Pith. sign in

REVIEW 5 major objections 6 minor 29 references

Enhancing Knowledge Graph Completion with GNN Distillation and Probabilistic Interaction Modeling

T0 review · 5 major / 6 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 This paper claims that two add-on modules—iterative message distillation and abstract probabilistic interaction modeling—reliably improve knowledge graph completion across GNN-based and embedding-based baselines.

desk verdict A plausible pair of cheap add-on modules, but the undefined Distill operator and the unquantified 'significant' claims make the central results unverifiable as written. read the letter →

arxiv 2505.12272 v1 pith:VWVALRZU submitted 2025-05-18 cs.AI cs.LG

Lingzhi Wang , Pengcheng Huang , Haotian Li , Yuliang Wei , Guodong Xin , Rui Zhang , Donglin Zhang , Zhenzhou Ji
show 1 more author
Wei Wang
This is my paper · ORCID
classification cs.AIcs.LG
keywords knowledgegraphcompletionGNNdistillationover-smoothingabstractprobabilisticinteractionmodelingsignaturestransitionmatricesWN18RRFB15K-237
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

Knowledge graph completion predicts missing links in incomplete graphs, but deep graph neural networks tend to collapse node representations through over-smoothing, and embedding models miss abstract relational structure. This paper tries to establish that two add-on modules fix those failures: an iterative GNN message-distillation step that filters out redundant features at each layer, and an abstract probabilistic interaction module that scores triples through learned relation-conditioned transition matrices over sparse entity interaction modes. If the reported results hold, any existing GNN or embedding KGC model can be strengthened by grafting on these modules without redesigning the model itself. The paper supports the claim with experiments on WN18RR and FB15K-237 across KB-GAT, RGCN, CompGCN, and SimKGc.

What carries the argument

The load-bearing object on the GNN side is the distillation operator $Distill(\tilde{m}_i^{(l,k)}; \alpha_i^{(l,k)})$, an iterative message filter that runs $K$ rounds per layer and progressively retains a fraction $\alpha$ of the most informative features, with $\alpha$ following a linear or exponential decay; the paper does not specify the operator's concrete form. On the embedding side the load-bearing object is the APIM score $f(h,r,t)=\tilde{a}_h^\top P_r \tilde{a}_t$, built from a sigmoid entity signature $a_e=\sigma(W_a h_e)$, a top-$k$ sparsification mask, and a relation-specific transition matrix $P_r=\tanh(\Theta_r)$. The distillation operator is meant to prevent over-smoothing while the bilinear expectation score is meant to capture abstract relational patterns. These two objects carry the argument: the rest of the paper is the claim that inserting them into existing models and training with a combined loss yields the reported benchmark gains.

What would settle it

Take the four-layer KB-GAT setup on FB15K-237 and replace the distillation operator with random feature masking that removes the same fraction of message dimensions under the same decay schedule; if MRR stays near the reported 0.3588 or higher, then the specific distillation mechanism is not the source of the gain. A second check: run the same models with APIM removed and distillation kept, and then with distillation removed and APIM kept, on both datasets; the gap between those two ablations reveals which module carries the improvement.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that two mechanisms—distillation and probabilistic interaction modeling—are complementary and portable. GNN distillation treats the aggregated message at each layer as a signal to be refined over several rounds: only features judged important survive, under a linear or exponential decay schedule, so nodes keep local distinctiveness while still absorbing global context. APIM encodes each entity as a sparse probabilistic signature over latent interaction modes, gives each relation a trainable transition matrix between modes, and scores a triple as the expected interaction $\tilde{a}_h^\top P_r \tilde{a}_t$; the module is trained with binary cross-entropy and a Frobenius penalty. Applied separately or together to four-layer GNNs and to the embedding/LLM-based SimKGc, the enhanced variants outperform their unmodified baselines on MRR and Hits@N, with the largest reported single gain being the WN18RR SimKGc MRR improvement from 0.626 to 0.658. The paper further reports that retaining about 20 interaction modes and a linear distillation decay from 1.0 to 0.2 are the best operating points.

Load-bearing premise

The argument collapses if the distillation operator does not actually preserve informative features while removing redundant ones, because the paper never defines how the operator or its importance scores work.

Editorial extensions

If this is right

  • GNN distillation should make deeper KGC architectures usable: the paper's four-layer models with distillation beat their unmodified counterparts, so the technique is a candidate remedy for over-smoothing in even deeper networks.
  • The merged APIM-plus-distillation variant is reported to exceed the additive effects of the individual modules, suggesting that the two mechanisms do not merely duplicate each other.
  • Moderate hyperparameter settings matter: retaining about 20 interaction modes and linearly decaying the distillation ratio from 1.0 to 0.2 gave the best results, while more aggressive filtering degraded performance.
  • Because the distillation module is inserted at every GNN layer and APIM is appended at the end, the combined loss $\mathcal{L}=\mathcal{L}_{\text{GNN}}+\lambda_{\text{APIM}}\mathcal{L}_{\text{APIM}}$ is a reusable recipe for other GNN-based KGC models.

Reading between the lines

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

  • The author-defined distillation operator is a placeholder: a concrete implementation is needed, and a natural candidate is hard top-k masking of message dimensions with an energy threshold analogous to the APIM top-20 rule; this would make the over-smoothing mechanism directly testable.
  • Since the APIM score is a relation-conditioned bilinear form over entity signatures, it is mathematically close to the factored bilinear embedding family; models in that family could absorb APIM almost for free, and gains should be largest on datasets where relational patterns are hierarchical, matching the larger reported improvements on WN18RR.
  • The paper does not isolate the distillation effect through a module-level ablation; a natural test of the claimed synergy would compare KB-GAT-DIST, KB-GAT-APIM, and KB-GAT-MERG against a version that trains both branches but replaces distillation with the identity map.
  • If the distillation mechanism is what the paper says, it should transfer to node classification on non-relational graphs, where over-smoothing is also a known failure mode; running GCN or GAT on citation graphs would be a cheap transfer test.
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

5 major / 6 minor

Summary. The manuscript proposes two modules for knowledge graph completion: GNN distillation, which filters messages inside each GNN layer to mitigate over-smoothing, and Abstract Probabilistic Interaction Modeling (APIM), which projects entity embeddings to sparse probabilistic signatures and scores triples as an expectation under relation-specific transition matrices. The methods are attached to KB-GAT, RGCN, CompGCN, and SimKGC, with APIM, distillation, and merged variants, and are evaluated on WN18RR and FB15K-237. The paper claims consistent significant gains over baselines and a synergistic effect for the merged variant.

Significance. If the empirical claims were fully supported, this would be a useful and model-agnostic contribution: the two modules are simple, they can be grafted onto both GNN-based and embedding-based KGC models, and the evaluation covers standard benchmarks with five-run averages, ablations on top-k and decay schedules, and an appendix analysis. The main weakness is not the idea but verifiability: the distillation operator is unspecified, the merged-variant synergy claim is contradicted by Table 1, and the word "significant" is used without error bars or significance tests. The central claim may be true, but the manuscript as written does not allow a reader to check it.

major comments (5)
  1. [3.1, Eqs. (2)–(4)] Section 3.1, Eqs. (2)–(4): the operator Distill is never defined. Equation (2) writes \tilde m_i^{(l)} = Distill(m_i^{(l)}; \alpha_i^{(l)}), and Eq. (4) gives a scalar schedule for \alpha(k), but the paper does not state what Distill does to a message, how \alpha_i^{(l,k)} is computed from the message or node, or which features are retained. Because every DIST and MERGE row in Table 1 and the over-smoothing explanation depend on this operator, the GNN-distillation half of the central claim is currently unreproducible. Please provide the exact filtering operation, such as top-percentile masking, thresholding on message norms, or soft shrinkage, together with the formula for the per-message importance.
  2. [Table 1, Observation 2] Table 1, Observation 2: the claimed synergistic performance of MERGE is not visible in the reported numbers. On WN18RR, KB-GAT-MERG (MRR 0.466) is below both KB-GAT-APIM (0.478) and KB-GAT-DIST (0.469); CompGCN-MERG (0.444) is below CompGCN-APIM (0.472) and CompGCN-DIST (0.466); and RGCN-MERG (0.436) is below RGCN-DIST (0.455). The merged variant is often not even the best among the three variants, so the statement that it exceeds the additive effects of its components is contradicted by the paper's own data. Please either correct the claim or identify the specific sense of synergy that the table supports.
  3. [Table 1 and Observation 1] Table 1 and Observation 1: the paper says results are averaged over 5 runs and repeatedly calls gains significant, but no standard deviations, confidence intervals, or significance tests are reported. Differences such as Simkgc-APIM's +0.001 MRR on FB15K-237 and KB-GAT-APIM's +0.005 MRR cannot be assessed without variance information, so the claims of "significant" improvements and "consistent and significant improvements across all evaluated architectures" are not supported. Please add error bars or paired significance tests, or soften the statistical language.
  4. [Section 5 vs. Section 4.3] Section 5, Key Findings: the discussion names "linear decay from 1.0 to 0.4" as the moderate filtering ratio that yields the best performance, but Section 4.3 and Figure 5 identify the 1.0-to-0.2 setting as the best (MRR 0.3588 on FB15K-237) and describe the 1.0-to-0.4 setting as underperforming. This is an internal contradiction in the paper's practical recommendation and should be corrected.
  5. [Appendix D] Appendix D: the cumulative-energy analysis does not provide independent evidence for the top-20 choice. The signatures a_e are taken from a model already trained with the TopK mask, so observing E(20) ≥ 0.85 partly reflects the imposed sparsity rather than a property of the underlying interaction distribution. In addition, the claim of "statistical significance" is not accompanied by any test. Please present this analysis as descriptive or provide an external validation, for example by comparing energy retention against random mode subsets or other values of K.
minor comments (6)
  1. [Throughout] There are numerous typos and inconsistencies, including "Emebdding-based" in Section 2.3, "gradially" and "intial" in Section 3.1, "an transforming" in Section 2.2, "Methodologys" in Section 3, "ComPGCN" versus "CompGCN" in Sections 3.3 and 4, and "the the" in Section 1.
  2. [Section 4.2, Observation 1] Observation 1 is incomplete: the bullet says the methods "achieve comparable performance to their corresponding," which reads as a truncated sentence and should be finished.
  3. [Section 3.2, Eq. (7)] Equation (7) uses TopK(·,k) and the text says the mask preserves the indices of the k largest values in α_e, but the variable should be the signature a_e; the formula itself is otherwise clear.
  4. [Section 3.2, Eq. (9)] The phrase "probabilistic interaction" is used although Eq. (9) normalizes P_r via tanh, which allows negative entries; please clarify whether P_r is a probability matrix or a signed interaction-weight matrix.
  5. [Section 3.2, Eq. (11)] Equation (11) writes λ·|P_r|^2_F; use \|P_r\|_F^2 for the Frobenius norm to avoid ambiguity.
  6. [Figure 5] Figure 5's axis ticks are dense and the bar labels partially overlap; add clear group labels and a caption defining what "Linear decay 1.0 0.2" means.

Circularity Check

1 steps flagged · score 2.0 of 10

Main empirical claims are independent of the method definitions; the only reduction-by-construction is the Appendix D Top-20 energy validation, which is a non-load-bearing restatement of top-k.

  1. self definitional [Appendix D, Eq. (7), Eq. (14)-(15); Section 4.1 hyperparameters]
    "eae =TopK(a e,k)⊙a e (7) where TopK(·,k) generates a binary mask preserving the indices of the k largest values in αe ... E(k) = Pk i=1 a(i) / PK j=1 a(j) (15) ... The energy curve shows the Top-20 mode selection robustly preserves≥85% of signature energy across heterogeneous knowledge graphs, satisfying the≥85% retention threshold with statistical significance."

    Equation (15) measures the l1-energy retained by the top-k entries of the very same signature ae used to construct eae in Equation (7). Because TopK is defined as keeping the k largest entries, the retained energy ratio is the maximum possible by construction; the claim that Top-20 preserves >=85% energy is a restatement of the definition applied to the trained signatures, not an independent validation. The 'statistical significance' wording adds no test of a separate prediction.

full rationale

The central claim -- that adding GNN distillation and APIM improves KGC performance on WN18RR and FB15K-237 -- is evaluated on fixed external test splits and is not forced by any equation. There are no load-bearing self-citations and no author-imported uniqueness theorems. The undefined Distill operator in Eq. (2) is a serious completeness and reproducibility gap, but an underspecified operator is not circular: the benchmark results are still empirical claims, not reductions to the definition. Similarly, the APIM score in Eq. (10) is a bilinear form, but the paper does not derive its success from that algebraic structure; it reports test-set measurements. The one genuine definitional reduction is the Appendix D energy-retention 'validation' of Top-20: it re-states that the selected subset has maximal cumulative energy for the same ae used to choose it, so it supplies no independent evidence. The absence of error bars and the inconsistency on the best decay setting are correctness and statistical concerns, not circularity. Overall, no central claim is definitionally circular; the circularity is minor and non-load-bearing, so the score is 2.

Assumptions & free parameters 5 free parameters · 5 assumptions · 3 invented entities

The central claims rest on a handful of hand-set hyperparameters (K=100, top-k=20, decay schedules, loss weights) and on an underspecified distillation operator. The APIM score is a trainable bilinear function with a sigmoid/tanh gloss, so the probabilistic interpretation is nominal and not independently validated.

free parameters (5)
  • K (number of latent interaction modes) = 100
    Set in Section 4.1 without ablation; dimensionality of the signature and transition matrices.
  • TopK count k = 20
    Selected by ablation in Section 4.3 and claimed to preserve 85% energy in Appendix D; this is a post hoc choice tuned on the evaluation data.
  • Distillation decay schedule (alpha_start, Delta, gamma) = 1.0, 0.2, 0.74
    Linear decay with Delta=0.2 is used in the main results, exponential gamma=0.74 appears in Appendix C, and Section 5 names a different optimum of 0.4, indicating tuning instability.
  • Loss weights lambda_APIM and lambda = not reported
    Defined in Equations (11)-(13) but actual values are never given, making reproduction impossible.
  • GNN depth = 4 layers
    Chosen as twice the commonly optimal depth to demonstrate over-smoothing (Section 4.2); no ablation across depths is shown.
assumptions (5)
  • domain assumption Message passing framework of Equation (1) is a valid and sufficient backbone for KGC models.
    The method is built on MPNNs as defined in Section 2.2; the paper does not question this modeling choice.
  • ad hoc to paper The Distill operator can compute an importance score alpha_i and filter messages so that over-smoothing is mitigated.
    Equations (2)-(4) assume such an operator exists but never define it; the over-smoothing claim depends entirely on this assumption.
  • domain assumption Filtered evaluation under the closed-world assumption is the correct protocol for KGC.
    Section 2.1 adopts the closed-world assumption and Appendix B describes filtered ranking; this is standard for the task.
  • ad hoc to paper Cumulative energy E(k) of sorted signature values is a valid measure of information preservation, and the 85% threshold justifies top-20 selection.
    Appendix D computes E(k) on trained signatures and asserts statistical significance without a test; this is a post hoc justification of a tuned hyperparameter.
  • domain assumption Sigmoid-bounded signatures and tanh transition matrices represent meaningful probabilistic interactions between entities and relations.
    Equations (6)-(9) interpret bounded values as probabilities, but tanh outputs can be negative and no probabilistic normalization is enforced.
invented entities (3)
  • Abstract probabilistic interaction signature a_e
    purpose: Represents each entity as a sparse probability-like vector over K latent interaction modes (Equations 6-7).
    It is a learned transformation of the entity embedding; no external prediction or measurement anchors it.
  • Relation transition matrix P_r
    purpose: Encodes conditional mode-to-mode interaction probabilities for each relation (Equations 8-9).
    It is a trainable bilinear parameter whose probabilistic interpretation is nominal because tanh can yield negative entries.
  • Latent interaction modes
    purpose: A K-dimensional latent space in which entities and relations interact (Equations 6-10).
    The modes are never observed or probed independently; Appendix D only visualizes relation-specific weights on the same trained model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Knowledge Graph Completion with GNN Distillation and Probabilistic Interaction Modeling." pith.science (2026). https://pith.science/paper/VWVALRZU

@misc{pith2026250512272,
  author       = {Pith},
  title        = {Pith review of: Enhancing Knowledge Graph Completion with GNN Distillation and Probabilistic Interaction Modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VWVALRZU}},
  note         = {Machine review of arXiv:2505.12272}
}
read the original abstract

Knowledge graphs (KGs) serve as fundamental structures for organizing interconnected data across diverse domains. However, most KGs remain incomplete, limiting their effectiveness in downstream applications. Knowledge graph completion (KGC) aims to address this issue by inferring missing links, but existing methods face critical challenges: deep graph neural networks (GNNs) suffer from over-smoothing, while embedding-based models fail to capture abstract relational features. This study aims to overcome these limitations by proposing a unified framework that integrates GNN distillation and abstract probabilistic interaction modeling (APIM). GNN distillation approach introduces an iterative message-feature filtering process to mitigate over-smoothing, preserving the discriminative power of node representations. APIM module complements this by learning structured, abstract interaction patterns through probabilistic signatures and transition matrices, allowing for a richer, more flexible representation of entity and relation interactions. We apply these methods to GNN-based models and the APIM to embedding-based KGC models, conducting extensive evaluations on the widely used WN18RR and FB15K-237 datasets. Our results demonstrate significant performance gains over baseline models, showcasing the effectiveness of the proposed techniques. The findings highlight the importance of both controlling information propagation and leveraging structured probabilistic modeling, offering new avenues for advancing knowledge graph completion. And our codes are available at https://anonymous.4open.science/r/APIM_and_GNN-Distillation-461C.

Figures

Figures reproduced from arXiv: 2505.12272 by the authors.

Figure 1
Figure 1. A general structure of the MPNNs-based KGC [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The visualisation of the GNN Distillation [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The enhanced KB-GAT framework integrates a GNN distillation module that refines message representa [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Impact of APIM’s Top-K mode retention thresholds on KGC performance (WN18RR) MRR Hit@1 Hit@3 Hit@10 0.0 0.1 0.2 0.3 0.4 0.5 0.3176 0.2337 0.3444 0.4865 0.3588 0.2687 0.3922 0.5390 0.3027 0.2246 0.3262 0.4512 Linear decay 1.0 0.4 Linear decay 1.0 0.2 Linear decay 1.0 0.…
Figure 5
Figure 5. Figure 5: Impact of GNN Distillation’s filtering ratio on [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Comparison of linear and exponential decay strategies on the FB15K-237 dataset across three GNN [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Mode retention validation results on FB15K-237 using CompGCN. In energy curve (a), the red dashed [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 12 canonical work pages

  1. [1]

    Ivana Bala z evi \'c , Carl Allen, and Timothy M Hospedales. 2019. Tucker: Tensor factorization for knowledge graph completion. arXiv preprint arXiv:1901.09590

    arXiv 2019

  2. [2]

    Kurt Bollacker, Colin Evans, Praveen Paritosh, Tim Sturge, and Jamie Taylor. 2008. Freebase: a collaboratively created graph database for structuring human knowledge. In Proceedings of the 2008 ACM SIGMOD international conference on Management of data, pages 1247--1250

    2008

  3. [3]

    Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and Oksana Yakhnenko. 2013. Translating embeddings for modeling multi-relational data. Advances in neural information processing systems, 26

    2013

  4. [4]

    Faxian Cao, Yongqiang Cheng, Adil Mehmood Khan, and Zhijing Yang. 2023. https://arxiv.org/abs/2305.11387 Justices for information bottleneck theory . Preprint, arXiv:2305.11387

    work page Pith review arXiv 2023

  5. [5]

    Hai Cui, Tao Peng, Feng Xiao, Jiayu Han, Ridong Han, and Lu Liu. 2023. Incorporating anticipation embedding into reinforcement learning framework for multi-hop knowledge graph question answering. Information Sciences, 619:745--761

    work page 2023

  6. [6]

    Tim Dettmers, Pasquale Minervini, Pontus Stenetorp, and Sebastian Riedel. 2018. Convolutional 2d knowledge graph embeddings. In Proceedings of the AAAI conference on artificial intelligence, volume 32

    2018

  7. [7]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://doi.org/10.18653/v1/N19-1423 BERT : Pre-training of deep bidirectional transformers for language understanding . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long a...

    doi:10.18653/v1/n19-1423 2019

  8. [8]

    Jakub Dutkiewicz and Czes aw J e drzejek. 2024. Knowledge graphs in information retrieval. In Proceedings of The 25th European Conference on Knowledge Management. Academic Conferences International

    work page 2024

Show all 29 references
  1. [9]

    Hao Fei, Yafeng Ren, Yue Zhang, Donghong Ji, and Xiaohui Liang. 2021. Enriching contextualized language model from knowledge graph for biomedical information extraction. Briefings in bioinformatics, 22(3):bbaa110

    2021

  2. [10]

    Will Hamilton, Zhitao Ying, and Jure Leskovec. 2017. Inductive representation learning on large graphs. Advances in neural information processing systems, 30

    2017

  3. [11]

    Lei Hu, Wenwen Li, Jun Xu, and Yunqiang Zhu. 2024. Geoentity-type constrained knowledge graph embedding for predicting natural-language spatial relations. International Journal of Geographical Information Science, pages 1--24

    2024

  4. [12]

    Jens Lehmann, Robert Isele, Max Jakob, Anja Jentzsch, Dimitris Kontokostas, Pablo N Mendes, Sebastian Hellmann, Mohamed Morsey, Patrick Van Kleef, S \"o ren Auer, et al. 2015. Dbpedia--a large-scale, multilingual knowledge base extracted from wikipedia. Semantic web, 6(2):167--195

    2015

  5. [13]

    Juanhui Li, Harry Shomer, Jiayuan Ding, Yiqi Wang, Yao Ma, Neil Shah, Jiliang Tang, and Dawei Yin. 2023. Are message passing neural networks really helpful for knowledge graph completion? In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistic...

    2023

  6. [14]

    Qimai Li, Zhichao Han, and Xiao-Ming Wu. 2018. Deeper insights into graph convolutional networks for semi-supervised learning. In Proceedings of the AAAI conference on artificial intelligence, volume 32

    2018

  7. [15]

    Deepak Nathani, Jatin Chauhan, Charu Sharma, and Manohar Kaul. 2019. Learning attention-based embeddings for relation prediction in knowledge graphs. arXiv preprint arXiv:1906.01195

    2019 arXiv

  8. [16]

    Anam Nazir and Ze Wang. 2023. A comprehensive survey of chatgpt: advancements, applications, prospects, and challenges. Meta-radiology, page 100022

    2023

  9. [17]

    Byungkook Oh, Seungmin Seo, Jimin Hwang, Dongho Lee, and Kyong-Ho Lee. 2022. Open-world knowledge graph completion for unseen entities and relations via attentive feature aggregation. Information sciences, 586:468--484

    2022

  10. [18]

    Kenta Oono and Taiji Suzuki. 2019. Graph neural networks exponentially lose expressive power for node classification. arXiv preprint arXiv:1905.10947

    2019 arXiv

  11. [19]

    Jeff Z Pan, Guido Vetere, Jose Manuel Gomez-Perez, and Honghan Wu. 2017. Exploiting linked data and knowledge graphs in large organisations. Springer

    2017

  12. [20]

    Michael Schlichtkrull, Thomas N Kipf, Peter Bloem, Rianne Van Den Berg, Ivan Titov, and Max Welling. 2018. Modeling relational data with graph convolutional networks. In The semantic web: 15th international conference, ESWC 2018, Heraklion, Crete, Greece, June 3--7, 2018, proc...

    2018

  13. [21]

    Tong Shen, Fu Zhang, and Jingwei Cheng. 2022. A comprehensive overview of knowledge graph completion. Knowledge-Based Systems, 255:109597

    2022

  14. [22]

    Kristina Toutanova and Danqi Chen. 2015. 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

  15. [23]

    Th \'e o Trouillon, Johannes Welbl, Sebastian Riedel, \'E ric Gaussier, and Guillaume Bouchard. 2016. Complex embeddings for simple link prediction. In International conference on machine learning, pages 2071--2080. PMLR

    2016

  16. [24]

    Petar Veli c kovi \'c , Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. 2017. Graph attention networks. arXiv preprint arXiv:1710.10903

    2017 arXiv

  17. [25]

    Denny Vrande c i \'c and Markus Kr \"o tzsch. 2014. Wikidata: a free collaborative knowledgebase. Communications of the ACM, 57(10):78--85

    2014

  18. [26]

    Liang Wang, Wei Zhao, Zhuoyu Wei, and Jingming Liu. 2022. Simkgc: Simple contrastive knowledge graph completion with pre-trained language models. arXiv preprint arXiv:2203.02167

    2022 arXiv

  19. [27]

    Xu Wang, Zijin Luo, Rui He, and Yixin Shao. 2023. Novel medical question and answer system: graph convolutional neural network based with knowledge graph optimization. Expert Systems with Applications, 227:120211

    2023

  20. [28]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  21. [29]

    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 gl...

Pith tools

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