Pith. sign in

REVIEW 5 major objections 5 minor 61 references

S$^2$DN: Learning to Denoise Unconvincing Knowledge for Inductive Knowledge Graph Completion

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read This paper claims that inductive knowledge graph completion can be made more accurate by smoothing semantically similar relations into shared representations and by pruning unreliable edges from the local subgraph, and reports that S2DN…

desk verdict A sensible denoising method for inductive KGC, but the SOTA claim is not supported by the reported experiments; both modules ablate well, and the robustness study is useful. read the letter →

arxiv 2412.15822 v1 pith:G7CMEO54 submitted 2024-12-20 cs.LG cs.AIcs.CL

classification cs.LGcs.AIcs.CL
keywords inductiveknowledgegraphcompletionsemanticsmoothingstructurerefiningdenoisingsubgraphreasoningGumbel-softmaxedgereliabilitynoisygraphs
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

Inductive knowledge graph completion asks a model to predict missing facts about entities it has never seen during training. The paper claims that the two biggest obstacles in this setting are inconsistent representations of semantically similar relations and noisy edges that surround newly emerged entities, and that both can be fixed by a single network that denoises meaning and structure together. The proposed S2DN smooths relation embeddings so similar relations share one semantic representation, while a structure-refining module scores every edge in the local subgraph and drops the ones that look unreliable. On WN18RR, FB15k-237, and NELL-995, the paper reports that S2DN outperforms existing rule-based and GNN-based inductive KGC methods, and that it degrades less than baselines when false facts are deliberately injected. If true, this would mean that explicit denoising of relations and edges is a practical route to robust reasoning about unseen entities, not just a stronger encoder.

What carries the argument

The load-bearing mechanism is a pair of denoising modules applied to the enclosing subgraph of a candidate triple. The semantic smoothing module learns a categorical assignment of each observed relation to a smaller set of semantic clusters via Gumbel-softmax reparameterization, blurring relations like 'located in' and 'lie in' into one representation; an information-bottleneck style argument and a KL penalty justify keeping the smoothed embedding close to the original. The structure refining module computes an edge-reliability score as the sigmoid of the inner product of MLP-transformed node-feature vectors, models all edges as independent Bernoulli variables, relaxes them with the concrete distribution for differentiability, and removes edges whose probability falls below 0.5. The smoothed graph is encoded with a relational GNN and the refined graph with a GCN, and their pooled representations are concatenated and fed to an MLP classifier. This machinery is what lets the model discard unreliable interactions while preserving relation semantics for unseen entities.

What would settle it

Build a poisoned version of WN18RR V1 where the injected false edges deliberately connect nodes whose MLP-transformed features are similar (for example, by adding noise that targets nodes on the same shortest-path distance labels), and compare S2DN to its own no-structure-refining variant: if the full model fails to beat the variant, the feature-similarity reliability assumption is refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that 'unconvincing knowledge' in inductive knowledge graph completion has two independent sources, semantic inconsistency and structural noise, and that a network addressing both can beat the state of the art. Semantic smoothing maps relations with similar meanings into a shared embedding using a Gumbel-softmax over the relation set, with a KL term keeping the smoothed representation from drifting too far from the original. Structure refining estimates a reliability probability for each edge in the enclosing subgraph from the inner product of MLP-transformed node features, treats edges as independent Bernoulli variables, and prunes edges with reliability below 0.5 before applying a GCN. The smoothed subgraph is encoded by a relational GNN and the refined subgraph by a GCN, and the two global readouts are concatenated and classified. The paper reports this design outperforms the compared baselines on the three benchmark datasets and that ablations removing either module cause clear drops in Hits@1, Hits@10, and MRR.

Load-bearing premise

The model assumes that how reliable an edge is can be judged from the similarity of learned node-feature vectors, and that edges are independent, so a false fact connecting two feature-similar nodes will survive pruning and can mislead the prediction.

Editorial extensions

If this is right

  • The performance gains reported over rule-based and GNN baselines imply that inductive KGC systems do not need cleaner graphs; a network that smooths relations and prunes edges before reasoning can recover much of the lost accuracy.
  • The contamination experiments imply that filtering edges by learned node-feature similarity is an effective defense against randomly injected false facts, both semantic replacements and randomly sampled triples.
  • The ablation results imply that neither module alone is sufficient: removing semantic smoothing hurts most on semantically noisy graphs, while removing structure refining hurts most on structurally noisy graphs.
  • Because relation smoothing is learned from context rather than fixed vocabularies, the method is designed to transfer to new relation sets, which is exactly what inductive completion requires.

Reading between the lines

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

  • A testable extension the authors do not run: apply the same two modules to transductive KGC benchmarks, where entities are seen during training; if smoothing relations still helps there, the benefit is not specific to inductive generalization but to noise generally.
  • An adversarial test the paper does not run: insert false edges specifically between nodes with similar learned features, since the structure refiner assumes feature-similar connected nodes are trustworthy; correlated noise of this kind may defeat the pruning.
  • The smoothed relation assignment learned by the network could be read as an emergent relation ontology; comparing it against human-curated relation clusters on FB15k-237 would be a direct external check of the semantic-consistency claim.
  • For the biomedical application the paper names as future work, the natural next experiment is to poison a drug-interaction or protein-function KG with literature-derived false associations and check whether the same robustness pattern holds when relations are more numerous and heterogeneous.
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

5 major / 5 minor

Summary. The paper proposes S2DN, a subgraph-based inductive knowledge graph completion model that combines two modules: a semantic smoothing module that blurs relation embeddings with similar semantics through a Gumbel-softmax categorization, and a structure refining module that estimates edge reliability as a Bernoulli probability from MLP-transformed node features and prunes low-weight edges. The smoothed subgraph is processed by an RGNN and the refined subgraph by a GCN; the two global readouts are concatenated and scored by a classifier. Experiments on WN18RR, FB15k-237, and NELL-995, each with four inductive splits, report Hits@1, Hits@10, and MRR, together with ablations and robustness tests under injected semantic and structural noise. The paper claims that S2DN surpasses state-of-the-art models.

Significance. If the reported gains are reproducible, the two modules are a plausible and generally applicable contribution to inductive KGC; the released code, the use of public benchmarks, the ablations of both modules, and the five-run repetition are strengths, and I did not find evidence of circular fitting to the test labels. However, the headline 'surpasses SOTA' claim is not supported by the paper's own tables, and the baseline comparison is incomplete and uses a protocol that prevents direct comparison with the strongest published results. The robustness study also covers only independently injected noise. Because the core method is sound and the issues are correctable by re-running comparisons and qualifying claims, the contribution could be publishable after substantial revision.

major comments (5)
  1. [Abstract; Conclusion; Tables 1 and 9] The claim that S2DN 'surpasses the performance of state-of-the-art models' is contradicted by the paper's own reported numbers. In Table 9 (NELL-995 V4), RMPI achieves Hits@10 84.20 and MRR 72.63 versus S2DN's 74.89 and 64.12; in Table 1 (WN18RR V4), RMPI's Hits@10 82.33 exceeds S2DN's 82.15, and in Table 9 (NELL-995 V3) RMPI's MRR 80.07 exceeds S2DN's 76.22. The abstract and conclusion should either restrict the claim to 'most settings' or support an aggregated claim with a significance test over all twelve splits.
  2. [Appendix B.1.1; Appendix B.2.3; Tables 1, 2, and 9] The comparison does not establish superiority relative to the strongest published inductive KGC models because of a protocol mismatch. Appendix B.1.1 states that ranking is done against 50 randomly sampled negative entities, and Appendix B.2.3 excludes RED-GNN and AdaProp solely because they rank against all entities; GraIL is discussed and used in the visualization but does not appear in the main result tables. Under the chosen protocol the numbers may be internally consistent, but they are not directly comparable to published SOTA numbers from these baselines. The authors should either rerun the omitted baselines under a common protocol or explicitly state in the abstract and conclusion that the comparison excludes these methods.
  3. [Appendix B; Tables 1, 2, and 9] The manuscript states in Appendix B that each experiment was repeated five times, but all reported metrics are single numbers without standard deviations or significance tests. Several decisive comparisons are close (for example, WN18RR V4 Hits@10 82.15 versus 82.33, and NELL-995 V1 Hits@1 45.00 versus RMPI's 47.50), so without variance information the reader cannot judge whether the gains are meaningful. The authors should report mean and standard deviation over the five runs and, where appropriate, paired significance tests.
  4. [Eqs. (5)-(6); 'Robustness of S2DN (RQ3)'; Tables 3, 4, and 8] The robustness claim is demonstrated only for independently injected noise. The structure refining module treats edges as mutually independent Bernoulli variables and scores reliability by the inner product of MLP-transformed node features, so correlated noise (for example, clusters of false facts among similar entities) is not covered by the experiments, which inject random relation replacements and random negative triples. If the robustness claim is meant to cover realistic contamination, the authors should either add experiments with correlated noise or restrict the claim to the tested noise model.
  5. [Theoretical Discussion of Smoothing; Eq. (12)] The information-bottleneck discussion is not tightly connected to the actual loss. Lemma 1 bounds I(\tilde{E}; E_n), but the objective in Eq. (12) replaces mutual-information terms with a KL divergence between smoothed and original relation embeddings and a cross-entropy term; the paper does not show that minimizing Eq. (12) is equivalent to minimizing the bound in Eq. (10). This section should be labeled as intuition or made rigorous before being presented as a theoretical justification.
minor comments (5)
  1. [Appendix C.4.3; Table 5/6; Section B.1.2] The hyperparameter analysis says that S2DN achieves optimal performance with k = 4 for NELL-995 V1 and later designates NELL-995 as 4-hop, whereas Table 5/6 and Section B.1.2 set k = 2 for NELL-995; this inconsistency should be resolved.
  2. [Appendix A/B; Tables 5 and 6] Tables 5 and 6 are duplicate 'hyperparameter details of S2DN' tables; the duplicate should be removed or renumbered.
  3. [Appendix B.2.1; Appendix B.2.2] The baseline TAGT is written as TACT in Appendix B.2.1 and B.2.2; please standardize the name throughout.
  4. [Section 'Robustness of S2DN (RQ3)'] The robustness section contains dangling references such as 'as shown in Section ,' and the semantic-consistency paragraph uses '%' placeholders without a target; the section numbers need to be filled in.
  5. [Abstract; Conclusion] The phrase 'surpasses the performance of state-of-the-art models' should be qualified even after additional baselines are added, because the paper's own tables show that RMPI wins on several individual splits.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: S2DN's modules are trained end-to-end on external benchmark labels, and its theoretical lemma is a standard data-processing inequality rather than a self-referential prediction.

full rationale

The central derivation chain is not circular. S2DN's semantic smoothing weights and structure-refining edge probabilities are trainable parameters optimized by the cross-entropy loss (Eq. 12) on labeled benchmark triples; they are not fitted to test metrics, and no reported Hits@10, Hits@1, or MRR value is an algebraic consequence of an input parameter. The information-bottleneck discussion (Lemma 1) is an application of the data-processing inequality to the Markov chain <(Y, E_n) -> E -> ~E>, adapted from Achille and Soatto 2018; it motivates the KL-plus-cross-entropy surrogate but is not itself used to generate the benchmark numbers, so a loose equivalence claim there is a rigor issue rather than circularity. Self-citations (Ma et al. 2018 image smoothing; Ma et al. 2022 and Lin et al. 2020 KG applications; Yan et al. 2022 as an excluded baseline) are contextual or related-work references, and none is load-bearing for the SOTA claim. The main empirical concerns—omission of RED-GNN, AdaProp, and GraIL from the main tables and two splits where RMPI beats S2DN—are comparability and evidence problems, not circular reductions; the paper itself discloses the evaluation-protocol reason in Appendix B.2.3. Accordingly, no circular step meets the quote-and-reduction standard, and the honest finding is no significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The method uses standard hyperparameters and model assumptions. The main ad hoc choices are the relation smoothing mechanism and the edge reliability scoring; neither is grounded in an external semantic resource or noise model.

free parameters (6)
  • lambda (regularization coefficient) = 0.5 (WN18RR, NELL-995), 0.1 (FB15k-237)
    Tuned by grid search; controls KL and L2 trade-off in Eq. (12).
  • tau (Gumbel softmax temperature) = not reported
    Temperature in Eq. (1); controls sharpness of relation smoothing and affects performance, but is omitted from the hyperparameter table.
  • pi (edge pruning threshold) = 0.5
    Edges with reliability below 0.5 are removed; tuned over {0.1, 0.3, 0.5, 0.7, 0.9} and fixed at 0.5 for all datasets.
  • k-hop subgraph size = 4 (WN18RR), 3 (FB15k-237), 2 or 4 (NELL-995, reported inconsistently)
    Dataset-specific, tuned in C.4.3; Table 6 says 2 for NELL-995 but C.4.3 says 4, an internal inconsistency.
  • learning rate = 0.1, 0.0005, 0.001 per dataset
    Tuned per dataset in C.4.1, with strong sensitivity on FB15k-237 and NELL-995.
  • batch size = 8 (WN18RR, NELL-995), 32 (FB15k-237)
    Tuned per dataset in C.4.2.
assumptions (5)
  • domain assumption The KG can be represented as a heterogeneous graph, and inductive splits keep the relation set shared while entity sets are disjoint.
    Problem definition and Appendix B.1.1; this is the standard inductive KGC setting.
  • domain assumption The enclosing subgraph around a target link contains enough evidence to predict the link.
    GraIL-style subgraph reasoning is used throughout the method and determines the input to both modules.
  • ad hoc to paper Relations with similar semantics can be identified by softmax over the inner product of relation embeddings with a learned weight matrix.
    Eq. (1); no external semantic resource is used, so semantic similarity is defined by the learned geometry, which risks circularity if the downstream task shapes it.
  • ad hoc to paper Edges in the refined graph are mutually independent Bernoulli variables with reliability probability from node-feature inner products.
    Eqs. (5) to (7); independence and feature-similarity scoring are strong modeling assumptions.
  • standard math Task-irrelevant noise En is independent of label Y, so H(Y|En) = H(Y), in the information-bottleneck lemma.
    Appendix A.1, used in the proof of Lemma 1; standard assumption but unverified for real KG noise.

how reviews work

0 comments
Cite this review

Pith. "Pith review of S$^2$DN: Learning to Denoise Unconvincing Knowledge for Inductive Knowledge Graph Completion." pith.science (2026). https://pith.science/paper/G7CMEO54

@misc{pith2026241215822,
  author       = {Pith},
  title        = {Pith review of: S$^2$DN: Learning to Denoise Unconvincing Knowledge for Inductive Knowledge Graph Completion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G7CMEO54}},
  note         = {Machine review of arXiv:2412.15822}
}
abstract

Inductive Knowledge Graph Completion (KGC) aims to infer missing facts between newly emerged entities within knowledge graphs (KGs), posing a significant challenge. While recent studies have shown promising results in inferring such entities through knowledge subgraph reasoning, they suffer from (i) the semantic inconsistencies of similar relations, and (ii) noisy interactions inherent in KGs due to the presence of unconvincing knowledge for emerging entities. To address these challenges, we propose a Semantic Structure-aware Denoising Network (S$^2$DN) for inductive KGC. Our goal is to learn adaptable general semantics and reliable structures to distill consistent semantic knowledge while preserving reliable interactions within KGs. Specifically, we introduce a semantic smoothing module over the enclosing subgraphs to retain the universal semantic knowledge of relations. We incorporate a structure refining module to filter out unreliable interactions and offer additional knowledge, retaining robust structure surrounding target links. Extensive experiments conducted on three benchmark KGs demonstrate that S$^2$DN surpasses the performance of state-of-the-art models. These results demonstrate the effectiveness of S$^2$DN in preserving semantic consistency and enhancing the robustness of filtering out unreliable interactions in contaminated KGs.

Figures

Figures reproduced from arXiv: 2412.15822 by the authors.

Figure 1
Figure 1. (a) S2DN outperforms GraIL in terms of Hits@10 on noisy KGs with different noise ratios (i.e., high robust￾ness). (b) The relation edited by shows a high percentage of being converted to other relations (enumerated on the x-axis) with similar semantics (i.e., high semantic consistency). 2023b). For example, LAN (Wang et al. 2019) learns the em￾beddings of unseen entities by aggregating information from neighboring n… view at source ↗
Figure 2
Figure 2. The S2DN framework comprises two modules for inductively predicting links in a given KG : (1) Smoothing relational semantics by blurring similar relations adaptively; (2) Refining the structure of subgraphs by learning reliable interactions dynamically. subgraphs are concatenated and fed into a classifier to pre￾dict the interaction probability of the target link (u, r, v). Semantic Smoothing. KGs often suffer from … view at source ↗
Figure 3
Figure 3. The architecture of Semantic Smoothing and Structure Refining modules of S2DN. by excessive smoothing of relations and contain further con￾sistencies, we incorporate a trade-off objective designed to preserve generic information during the optimization pro￾cess. After obtaining the smoothed relations, we refined the enclosing subgraph g by the new relations R˜. Then a L-layer RGNN (Schlichtkrull et al. 2018; Xu et a… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The transition ratio between the original and [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: The big red and green nodes represent the source [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: An example to show the impact of semantic incon [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: The big red and green nodes represent the source and target entities. The small nodes in red, orange, and blue are [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: The sensitivity of hyperparameters across all datasets (V1 version). The bar indicates the [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 55 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Achille, A.; and Soatto, S. 2018. Emergence of invariance and disentanglement in deep representations. Journal of Machine Learning Research, 19(50): 1--34

  4. [4]

    F.; and Saffari, A

    Baek, J.; Aji, A. F.; and Saffari, A. 2023. Knowledge-augmented language model prompting for zero-shot knowledge graph question answering. arXiv preprint arXiv:2306.04136

  5. [5]

    Bai, J.; Luo, C.; Li, Z.; Yin, Q.; Yin, B.; and Song, Y. 2023. Knowledge graph reasoning over entities and numerical values. In Proceedings of ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 57--68

  6. [6]

    Cao, Q.; Li, B.; Liang, X.; Wang, K.; and Lin, L. 2022. Knowledge-Routed Visual Question Reasoning: Challenges for Deep Representation Embedding. IEEE Transactions on Neural Networks and Learning Systems, 33(7): 2758--2767

  7. [7]

    Chen, J.; He, H.; Wu, F.; and Wang, J. 2021. Topology-aware correlations between relations for inductive link prediction in knowledge graphs. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, 6271--6278

  8. [8]

    K.; and Sun, Y

    Cheng, K.; Ahmed, N. K.; and Sun, Y. 2023. Neural compositional rule learning for knowledge graph reasoning. Proceedings of International Conference on Learning Representations

Show all 61 references
  1. [9]

    Cheng, K.; Liu, J.; Wang, W.; and Sun, Y. 2022. Rlogic: Recursive logical rule learning from knowledge graphs. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 179--189

  2. [10]

    Cover, T. M. 1999. Elements of information theory. John Wiley & Sons

  3. [11]

    Dettmers, T.; Minervini, P.; Stenetorp, P.; and Riedel, S. 2018. Convolutional 2d knowledge graph embeddings. In Proceedings of the AAAI conference on artificial intelligence, volume 32

  4. [12]

    Du, Z.; Zhou, C.; Yao, J.; Tu, T.; Cheng, L.; Yang, H.; Zhou, J.; and Tang, J. 2021. Cogkr: Cognitive graph for multi-hop knowledge reasoning. IEEE Transactions on Knowledge and Data Engineering

  5. [13]

    Fan, Z.; Xu, K.; Dong, Z.; Peng, H.; Zhang, J.; and Yu, P. S. 2023. Graph Collaborative Signals Denoising and Augmentation for Recommendation. In Proceedings of International ACM SIGIR Conference on Research and Development in Information Retrieval, 2037--2041

  6. [14]

    Gao, Y.; Du, Y.; Hu, Y.; Chen, L.; Zhu, X.; Fang, Z.; and Zheng, B. 2022. Self-guided learning to denoise for robust recommendation. In Proceedings of International ACM SIGIR Conference on Research and Development in Information Retrieval, 1412--1422

  7. [15]

    Z.; Chen, M.; Jiang, S.; Zhang, W.; and Chen, H

    Geng, Y.; Chen, J.; Pan, J. Z.; Chen, M.; Jiang, S.; Zhang, W.; and Chen, H. 2023. Relational message passing for fully inductive knowledge graph completion. In Proceedings of IEEE International Conference on Data Engineering (ICDE), 1221--1233. IEEE

  8. [16]

    Glorot, X.; and Bengio, Y. 2010. Understanding the difficulty of training deep feedforward neural networks. In Proceedings of International Conference on Artificial Intelligence and Statistics, 249--256. JMLR Workshop and Conference Proceedings

  9. [17]

    Guo, T.; Xu, C.; Shi, B.; Xu, C.; and Tao, D. 2019. Smooth deep image generator from noises. In Proceedings of the AAAI Conference on Artificial Intelligence, 01, 3731--3738

  10. [18]

    Hong, Y.; Bu, C.; and Wu, X. 2021. High-Quality Noise Detection for Knowledge Graph Embedding with Rule-Based Triple Confidence. In Pacific Rim International Conference on Artificial Intelligence, PRICAI 2021, 572–585

  11. [19]

    Jang, E.; Gu, S.; and Poole, B. 2017. Categorical reparameterization with gumbel-softmax. In Proceedings of International Conference on Learning Representations

  12. [20]

    Ji, S.; Pan, S.; Cambria, E.; Marttinen, P.; and Yu, P. S. 2022. A Survey on Knowledge Graphs: Representation, Acquisition, and Applications. IEEE Transactions on Neural Networks and Learning Systems, 33(2): 494--514

  13. [21]

    P.; and Ba, J

    Kingma, D. P.; and Ba, J. 2014. Adam: A method for stochastic optimization. In Proceedings of International Conference on Learning Representations

  14. [22]

    N.; and Welling, M

    Kipf, T. N.; and Welling, M. 2017. Semi-supervised classification with graph convolutional networks. In Proceedings of International Conference on Learning Representations, 1--13

  15. [23]

    Li, Y.; Yu, K.; Zhang, Y.; Liang, J.; and Wu, X. 2023. Adaptive Prototype Interaction Network for Few-Shot Knowledge Graph Completion. IEEE Transactions on Neural Networks and Learning Systems, 1--14

  16. [24]

    Li, Z.; Sun, X.; Luo, Y.; Zhu, Y.; Chen, D.; Luo, Y.; Zhou, X.; Liu, Q.; Wu, S.; Wang, L.; et al. 2024. GSLB: the graph structure learning benchmark. Advances in Neural Information Processing Systems, 36

  17. [25]

    Lin, Q.; Liu, J.; Xu, F.; Pan, Y.; Zhu, Y.; Zhang, L.; and Zhao, T. 2022. Incorporating context graph with logical reasoning for inductive relation prediction. In Proceedings of International ACM SIGIR Conference on Research and Development in Information Retrieval, 893--903

  18. [26]

    Lin, X.; Quan, Z.; Wang, Z.-J.; Ma, T.; and Zeng, X. 2020. KGNN: Knowledge Graph Neural Network for Drug-Drug Interaction Prediction. In Proceedings of International Joint Conference on Artificial Intelligence, 2739--2745

  19. [27]

    Liu, L.; Chen, Y.; Das, M.; Yang, H.; and Tong, H. 2023. Knowledge Graph Question Answering with Ambiguous Query. In Proceedings of the ACM Web Conference, 2477--2486

  20. [28]

    Ma, K.; Fu, H.; Liu, T.; Wang, Z.; and Tao, D. 2018. Deep blur mapping: Exploiting high-level semantics by deep neural networks. IEEE Transactions on Image Processing, 27(10): 5155--5166

  21. [29]

    Y.; and Zeng, X

    Ma, T.; Lin, X.; Song, B.; Philip, S. Y.; and Zeng, X. 2022. Kg-mtl: Knowledge graph enhanced multi-task learning for molecular interaction. IEEE Transactions on Knowledge and Data Engineering, 35(7): 7068--7081

  22. [30]

    Mai, S.; Zheng, S.; Yang, Y.; and Hu, H. 2021. Communicative message passing for inductive relation reasoning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, 4294--4302

  23. [31]

    Meilicke, C.; Fink, M.; Wang, Y.; Ruffinelli, D.; Gemulla, R.; and Stuckenschmidt, H. 2018. Fine-grained evaluation of rule-and embedding-based systems for knowledge graph completion. In Proceedings of International Semantic Web Conference, 3--20. Springer

  24. [32]

    Nair, V.; and Hinton, G. E. 2010. Rectified linear units improve restricted boltzmann machines. In Proceedings of International Conference on Machine Learning, 807--814

  25. [33]

    Pan, Y.; Liu, J.; Zhang, L.; Zhao, T.; Lin, Q.; Hu, X.; and Wang, Q. 2022. Inductive relation prediction with logical reasoning using contrastive representations. In Proceedings of Conference on Empirical Methods in Natural Language Processing, 4261--4274

  26. [34]

    Pujara, J.; Augustine, E.; and Getoor, L. 2017. Sparsity and noise: Where knowledge graph embeddings fall short. In Proceedings of Conference on Empirical Methods in Natural Language Processing, 1751--1756

  27. [35]

    Qu, M.; Chen, J.; Xhonneux, L.-P.; Bengio, Y.; and Tang, J. 2020. Rnnlogic: Learning logic rules for reasoning on knowledge graphs. In Proceedings of International Conference on Learning Representations

  28. [36]

    Quan, Y.; Ding, J.; Gao, C.; Yi, L.; Jin, D.; and Li, Y. 2023. Robust Preference-Guided Denoising for Graph based Social Recommendation. In The World Wide Web Conference, 1097--1108

  29. [37]

    Sadeghian, A.; Armandpour, M.; Ding, P.; and Wang, D. Z. 2019. Drum: End-to-end differentiable rule mining on knowledge graphs. Advances in Neural Information Processing Systems, 32

  30. [38]

    N.; Bloem, P.; Van Den Berg, R.; Titov, I.; and Welling, M

    Schlichtkrull, M.; Kipf, T. N.; Bloem, P.; Van Den Berg, R.; Titov, I.; and Welling, M. 2018. Modeling relational data with graph convolutional networks. In Proceedings of Extended Semantic Web Conference, 593--607

  31. [39]

    Sun, Q.; Li, J.; Peng, H.; Wu, J.; Fu, X.; Ji, C.; and Philip, S. Y. 2022. Graph structure learning with variational information bottleneck. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, 4165--4174

  32. [40]

    Teru, K.; Denis, E.; and Hamilton, W. 2020. Inductive relation prediction by subgraph reasoning. In Proceedings of International Conference on Machine Learning, 9448--9457

  33. [41]

    Tian, C.; Xie, Y.; Li, Y.; Yang, N.; and Zhao, W. X. 2022. Learning to denoise unreliable interactions for graph collaborative filtering. In Proceedings of International ACM SIGIR Conference on Research and Development in Information Retrieval, 122--132

  34. [42]

    C.; and Bialek, W

    Tishby, N.; Pereira, F. C.; and Bialek, W. 2000. The information bottleneck method. arXiv preprint physics/0004057

  35. [43]

    Toutanova, K.; Chen, D.; Pantel, P.; Poon, H.; Choudhury, P.; and Gamon, M. 2015. Representing text for joint embedding of text and knowledge bases. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, 1499--1509

  36. [44]

    Wang, P.; Han, J.; Li, C.; and Pan, R. 2019. Logic attention based neighborhood aggregation for inductive knowledge graph embedding. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 33, 7152--7159

  37. [45]

    Wang, Q.; Mao, Z.; Wang, B.; and Guo, L. 2017. Knowledge graph embedding: A survey of approaches and applications. IEEE Transactions on Knowledge and Data Engineering, 29(12): 2724--2743

  38. [46]

    Wang, W.; Feng, F.; He, X.; Nie, L.; and Chua, T.-S. 2021. Denoising implicit feedback for recommendation. In Proceedings of ACM International Conference on Web Search and Data Mining, 373--381

  39. [47]

    Xiong, W.; Hoang, T.; and Wang, W. Y. 2017. Deeppath: A reinforcement learning method for knowledge graph reasoning. Proceedings of the Conference on Empirical Methods in Natural Language Processing, 564--573

  40. [48]

    Xu, X.; Zhang, P.; He, Y.; Chao, C.; and Yan, C. 2022. Subgraph neighboring relations infomax for inductive link prediction on knowledge graphs. In Proceedings of International Joint Conference on Artificial Intelligence, 2341--2347

  41. [49]

    Xue, B.; and Zou, L. 2023. Knowledge Graph Quality Management: A Comprehensive Survey. IEEE Transactions on Knowledge and Data Engineering, 35(5): 4969--4988

  42. [50]

    Yan, Z.; Ma, T.; Gao, L.; Tang, Z.; and Chen, C. 2022. Cycle representation learning for inductive relation prediction. In Proceedings of International Conference on Machine Learning, 24895--24910. PMLR

  43. [51]

    Yang, F.; Yang, Z.; and Cohen, W. W. 2017. Differentiable learning of logical rules for knowledge base reasoning. Advances in Neural Information Processing Systems, 30

  44. [52]

    Yang, Y.; Huang, C.; Xia, L.; and Huang, C. 2023. Knowledge graph self-supervised rationalization for recommendation. In Proceedings of ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 3046--3056

  45. [53]

    Zhang, M.; and Chen, Y. 2018. Link prediction based on graph neural networks. Advances in Neural Information Processing Systems, 31

  46. [54]

    Zhang, Q.; Dong, J.; Tan, Q.; and Huang, X. 2024. Integrating Entity Attributes for Error-Aware Knowledge Graph Embedding. IEEE Transactions on Knowledge and Data Engineering, 36(4): 1667--1682

  47. [55]

    Zhang, W.; Paudel, B.; Wang, L.; Chen, J.; Zhu, H.; Zhang, W.; Bernstein, A.; and Chen, H. 2019. Iteratively learning embeddings and rules for knowledge graph reasoning. In The World Wide Web Conference, 2366--2377

  48. [56]

    Zhang, W.; Zhu, Y.; Chen, M.; Geng, Y.; Huang, Y.; Xu, Y.; Song, W.; and Chen, H. 2023 a . Structure Pretraining and Prompt Tuning for Knowledge Graph Transfer. In Proceedings of the ACM Web Conference, 2581--2590

  49. [57]

    Zhang, X.; and Zitnik, M. 2020. Gnnguard: Defending graph neural networks against adversarial attacks. Advances in Neural Information Processing Systems, 33: 9263--9275

  50. [58]

    Zhang, Y.; and Yao, Q. 2022. Knowledge graph reasoning with relational digraph. In Proceedings of the ACM Web Conference, 912--924

  51. [59]

    Zhang, Y.; Zhou, Z.; Yao, Q.; Chu, X.; and Han, B. 2023 b . Adaprop: Learning adaptive propagation for graph neural network based knowledge graph reasoning. In Proceedings of ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 3446--3457

  52. [60]

    Zhu, X.; Du, Y.; Mao, Y.; Chen, L.; Hu, Y.; and Gao, Y. 2023. Knowledge-refined Denoising Network for Robust Recommendation. In Proceedings of ACM SIGIR Conference on Research and Development in Information Retrieval, 362--371

  53. [61]

    Zhu, Z.; Yuan, X.; Galkin, M.; Xhonneux, L.-P.; Zhang, M.; Gazeau, M.; and Tang, J. 2024. A* net: A scalable path-based reasoning approach for knowledge graphs. Advances in Neural Information Processing Systems, 36

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.