Pith. sign in

REVIEW 1 cited by

When LLMs meet open-world graph learning: a new perspective for unlabeled data uncertainty

T0 review · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read OGA combines prototype-based unknown-class rejection with LLM-generated, structure-guided annotations so text-attributed graphs can be retrained in open-world settings.

arxiv 2505.13989 v2 pith:WQ5W7LNS submitted 2025-05-20 cs.LG cs.AI

Yanzhe Wen , Xunkai Li , Qi Zhang , Zhu Lei , Guang Zeng , Rong-Hua Li , Guoren Wang This is my paper · ORCID
classification cs.LGcs.AI
keywords graphopen-worldunknown-classdatalabellearningllmsnodes
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 tackles a practical problem: a graph of items (papers, products, Wikipedia pages) where only a few items have labels, and some items belong to categories that have never been seen. The proposed system, OGA, works in two steps. First, a module called ALT embeds each node using a pretrained text-plus-graph model, builds a set of 'concepts' (average embeddings of each known label), and pushes unlabeled nodes toward the nearest concept. If no concept is close enough, the node is declared to belong to an unknown class. Second, a module called GLA groups the rejected nodes into communities using both graph links and text similarity, asks an LLM to name a few representative nodes in each community, then propagates those names to the rest of the community and merges similar names across communities. The final labels are added to the graph and a GNN is retrained on the enlarged label set. The authors report improved accuracy on known classes, higher coverage and precision for finding unknown classes, and better downstream accuracy after retraining compared to several existing open-world graph methods. The paper also includes four theorems intended to justify the design. However, the experimental section has internal inconsistencies: the sensitivity analysis recommends a sharpness value between 0.5 and 0.8 while the settings table reports 10; the loss-weight analysis recommends alpha=0.8 while the settings table reports 0.4; and one ablation table's numbers contradict its own text. The pretrained graph-language encoder that the whole method builds on is never named, and no code is released.
Extended reading notes

Core claim

The central claim is that OGA is the first LLM-enhanced open-world graph learning pipeline that integrates unknown-class rejection (UCR) with unknown-class annotation (UCA), and that it achieves state-of-the-art performance in four aspects: OGA improves known-class classification by 4.98% over the best baselines, improves unknown-class identification by 6.2% in coverage and 4.6% in precision, produces annotations with comparable semantic quality to ground truth, and improves downstream accuracy by 10.1% on average after retraining (Sec. 1, Tables 1-4).

Load-bearing premise

The method depends on an unnamed pretrained graph-language encoder (Eq. (1), Sec. 3.2) to produce embeddings that are already good enough for prototype-based rejection. The paper never identifies which encoder is used and never compares OGA against the same encoder used as a standalone classifier. If the encoder already separates known from unknown nodes, the gains attributed to ALT and GLA could be largely inherited from the pretrained model, making the central SOTA claim an artifact of the encoder choice rather than the proposed pipeline.

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.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method's central pipeline rests on a small number of hand-picked hyperparameters and on an unnamed pretrained graph-language encoder. The reported hyperparameter values are internally inconsistent across the appendix, and the theoretical guarantees rely on unverified regularity assumptions. No new physical or formal entities are introduced; the ontology framing reuses class prototypes as concepts.

free parameters (6)
  • lambda (sharpness) = Reported as 10 in A.11; sensitivity analysis suggests optimal 0.5-0.8 (A.14.1)
    Controls the sharpness of the prototype softmax in Eq. (3); tuned per dataset; reported value conflicts with the sensitivity analysis.
  • epsilon (rejection threshold) = 0.6 (A.11)
    Confidence threshold below which a node is rejected as unknown in Eq. (3); chosen by hand and directly sets the precision/coverage trade-off.
  • alpha (smoothness loss weight) = 0.4 (A.11); A.14.3 reports optimal 0.8
    Weight on the topology-aware smoothness loss in Eq. (4); the two reported values are inconsistent.
  • beta (separation loss weight) = 0.6 (A.11)
    Weight on the separation loss in Eq. (4).
  • gamma (semantic-topology balance) = 0.6 (optimal in A.14.4)
    Balances semantic and structural terms in the community detection objective Eq. (5).
  • kappa (propagation intensity) = 0.2 (A.11)
    Fine-tuned intensity factor in Eq. (1) for entity embedding propagation.
assumptions (4)
  • domain assumption A pretrained graph-language encoder provides unbiased, high-quality embeddings for all nodes (Sec. 3.2).
    Eq. (1) builds all representations on this encoder, which is never named or ablated; the entire ALT module inherits its properties.
  • domain assumption Homophily: high-degree nodes can inherit annotations from their neighbors (Sec. 3.3, Eq. (6)).
    GLA's efficiency scheme assumes neighbor labels are reliable enough to propagate without LLM calls; this holds only if the graph is homophilic.
  • standard math Theorems 1-4 assume a Lipschitz-continuous encoder, compact manifold, bounded intra-class variance sigma^2, and bounded Dirichlet energy delta (Appendix A.6).
    These assumptions are stated in the theorem statements but not verified for the actual datasets or the unnamed encoder.
  • standard math In the proof of Theorem 3, |Ck|-1 is approximated by |Ck| for large class counts (Appendix A.6.3).
    This approximation is used to reach the reported bound, but it is not justified for small class counts and does not explain the appearance of lambda.

how reviews work

0 comments
Cite this review

Pith. "Pith review of When LLMs meet open-world graph learning: a new perspective for unlabeled data uncertainty." pith.science (2026). https://pith.science/paper/WQ5W7LNS

@misc{pith2026250513989,
  author       = {Pith},
  title        = {Pith review of: When LLMs meet open-world graph learning: a new perspective for unlabeled data uncertainty},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WQ5W7LNS}},
  note         = {Machine review of arXiv:2505.13989}
}
read the original abstract

Recently, large language models (LLMs) have significantly advanced text-attributed graph (TAG) learning. However, existing methods inadequately handle data uncertainty in open-world scenarios, especially concerning limited labeling and unknown-class nodes. Prior solutions typically rely on isolated semantic or structural approaches for unknown-class rejection, lacking effective annotation pipelines. To address these limitations, we propose Open-world Graph Assistant (OGA), an LLM-based framework that combines adaptive label traceability, which integrates semantics and topology for unknown-class rejection, and a graph label annotator to enable model updates using newly annotated nodes. Comprehensive experiments demonstrate OGA's effectiveness and practicality.

Figures

Figures reproduced from arXiv: 2505.13989 by the authors.

Figure 1
Figure 1. A comparative overview of our proposed open-world learning pipeline and the conventional one. Our proposed OGA integrates LLM to introduce a new paradigm for unlabeled data uncertainty. During our investigation, we found that node-level out-of-distribution (OOD) detection and conventional open-world graph learning align most closely with the context of our research problem. Additionally, we identify that related fie… view at source ↗
Figure 2
Figure 2. The overview of our proposed OGA. For the first time, we introduce this concept into UCR, leveraging the flexible concepts, entities, and relations in ontology to enable adaptive open-label domain representation learning as follows: ◆ What is Ontology Represented in UCR? Ontology is a formal and explicit representation of concepts, entities, and their interrelations within a domain. In the context of UCR: ① The doma… view at source ↗
Figure 4
Figure 4. GLA sensitivity analysis (data). GLA Part (Data Perspective). In this section, we analyze the impact of the proportion of unlabeled known-class nodes on annotation, as illustrated in [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5 [PITH_FULL_IMAGE:figures/full_fig_p030_5.png]
Figure 6
Figure 6. Figure 6: Test accuracy convergence curves of GCN, GAT, and GraphSAGE on the Pubmed dataset under three conditions: lower bound, upper bound, and OGA-enhanced graph. The curves represent the model’s performance across epochs. effects introduced by the topology-aware smoothness l…
Figure 7
Figure 7. Figure 7: Effect of λ on key performance metrics. As λ increases, classification accuracy on known classes improves steadily until it reaches a saturation point, while the rejection rate also increases, indicating stronger separation in the ontology space. However, when λ become…
Figure 9
Figure 9. Figure 9: Unknown-Class Rejection and Accuracy on the WikiCS Dataset. In terms of OOD accuracy, the model exhibits a near-linear improvement up to a ratio of 0.6, reaching 93.56%, and achieves perfect separation (100%) when the unknown ratio exceeds 0.7. These results highlight …
Figure 10
Figure 10. Figure 10: Heatmap showing the impact of α and β on Accuracy, UCR, and Unknown Accuracy. The optimal performance is observed at α = 0.8 and β = 0.6, achieving the best balance between known-class accuracy and unknown-class rejection. In this section, we analyze the impact of the…
Figure 11
Figure 11. Figure 11: Effect of semantic-topology balance γ on modularity, semantic consistency, and LLM calls (Citeseer). A clear optimum is observed at γ = 0.6. A.15 Efficiency analysis A.15.1 Time Complexity of ALT Among the twelve baseline methods included in our evaluation, we selecti…

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. FedOGL: Combating Catastrophic Forgetting in Federated Open-World Multimodal Graph Learning

    cs.LG 2026-07 conditional novelty 5.0 of 10

    FedOGL blends replay, distillation, structure-basis gradient shielding, and prototype consolidation to cut catastrophic forgetting in federated open-world multimodal graph learning, reporting 42.67% less forgetting th...

Reference graph

Works this paper leans on

65 extracted references · 59 canonical work pages · cited by 1 Pith paper

  1. [1]

    Graph out-of-distribution detection goes neighborhood shaping

    Tianyi Bao, Qitian Wu, Zetian Jiang, Yiting Chen, Jiawei Sun, and Junchi Yan. Graph out-of-distribution detection goes neighborhood shaping. In International Conference on Machine Learning, ICML, 2024

    work page 2024

  2. [2]

    Motif prediction with graph neural networks

    Maciej Besta, Raphael Grob, Cesare Miglioli, Nicola Bernold, Grzegorz Kwasniewski, Gabriel Gjini, Raghavendra Kanakagiri, Saleh Ashkboos, Lukas Gianinazzi, Nikoli Dryden, and Torsten Hoefler. Motif prediction with graph neural networks. In Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD, 2022

    work page 2022

  3. [3]

    Decoupled graph energy- based model for node out-of-distribution detection on heterophilic graphs

    Yuhan Chen, Yihong Luo, Yifan Song, Pengwen Dai, Jing Tang, and Xiaochun Cao. Decoupled graph energy- based model for node out-of-distribution detection on heterophilic graphs. International Conference on Learning Representations, ICLR, 2025

    work page 2025

  4. [4]

    Label-free node classification on graphs with large language models (llms)

    Zhikai Chen, Haitao Mao, Hongzhi Wen, Haoyu Han, Wei Jin, Haiyang Zhang, Hui Liu, and Jiliang Tang. Label-free node classification on graphs with large language models (llms). International Conference on Learning Representations, ICLR, 2023

    work page 2023

  5. [5]

    Dslr: Diversity enhancement and structure learning for rehearsal-based graph continual learning

    Seungyoon Choi, Wonjoong Kim, Sungwon Kim, Yeonjun In, Sein Kim, and Chanyoung Park. Dslr: Diversity enhancement and structure learning for rehearsal-based graph continual learning. In Proceedings of the ACM Web Conference, WWW, 2024

    work page 2024

  6. [6]

    Spreading out-of-distribution detection on graphs

    Jongin Lim Daeho Um, Yuneil Yeo Sunoh Kim, and Yoonho Jung. Spreading out-of-distribution detection on graphs. ICLR 2025, 2023

    work page 2025

  7. [7]

    Continual learning of knowledge graph embeddings

    Angel Daruna, Mehul Gupta, Mohan Sridharan, and Sonia Chernova. Continual learning of knowledge graph embeddings. IEEE Robotics and Automation Letters, 6(2):1128–1135, 2021

    work page 2021

  8. [8]

    Lifelong learning of graph neural networks for open-world node classification

    Lukas Galke, Benedikt Franke, Tobias Zielke, and Ansgar Scherp. Lifelong learning of graph neural networks for open-world node classification. In International Joint Conference on Neural Networks, IJCNN, 2021

    work page 2021

Show all 65 references
  1. [9]

    An energy-centric framework for category-free out-of-distribution node detection in graphs

    Zheng Gong and Ying Sun. An energy-centric framework for category-free out-of-distribution node detection in graphs. In Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD, 2024

    2024

  2. [10]

    Hamilton, Rex Ying, and Jure Leskovec

    William L. Hamilton, Rex Ying, and Jure Leskovec. Inductive representation learning on large graphs. Curran Associates Inc., 2017

    2017

  3. [11]

    Universal graph continual learning

    Thanh Duc Hoang, Do Viet Tung, Duy-Hung Nguyen, Bao-Sinh Nguyen, Huy Hoang Nguyen, and Hung Le. Universal graph continual learning. Transactions on Machine Learning Research, 2023

    2023

  4. [12]

    Open-world lifelong graph learning

    Marcel Hoffmann, Lukas Galke, and Ansgar Scherp. Open-world lifelong graph learning. In International Joint Conference on Neural Networks, IJCNN, 2023

    2023

  5. [13]

    Open-world lifelong graph learning

    Marcel Hoffmann, Lukas Galke, and Ansgar Scherp. Open-world lifelong graph learning. In 2023 International Joint Conference on Neural Networks (IJCNN), pages 1–9. IEEE, 2023

    2023

  6. [14]

    Beyond the known: Novel class discovery for open-world graph learning

    Yucheng Jin, Yun Xiong, Juncheng Fang, Xixi Wu, Dongxiao He, Xing Jia, Bingchen Zhao, and Philip S Yu. Beyond the known: Novel class discovery for open-world graph learning. InInternational Conference on Database Systems for Advanced Applications, DASFAA, 2024

    2024

  7. [15]

    Beyond the known: Novel class discovery for open-world graph learning

    Yucheng Jin, Yun Xiong, Juncheng Fang, Xixi Wu, Dongxiao He, Xing Jia, Bingchen Zhao, and Philip S Yu. Beyond the known: Novel class discovery for open-world graph learning. InInternational Conference on Database Systems for Advanced Applications, pages 117–133. Springer, 2024

    2024

  8. [16]

    Semi-supervised classification with graph convolutional networks

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

    2016 arXiv

  9. [17]

    Semi-supervised classification with graph convolutional networks

    Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolutional networks. In International Conference on Learning Representations, ICLR, 2017

    2017

  10. [18]

    Gofa: A generative one-for-all model for joint graph language modeling

    Lecheng Kong, Jiarui Feng, Hao Liu, Chengsong Huang, Jiaxin Huang, Yixin Chen, and Muhan Zhang. Gofa: A generative one-for-all model for joint graph language modeling. International Conference on Learning Representations, ICLR, 2025

    2025

  11. [19]

    Disentangle-based continual graph representation learning

    Xiaoyu Kou, Yankai Lin, Shaobo Liu, Peng Li, Jie Zhou, and Yan Zhang. Disentangle-based continual graph representation learning. In The Conference on Empirical Methods in Natural Language Processing, EMNLP, 2020

    2020

  12. [20]

    Gated attention with asymmetric regularization for transformer- based continual graph learning

    Hongxiang Lin, Ruiqi Jia, and Xiaoqing Lyu. Gated attention with asymmetric regularization for transformer- based continual graph learning. In Proceedings of the International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR, 2023. 10 arXiv Temp...

    2023

  13. [21]

    Open-world semi- supervised novel class discovery

    Jiaming Liu, Yangqiming Wang, Tongze Zhang, Yulu Fan, Qinli Yang, and Junming Shao. Open-world semi- supervised novel class discovery. arXiv preprint arXiv:2305.13095, 2023

    2023 arXiv

  14. [22]

    Good-d: On unsupervised graph out-of-distribution detection

    Yixin Liu, Kaize Ding, Huan Liu, and Shirui Pan. Good-d: On unsupervised graph out-of-distribution detection. In ACM International Conference on Web Search and Data Mining, WSDM, 2023

    2023

  15. [23]

    Good-d: On unsupervised graph out-of-distribution detection

    Yixin Liu, Kaize Ding, Huan Liu, and Shirui Pan. Good-d: On unsupervised graph out-of-distribution detection. In Proceedings of the sixteenth ACM international conference on web search and data mining, pages 339–347, 2023

    2023

  16. [24]

    Arc: A generalist graph anomaly detector with in-context learning

    Yixin Liu, Shiyuan Li, Yu Zheng, Qingfeng Chen, Chengqi Zhang, and Shirui Pan. Arc: A generalist graph anomaly detector with in-context learning. 2024

    2024

  17. [25]

    Revisiting score propagation in graph out-of- distribution detection

    Longfei Ma, Yiyou Sun, Kaize Ding, Zemin Liu, and Fei Wu. Revisiting score propagation in graph out-of- distribution detection. In Advances in Neural Information Processing Systems, NeurIPS, 2024

    2024

  18. [26]

    Entropic out- of-distribution detection

    David Macêdo, Tsang Ing Ren, Cleber Zanchettin, Adriano LI Oliveira, and Teresa Ludermir. Entropic out- of-distribution detection. In 2021 international joint conference on neural networks (IJCNN), pages 1–8. IEEE, 2021

    2021

  19. [27]

    Graph continual learning with debiased lossless memory replay

    Chaoxi Niu, Guansong Pang, and Ling Chen. Graph continual learning with debiased lossless memory replay. European Conference on Artificial Intelligence, ECAI, 2024

    2024

  20. [28]

    Ftf-er: Feature-topology fusion-based experience replay method for continual graph learning

    Jinhui Pang, Changqing Lin, Xiaoshuai Hao, Rong Yin, Zixuan Wang, Zhihui Zhang, Jinglin He, and Huang Tai Sheng. Ftf-er: Feature-topology fusion-based experience replay method for continual graph learning. In Proceedings of the ACM International Conference on Multimedia, MM, 2024

    2024

  21. [29]

    Contrastive augmented graph2graph memory interaction for few shot continual learning

    Biqing Qi, Junqi Gao, Xinquan Chen, Dong Li, Jianxing Liu, Ligang Wu, and Bowen Zhou. Contrastive augmented graph2graph memory interaction for few shot continual learning. IEEE Transactions on Circuits and Systems for Video Technology, 2025

    2025

  22. [30]

    Reinforced continual learning for graphs

    Appan Rakaraddi, Lam Siew Kei, Mahardhika Pratama, and Marcus De Carvalho. Reinforced continual learning for graphs. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management, CIKM, 2022

    2022

  23. [31]

    Learning on graphs with out-of-distribution nodes

    Yu Song and Donglin Wang. Learning on graphs with out-of-distribution nodes. In Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD, 2022

    2022

  24. [32]

    Learning on graphs with out-of-distribution nodes

    Yu Song and Donglin Wang. Learning on graphs with out-of-distribution nodes. page 1635–1645, August 2022

    2022

  25. [33]

    Graph posterior network: Bayesian predictive uncertainty for node classification

    Maximilian Stadler, Bertrand Charpentier, Simon Geisler, Daniel Zügner, and Stephan Günnemann. Graph posterior network: Bayesian predictive uncertainty for node classification. Advances in Neural Information Processing Systems, NeurIPS, 2021

    2021

  26. [34]

    Graph-based continual learning

    Binh Tang and David S Matteson. Graph-based continual learning. International Conference on Learning Representations, ICLR, 2020

    2020

  27. [35]

    Spreading out-of-distribution detection on graphs

    Daeho Um, Jongin Lim, Sunoh Kim, Yuneil Yeo, and Yoonho Jung. Spreading out-of-distribution detection on graphs. In International Conference on Learning Representations, ICLR, 2025

    2025

  28. [36]

    Graph attention networks

    Petar Veliˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. Graph attention networks. arXiv preprint arXiv:1710.10903, 2017

    2017 arXiv

  29. [37]

    Smug: Sand mixing for unobserved class detection in graph few-shot learning

    Chenxu Wang, Xichan Nie, Jinfeng Chen, Pinghui Wang, Junzhou Zhao, and Xiaohong Guan. Smug: Sand mixing for unobserved class detection in graph few-shot learning. In Proceedings of the ACM Web Conference, WWW, 2024

    2024

  30. [38]

    Gold: Graph out-of-distribution detection via implicit adversarial latent generation

    Danny Wang, Ruihong Qiu, Guangdong Bai, and Zi Huang. Gold: Graph out-of-distribution detection via implicit adversarial latent generation. International Conference on Learning Representations, ICLR, 2025

    2025

  31. [39]

    Streaming graph neural networks via continual learning

    Junshan Wang, Guojie Song, Yi Wu, and Liang Wang. Streaming graph neural networks via continual learning. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management, CIKM, 2020

    2020

  32. [40]

    Open-world semi-supervised learning for node classification

    Yanling Wang, Jing Zhang, Lingxi Zhang, Lixin Liu, Yuxiao Dong, Cuiping Li, Hong Chen, and Hongzhi Yin. Open-world semi-supervised learning for node classification. In International Conference on Data Engineering, ICDE, 2024

    2024

  33. [41]

    Open-world semi-supervised learning for node classification

    Yanling Wang, Jing Zhang, Lingxi Zhang, Lixin Liu, Yuxiao Dong, Cuiping Li, Hong Chen, and Hongzhi Yin. Open-world semi-supervised learning for node classification. In 2024 IEEE 40th International Conference on Data Engineering (ICDE), pages 2723–2736. IEEE, 2024. 11 arXiv Tem...

    2024

  34. [42]

    Augmenting low-resource text classification with graph-grounded pre-training and prompting

    Zhihao Wen and Yuan Fang. Augmenting low-resource text classification with graph-grounded pre-training and prompting. In Proceedings of the International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR, 2023

    2023

  35. [43]

    Openwgl: Open-world graph learning

    Man Wu, Shirui Pan, and Xingquan Zhu. Openwgl: Open-world graph learning. InIEEE International Conference on Data Mining, ICDM, 2020

    2020

  36. [44]

    Openwgl: Open-world graph learning

    Man Wu, Shirui Pan, and Xingquan Zhu. Openwgl: Open-world graph learning. In 2020 IEEE international conference on data mining (icdm), pages 681–690. IEEE, 2020

    2020

  37. [45]

    Energy-based out-of-distribution detection for graph neural networks

    Qitian Wu, Yiting Chen, Chenxiao Yang, and Junchi Yan. Energy-based out-of-distribution detection for graph neural networks. International Conference on Learning Representations, ICLR, 2023

    2023

  38. [46]

    Energy-based out-of-distribution detection for graph neural networks

    Qitian Wu, Yiting Chen, Chenxiao Yang, and Junchi Yan. Energy-based out-of-distribution detection for graph neural networks. arXiv preprint arXiv:2302.02914, 2023

    2023 arXiv

  39. [47]

    Lego-learn: Label-efficient graph open-set learning

    Haoyan Xu, Kay Liu, Zhengtao Yao, Philip S Yu, Kaize Ding, and Yue Zhao. Lego-learn: Label-efficient graph open-set learning. arXiv preprint arXiv:2410.16386, 2024

    2024 arXiv

  40. [48]

    Graph synthetic out-of-distribution exposure with large language models

    Haoyan Xu, Zhengtao Yao, Ziyi Wang, Zhan Cheng, Xiyang Hu, Mengyuan Li, and Yue Zhao. Graph synthetic out-of-distribution exposure with large language models. arXiv preprint arXiv:2504.21198, 2025

    2025 arXiv

  41. [49]

    Glip-ood: Zero-shot graph ood detection with foundation model

    Haoyan Xu, Zhengtao Yao, Xuzhi Zhang, Ziyi Wang, Langzhou He, Yushun Dong, Philip S Yu, Mengyuan Li, and Yue Zhao. Glip-ood: Zero-shot graph ood detection with foundation model. arXiv preprint arXiv:2504.21186, 2025

    2025 arXiv

  42. [50]

    Open-world graph active learning for node classification

    Hui Xu, Liyao Xiang, Junjie Ou, Yuting Weng, Xinbing Wang, and Chenghu Zhou. Open-world graph active learning for node classification. ACM Transactions on Knowledge Discovery from Data, TKDD, 18(2):1–20, 2023

    2023

  43. [51]

    How powerful are graph neural networks? 2019

    Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. How powerful are graph neural networks? 2019

    2019

  44. [52]

    Bounded and uniform energy-based out-of-distribution detection for graphs

    Shenzhi Yang, Bin Liang, An Liu, Lin Gui, Xingkai Yao, and Xiaofang Zhang. Bounded and uniform energy-based out-of-distribution detection for graphs. In International Conference on Machine Learning, ICML, 2024

    2024

  45. [53]

    Samgpt: Text-free graph foundation model for multi-domain pre-training and cross-domain adaptation

    Xingtong Yu, Zechuan Gong, Chang Zhou, Yuan Fang, and Hui Zhang. Samgpt: Text-free graph foundation model for multi-domain pre-training and cross-domain adaptation. Proceedings of the ACM Web Conference, WWW, 2025

    2025

  46. [54]

    Cost-effective label-free node classification with llms

    Taiyan Zhang, Renchi Yang, Mingyu Yan, Xiaochun Ye, Dongrui Fan, and Yurui Lai. Cost-effective label-free node classification with llms. arXiv preprint arXiv:2412.11983, 2024

    2024 arXiv

  47. [55]

    Hierarchical prototype networks for continual graph representation learning

    Xikun Zhang, Dongjin Song, and Dacheng Tao. Hierarchical prototype networks for continual graph representation learning. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(4):4622–4636, 2022

    2022

  48. [56]

    Uncertainty aware semi-supervised learning on graph data

    Xujiang Zhao, Feng Chen, Shu Hu, and Jin-Hee Cho. Uncertainty aware semi-supervised learning on graph data. Advances in Neural Information Processing Systems, NeurIPS, 2020

    2020

  49. [57]

    Overcoming catastrophic forgetting in graph neural networks with experience replay

    Fan Zhou and Chengtai Cao. Overcoming catastrophic forgetting in graph neural networks with experience replay. In Proceedings of the Association for the Advancement of Artificial Intelligence, AAAI, 2021. 12 arXiv Template A PREPRINT A Outline The appendix is organized as foll...

    2021

  50. [58]

    Community-Level Analysis: Focus on each community-level label and its associated content

  51. [59]

    Label Merging: Merge similar community-level labels based on their semantic similarity, mea- sured using cosine distance between their embedding vectors

  52. [60]

    Use this supplementary information, but prioritize semantic similarity when merging

    Use of Neighboring Information: Neighboring community labels are provided to facilitate label fusion. Use this supplementary information, but prioritize semantic similarity when merging

  53. [61]

    Output Format: The output should be a comma-separated list of merged labels, each enclosed in parenthe- ses, in the same order as the input community-level labels

    Final Label Generation: The resulting label must be concise, meaningful, and representative of the combined communities. Output Format: The output should be a comma-separated list of merged labels, each enclosed in parenthe- ses, in the same order as the input community-level ...

  54. [62]

    Community-Level Analysis: Examine each community label along with its associated content

  55. [63]

    Label Merging: Merge labels that exhibit high semantic similarity, as determined by the cosine distance between their embedding vectors

  56. [64]

    However, prioritize semantic similarity over contextual proximity

    Utilization of Neighboring Information: Neighboring labels are provided to assist the merging process. However, prioritize semantic similarity over contextual proximity

  57. [65]

    Output Format: Return a comma-separated list of the merged labels, with each label enclosed in parenthe- ses, following the original order of the input

    Final Label Generation: Produce a concise and meaningful label that accurately represents the merged communities. Output Format: Return a comma-separated list of the merged labels, with each label enclosed in parenthe- ses, following the original order of the input. Example: (...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.