as zero-shot graph learners: Alignment of gnn representations with llm token embeddings. arXiv preprint arXiv:2408.14512, 2024

    2024 arXiv

  27. [35]

    Z. Wang, Z. Zhang, N. Chawla, C. Zhang, and Y . Ye. Gft: Graph foundation model with transferable tree vocabulary. Advances in Neural Information Processing Systems, 37:107403– 107443, 2024

    2024

  28. [36]

    J. Wei, Y . Tay, R. Bommasani, C. Raffel, B. Zoph, S. Borgeaud, D. Yogatama, M. Bosma, D. Zhou, D. Metzler, et al. Emergent abilities of large language models. arXiv preprint arXiv:2206.07682, 2022

    2022 arXiv

  29. [37]

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and S. Y . Philip. A comprehensive survey on graph neural networks. IEEE transactions on neural networks and learning systems , 32(1):4–24, 2020. 11

    2020

  30. [38]

    J. Xia, L. Wu, J. Chen, B. Hu, and S. Z. Li. Simgrace: A simple framework for graph contrastive learning without data augmentation. In Proceedings of the ACM Web Conference 2022, pages 1070–1079, 2022

    2022

  31. [39]

    Xia and C

    L. Xia and C. Huang. Anygraph: Graph foundation model in the wild. 2024

    2024

  32. [40]

    L. Xia, B. Kao, and C. Huang. Opengraph: Towards open graph foundation models. arXiv preprint arXiv:2403.01121, 2024

    2024 arXiv

  33. [41]

    K. Xu, W. Hu, J. Leskovec, and S. Jegelka. How powerful are graph neural networks? arXiv preprint arXiv:1810.00826, 2018

    2018 arXiv

  34. [42]

    Z. Yang, W. Cohen, and R. Salakhudinov. Revisiting semi-supervised learning with graph embeddings. In International conference on machine learning, pages 40–48. PMLR, 2016

    2016

  35. [43]

    Y . You, T. Chen, Y . Sui, T. Chen, Z. Wang, and Y . Shen. Graph contrastive learning with augmentations. Advances in neural information processing systems, 33:5812–5823, 2020

    2020

  36. [44]

    X. Yu, C. Zhou, Y . Fang, and X. Zhang. Text-free multi-domain graph pre-training: Toward graph foundation models. arXiv preprint arXiv:2405.13934, 2024

    2024 arXiv

  37. [45]

    H. Zhao, A. Chen, X. Sun, H. Cheng, and J. Li. All in one and one for all: A simple yet effective method towards cross-domain graph pretraining. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 4443–4454, 2024

    2024

  38. [46]

    J. Zhao, H. Mostafa, M. Galkin, M. Bronstein, Z. Zhu, and J. Tang. Graphany: A foundation model for node classification on any graph. arXiv preprint arXiv:2405.20445, 2024

    2024 arXiv

  39. [47]

    J. Zhu, Z. Ding, J. Yu, J. Tan, X. Li, and W. Qian. Relief: Reinforcement learning empowered graph feature prompt tuning. arXiv preprint arXiv:2408.03195, 2024

    2024 arXiv

  40. [48]

    Pre.” refers to pretraining, “Trans

    Y . Zhu, H. Shi, X. Wang, Y . Liu, Y . Wang, B. Peng, C. Hong, and S. Tang. Graphclip: Enhancing transferability in graph foundation models for text-attributed graphs. arXiv preprint arXiv:2410.10329, 2024. 12 A Datasets Table 3 summarizes the details. Here is a detailed descr...

    2024 arXiv

  41. [49]

    Guidelines: • The answer NA means that the abstract and introduction do not include the claims made in the paper

    Claims Question: Do the main claims made in the abstract and introduction accurately reflect the paper’s contributions and scope? Answer: [Yes] Justification: This paper introduces a relation-based graph foudation model, with the main contributions and scope outlined in the ab...

  42. [50]

    Limitations

    Limitations Question: Does the paper discuss the limitations of the work performed by the authors? 19 Answer: [Yes] Justification: We analyze the limitations in Section 6. Guidelines: • The answer NA means that the paper has no limitation while the answer No means that the pap...

  43. [51]

    Guidelines: • The answer NA means that the paper does not include theoretical results

    Theory assumptions and proofs Question: For each theoretical result, does the paper provide the full set of assumptions and a complete (and correct) proof? Answer: [NA] Justification: This paper does not include theoretical results. Guidelines: • The answer NA means that the p...

  44. [52]

    The results in our paper are reproducible

    Experimental result reproducibility Question: Does the paper fully disclose all the information needed to reproduce the main ex- perimental results of the paper to the extent that it affects the main claims and/or conclusions of the paper (regardless of whether the code and da...

  45. [53]

    Guidelines: • The answer NA means that paper does not include experiments requiring code

    Open access to data and code Question: Does the paper provide open access to the data and code, with sufficient instruc- tions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes] Justification: We provide our code, as wel...

  46. [54]

    Guidelines: • The answer NA means that the paper does not include experiments

    Experimental setting/details Question: Does the paper specify all the training and test details (e.g., data splits, hyper- parameters, how they were chosen, type of optimizer, etc.) necessary to understand the results? Answer: [Yes] Justification: We provide experimental detai...

  47. [55]

    Guidelines: • The answer NA means that the paper does not include experiments

    Experiment statistical significance Question: Does the paper report error bars suitably and correctly defined or other appropriate information about the statistical significance of the experiments? Answer: [Yes] Justification: We report the standard deviation of the results. G...

  48. [56]

    Guidelines: • The answer NA means that the paper does not include experiments

    Experiments compute resources Question: For each experiment, does the paper provide sufficient information on the com- puter resources (type of compute workers, memory, time of execution) needed to reproduce the experiments? Answer: [Yes] Justification: This paper provide info...

  49. [57]

    Guidelines: • The answer NA means that the authors have not reviewed the NeurIPS Code of Ethics

    Code of ethics Question: Does the research conducted in the paper conform, in every respect, with the NeurIPS Code of Ethics https://neurips.cc/public/EthicsGuidelines? Answer: [Yes] Justification: The research conducted in the paper conform, in every respect, with the NeurIPS...

  50. [58]

    Guidelines: • The answer NA means that there is no societal impact of the work performed

    Broader impacts Question: Does the paper discuss both potential positive societal impacts and negative societal impacts of the work performed? Answer: [Yes] Justification: We discuss the broader impacts of this paper in Appendix G. Guidelines: • The answer NA means that there ...

  51. [59]

    Guidelines: • The answer NA means that the paper poses no such risks

    Safeguards Question: Does the paper describe safeguards that have been put in place for responsible release of data or models that have a high risk for misuse (e.g., pretrained language models, image generators, or scraped datasets)? Answer: [NA] 23 Justification: Our paper ha...

  52. [60]

    Guidelines: • The answer NA means that the paper does not use existing assets

    Licenses for existing assets Question: Are the creators or original owners of assets (e.g., code, data, models), used in the paper, properly credited and are the license and terms of use explicitly mentioned and properly respected? Answer: [Yes] Justification: Yes, the paper p...

  53. [61]

    Guidelines: • The answer NA means that the paper does not release new assets

    New assets Question: Are new assets introduced in the paper well documented and is the documentation provided alongside the assets? Answer: [Yes] Justification: : New assets introduced in this paper are well documented. Guidelines: • The answer NA means that the paper does not...

  54. [62]

    Guidelines: • The answer NA means that the paper does not involve crowdsourcing nor research with human subjects

    Crowdsourcing and research with human subjects 24 Question: For crowdsourcing experiments and research with human subjects, does the paper include the full text of instructions given to participants and screenshots, if applicable, as well as details about compensation (if any)...

  55. [63]

    Guidelines: • The answer NA means that the paper does not involve crowdsourcing nor research with human subjects

    Institutional review board (IRB) approvals or equivalent for research with human subjects Question: Does the paper describe potential risks incurred by study participants, whether such risks were disclosed to the subjects, and whether Institutional Review Board (IRB) approvals...

  56. [64]

    Answer: [NA] Justification: We only use LLMs to improve writing

    Declaration of LLM usage Question: Does the paper describe the usage of LLMs if it is an important, original, or non-standard component of the core methods in this research? Note that if the LLM is used only for writing, editing, or formatting purposes and does not impact the ...

    2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.