Pith. sign in

REVIEW 2 major objections 4 minor 71 references

Treating each knowledge graph as a partial view of an entity, conditional diffusion can generate domain-general embeddings that transfer knowledge without erasing domain-specific context.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-12 04:28 UTC pith:UW66HPL6

load-bearing objection Clean generation-based alternative to consistency MKGC that delivers consistent 4.3% MRR gains and holds up under low-resource stress tests. the 2 major comments →

arxiv 2607.03154 v1 pith:UW66HPL6 submitted 2026-07-03 cs.CL cs.AI

Conditional Diffusion Guided Knowledge Transfer for Multi-Domain Knowledge Graph Completion

Jiawei Sheng , Taoyu Su , Xixun Lin , Xiaodong Li , Tingwen Liu This is my paper
classification cs.CL cs.AI
keywords knowledge graph completionmulti-domain learningdiffusion modelsknowledge transferrepresentation learningconditional generation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Knowledge graphs built for different languages or domains share some entities but leave many triples incomplete. Prior multi-domain methods force those shared entities to look alike across graphs, which can erase useful local context and hurt prediction when data are scarce. This paper instead treats every graph as a partial observation of the same underlying entity and uses a conditional diffusion model to generate a single domain-general embedding for that entity. The generation is guided by the support graphs yet is trained to reconstruct a domain-agnostic prior, so it stays unbiased; at the same time the generated embedding is required to remain predictive inside the target graph. On fourteen graphs across three benchmarks the resulting method improves average mean reciprocal rank by 4.3 percent, and the gains hold when equivalent entities, support graphs, or training triples are deliberately reduced.

Core claim

A conditional diffusion process that reconstructs domain-agnostic prior entity embeddings, while being guided by fused support-graph representations and jointly trained for target-graph link prediction, produces more informative entity embeddings than consistency constraints and yields a 4.3 percent average MRR gain on multi-domain knowledge-graph completion.

What carries the argument

DMKGC’s conditional diffusion transfer: prior embeddings serve as the unbiased reconstruction target; support-graph entity embeddings are attentively fused into a condition; the reverse process is trained both to match the prior (generation + single-domain regularization losses) and to keep the generated embedding task-predictive inside the target graph.

Load-bearing premise

The randomly initialized, domain-agnostic prior embeddings are assumed to be a rich enough and unbiased target for the diffusion model to reconstruct; if that prior is uninformative, the claimed advantage over simple consistency constraints disappears.

What would settle it

Replace the learned prior with pure Gaussian noise (or freeze the prior after random initialization) and re-run the identical training pipeline on DBP-5L; if the 4.3 percent MRR lift vanishes, the proxy-prior assumption is falsified.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes DMKGC, a generation-based framework for multi-domain knowledge graph completion (MKGC). Instead of enforcing consistency constraints on equivalent entities, it treats each KG as a partial view of an entity and uses a conditional diffusion model to generate domain-general entity embeddings. Domain-agnostic prior embeddings (shared random initializations before KG encoding) serve as the proxy generation target; support-KG encodings provide the condition via an attentive fuser; and the generated embeddings are jointly trained to remain task-predictive on the target KG (margin loss) while a single-domain regularization encourages consistency under partial conditions. Experiments on 14 KGs across three benchmarks (DBP-5L, E-PKG, DWY) report a 4.3% average MRR gain over strong multi-domain baselines, with ablations and low-resource/unseen-entity protocols supporting robustness.

Significance. If the empirical pattern holds, the work supplies a concrete alternative to the dominant consistency-based paradigm for MKGC. The generation framing, the use of priors as unbiased proxy targets, and the joint task+generation+regularization objective are clearly articulated and yield consistent gains across multilingual, industrial, and multi-source settings, including reduced equivalent-entity ratios, fewer support KGs, and fully unseen heads. Ablations isolate the encoder, condition, diffusion module, and regularization; comparisons against InfoNCE, DA-DIFF, and MMD under the same backbone further locate the contribution. Code is promised, and the low-resource protocols are practically relevant. These elements make the paper a useful addition to the multi-domain KGC literature even if the absolute novelty of diffusion for transfer is incremental.

major comments (2)
  1. The central modeling claim rests on the prior embeddings (Eq. 4) being a sufficiently informative and unbiased proxy generation target (Eqs. 14, 18). Table 4 shows that removing diffusion, condition, or regularization each hurts AVG-MRR, yet there is no independent diagnostic (e.g., reconstruction quality of the prior, or a controlled comparison that freezes the prior after random initialization) that isolates whether the prior itself carries useful structure versus simply acting as a convenient reconstruction anchor that is jointly optimized under L_task. A short diagnostic or discussion would strengthen the generation-based narrative.
  2. Main results (Tables 1–3) and low-resource figures report point estimates without error bars or multi-seed statistics, despite free parameters (T, s, ω1, ω2, p_u) and grid search. Given that the headline claim is a 4.3% average MRR improvement, reporting variance (or at least three-seed means) on the primary AVG-MRR numbers would make the significance of the gains clearer and is standard for this class of embedding models.
minor comments (4)
  1. Notation: the same symbol t is used both for the target KG index and (implicitly) for time steps in the diffusion process; a brief clarification or distinct symbols would help.
  2. Figure 3 caption and the surrounding text could more explicitly mark which embeddings are shared versus domain-specific; the current diagram is dense.
  3. Appendix Table 7 (dataset statistics) and the hyper-parameter table are useful; a one-sentence note on how virtual isolated entities are handled for missing equivalents would aid reproducibility.
  4. A few typos and minor phrasing issues appear (e.g., “surpassing domain-specific” in the conclusion; “LSGMA” vs “LSMGA” inconsistency in places).

Circularity Check

0 steps flagged

No significant circularity: purely empirical method with standard losses and external benchmarks; priors are jointly optimized parameters, not a definitional reduction of the reported MRR gains.

full rationale

The paper proposes a generation-based MKGC framework (DMKGC) that initializes shared random prior embeddings (Eq. 4), encodes them per KG (Eq. 5), and trains a conditional diffusion reverse process to reconstruct those priors under support-KG conditions (L_gen Eq. 14, L_reg Eq. 18) while also optimizing a standard margin-based KGC loss on the fused embeddings (L_task Eq. 17). The overall objective (Eq. 19) is a conventional weighted sum; nothing forces the test-set MRR numbers by construction. Ablations (Table 4) and comparisons to alternative transfer methods (Table 5) and low-resource settings (Figs. 4-5, Table 6) are independent empirical checks against external baselines (LSMGA, GLKGC, etc.). No self-definitional loop, no fitted parameter renamed as prediction, no load-bearing uniqueness theorem imported from the authors, and no ansatz smuggled via self-citation. The work is self-contained against the three public benchmarks.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 2 invented entities

The central empirical claim rests on a small set of modeling choices (shared random priors as unbiased targets, each KG as a partial view, diffusion as the transfer operator) plus a handful of free hyper-parameters tuned by grid search. No deep mathematical axioms are required; the load-bearing pieces are the domain assumptions about multi-KG structure and the ad-hoc proxy construction that makes the generation objective well-defined.

free parameters (5)
  • diffusion steps T
    Chosen by grid search in {2,4,...,64}; optimal values 16/32/16 on the three datasets; directly controls generation fidelity and compute.
  • condition strength s
    CFG scale tuned in {1..5}; optimal 2/1/2; controls how strongly support conditions bias the reverse process.
  • loss weights omega1, omega2
    Balance generation/regularization against task loss; tuned over {1,3,5}e-{4..1}; optimal 1e-2 and 1e-3.
  • unconditional ratio p_u
    Classifier-free guidance dropout probability fixed at 0.1; affects unconditional generation quality.
  • margin lambda
    Margin for the ranking loss fixed at 0.5 across all datasets.
axioms (4)
  • domain assumption Each knowledge graph supplies only a partial observational view of an underlying domain-general entity representation.
    Stated as the key insight in the abstract and §1; justifies conditioning diffusion on support KGs.
  • ad hoc to paper Shared randomly-initialized embeddings obtained before any KG encoding are domain-agnostic and therefore constitute an unbiased proxy generation target.
    Introduced in §3.1.1 and used as x0 in the diffusion objective (Eqs. 14, 18); no external justification is given.
  • domain assumption Equivalent entity pairs across KGs are given a priori and relations share a unified schema.
    Problem formulation §2.1; standard MKGC setting but required for the condition fusion to be well-defined.
  • standard math The re-parameterized x0-prediction form of the ELBO (Eq. 13) is a valid training objective for embedding generation.
    Follows standard DDPM simplifications cited from Ho et al. and later works; used throughout §3.3.
invented entities (2)
  • domain-general entity embedding produced by conditional diffusion no independent evidence
    purpose: Serves as the transferable representation that is unbiased across KGs yet still task-predictive.
    Core output of the reverse process; no independent measurement outside the joint training objective is supplied.
  • prior entity embedding used as proxy generation objective no independent evidence
    purpose: Provides a domain-agnostic reconstruction target so the diffusion does not collapse toward any single conditioned KG.
    Defined in Eq. 4 and reused as x0; its suitability is assumed rather than measured against an external ground-truth general embedding.

pith-pipeline@v1.1.0-grok45 · 27356 in / 3019 out tokens · 36979 ms · 2026-07-12T04:28:55.296101+00:00 · methodology

0 comments
read the original abstract

Multi-domain knowledge graph completion (MKGC) aims to improve missing triple prediction in a target KG by transferring knowledge from other support KGs. Existing methods typically enforce consistency constraints on equivalent entities across KGs to transfer knowledge, which risks suppressing domain-specific contextual information of entities. This design can also compromise entity representation information from all KG domains, impeding performance improvements, especially in low-resource data scenarios. To address this, we pioneer a generation-based paradigm for MKGC and propose DMKGC, a conditional diffusion-guided knowledge transfer framework. Our key insight is to treat each KG as a partial view of the entity entire information, and generate informative domain-general entity embeddings through diffusion models conditioned on support KGs. Particularly, we first initialize domain-agnostic entity embeddings as prior entity embeddings, and then encode them within individual KGs. Afterward, we fuse equivalent entities from support KGs as the conditional diffusion generation guidance. We leverage the prior entity embeddings as the proxy generation objective, which ensures this conditional generation to be unbiased towards any conditioned KGs. Simultaneously, we also train the generated embeddings to be predictive across KGs, thus preserving domain-specific information. Extensive experiments on 14 KGs in 3 benchmarks demonstrate a 4.3\% average MRR improvement in tail entity prediction over state-of-the-art methods, with sustained gains in low-resource data settings.

Figures

Figures reproduced from arXiv: 2607.03154 by Jiawei Sheng, Taoyu Su, Tingwen Liu, Xiaodong Li, Xixun Lin.

Figure 1
Figure 1. Figure 1: A toy example of the MKGC task, which predicts [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: The overview of our proposed framework, DMKGC. It contains (a) [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Results (%) with a limited number of equivalent [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Results (%) with a limited number of support KGs [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Impact of timestep number 𝑇 , condition strength 𝑠, and harmonic factors 𝜔1, 𝜔2. Averaged MRR (%) is reported with ± std of the 5 KGs on DBP-5L. A.5 Further Analysis (RQ3) A.5.1 Impact on Hyper-parameters. We show the impact of diffu￾sion time step 𝑇 , conditional strength factor 𝑠, and training balance factors 𝜔1 and 𝜔2, in [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

71 extracted references · 13 canonical work pages · 1 internal anchor

  1. [1]

    Antoine Bordes, Nicolas Usunier, Alberto Garcia-Durán, Jason Weston, and Ok- sana Yakhnenko. 2013. Translating Embeddings for Modeling Multi-Relational Data. InProceedings of NeurIPS. 2787–2795

    2013

  2. [2]

    Zongsheng Cao, Jing Li, Zigan Wang, and Jinliang Li. 2024. DiffusionE: Reasoning on Knowledge Graphs via Diffusion-based Graph Neural Networks. InProceedings of KDD. ACM, 222–230. doi:10.1145/3637528.3671997

    doi:10.1145/3637528.3671997 2024

  3. [3]

    Muhao Chen, Yingtao Tian, Mohan Yang, and Carlo Zaniolo. 2017. Multilin- gual Knowledge Graph Embeddings for Cross-lingual Knowledge Alignment. In Proceedings of IJCAI

    2017

  4. [4]

    Pan, and Hua- jun Chen

    Mingyang Chen, Wen Zhang, Yuxia Geng, Zezhong Xu, Jeff Z. Pan, and Hua- jun Chen. 2023. Generalizing to Unseen Elements: A Survey on Knowl- edge Extrapolation for Knowledge Graphs. InProceedings of IJCAI. 6574–6582. doi:10.24963/IJCAI.2023/737

    work page doi:10.24963/ijcai.2023/737 2023

  5. [5]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey E. Hinton. 2020. A Simple Framework for Contrastive Learning of Visual Representations. In Proceedings of ICML, Vol. 119. PMLR, 1597–1607. http://proceedings.mlr.press/ v119/chen20j.html

    2020

  6. [7]

    Xuelu Chen, Muhao Chen, Changjun Fan, Ankith Uppunda, Yizhou Sun, and Carlo Zaniolo. 2020. Multilingual Knowledge Graph Completion via Ensemble Knowledge Transfer. InFindings of EMNLP. 3227–3238

    2020

  7. [8]

    Tim Dettmers, Minervini Pasquale, Stenetorp Pontus, and Sebastian Riedel. 2018. Convolutional 2D Knowledge Graph Embeddings. InProceedings of AAAI. 1811– 1818

    2018

  8. [9]

    Borgwardt, Malte J

    Arthur Gretton, Karsten M. Borgwardt, Malte J. Rasch, Bernhard Schölkopf, and Alexander J. Smola. 2012. A Kernel Two-Sample Test.J. Mach. Learn. Res.13 (2012), 723–773. doi:10.5555/2503308.2188410

    work page doi:10.5555/2503308.2188410 2012

  9. [10]

    Hashimoto

    Ishaan Gulrajani and Tatsunori B. Hashimoto. 2023. Likelihood-Based Diffusion Language Models. InProceedings of NeurIPS, Alice Oh, Tristan Naumann, Amir Globerson, Kate Saenko, Moritz Hardt, and Sergey Levine (Eds.)

    2023

  10. [11]

    Lingbing Guo, Zhongpu Bo, Zhuo Chen, Yichi Zhang, Jiaoyan Chen, Yarong Lan, Mengshu Sun, Zhiqiang Zhang, Yangyifei Luo, Qian Li, Qiang Zhang, Wen Zhang, and Huajun Chen. 2024. MKGL: Mastery of a Three-Word Language. InProceedings of NeurIPS. http://papers.nips.cc/paper_files/paper/2024/hash/ fe03053bd2cf5b5c56de1e463bc53e1a-Abstract-Conference.html

    2024

  11. [12]

    Boyong He, Yuxiang Ji, Zhuoyue Tan, and Liaoni Wu. 2025. Diffusion Do- main Teacher: Diffusion Guided Domain Adaptive Object Detector.CoRR abs/2506.04211 (2025). arXiv:2506.04211 doi:10.48550/ARXIV.2506.04211

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2506.04211 2025

  12. [13]

    Jinyan He and Haitong Yang. 2024. Multilingual Knowledge Graph Completion based on Global-Local Structure Encoding. InProceedings of ISCTIS. 647–650. doi:10.1109/ISCTIS63324.2024.10699111

    work page doi:10.1109/isctis63324.2024.10699111 2024

  13. [14]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising Diffusion Probabilistic Models. InProceedings of NeurIPS

    2020

  14. [15]

    Jonathan Ho and Tim Salimans. 2022. Classifier-Free Diffusion Guidance.CoRR abs/2207.12598 (2022). arXiv:2207.12598 doi:10.48550/ARXIV.2207.12598

    Pith/arXiv arXiv doi:10.48550/arxiv.2207.12598 2022

  15. [16]

    Tao Huang, Yuan Zhang, Mingkai Zheng, Shan You, Fei Wang, Chen Qian, and Chang Xu. 2023. Knowledge Diffusion for Distillation. In Proceedings of NeurIPS. http://papers.nips.cc/paper_files/paper/2023/hash/ cdddf13f06182063c4dbde8cbd5a5c21-Abstract-Conference.html

    2023

  16. [17]

    Wei Huang, Meiyu Liang, Peining Li, Xu Hou, Yawen Li, Junping Du, Zhe Xue, and Zeli Guan. 2025. DiffusionCom: Structure-Aware Multimodal Diffusion Model for Multimodal Knowledge Graph Completion.CoRRabs/2504.06543 (2025). arXiv:2504.06543 doi:10.48550/ARXIV.2504.06543

    Pith/arXiv arXiv doi:10.48550/arxiv.2504.06543 2025

  17. [18]

    Zijie Huang, Zheng Li, Haoming Jiang, Tianyu Cao, Hanqing Lu, Bing Yin, Karthik Subbian, Yizhou Sun, and Wei Wang. 2022. Multilingual Knowledge Graph Completion with Self-Supervised Adaptive Graph Alignment. InProceedings of ACL. 474–485. doi:10.18653/V1/2022.ACL-LONG.36

    work page doi:10.18653/v1/2022.acl-long.36 2022

  18. [19]

    Yangqin Jiang, Yuhao Yang, Lianghao Xia, and Chao Huang. 2024. DiffKG: Knowledge Graph Diffusion Model for Recommendation. InProceedings of ACM WSDM. ACM, 313–321. doi:10.1145/3616855.3635850

    doi:10.1145/3616855.3635850 2024

  19. [20]

    Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization.. InInternational Conference on Learning Representations

    2014

  20. [21]

    Diederik P Kingma and Max Welling. 2014. Auto-Encoding Variational Bayes. In Proceedings of ICLR

    2014

  21. [22]

    Kipf and Max Welling

    Thomas N. Kipf and Max Welling. 2017. Semi-Supervised Classification with Graph Convolutional Networks. InInternational Conference on Learning Repre- sentations

    2017

  22. [23]

    Dawei Li, Zhen Tan, Tianlong Chen, and Huan Liu. 2024. Contextualization Distillation from Large Language Model for Knowledge Graph Completion, Yvette Graham and Matthew Purver (Eds.). 458–477. https://aclanthology.org/2024. findings-eacl.32

    2024

  23. [24]

    Krishnan, Matthew D

    Dawen Liang, Rahul G. Krishnan, Matthew D. Hoffman, and Tony Jebara. 2018. Variational Autoencoders for Collaborative Filtering. InProceedings of WWW. ACM, 689–698. doi:10.1145/3178876.3186150

    doi:10.1145/3178876.3186150 2018

  24. [25]

    Lei Liang, Zhongpu Bo, Zhengke Gui, Zhongshu Zhu, Ling Zhong, Peilong Zhao, Mengshu Sun, Zhiqiang Zhang, Jun Zhou, Wenguang Chen, Wen Zhang, and Huajun Chen. 2025. KAG: Boosting LLMs in Professional Domains via Knowledge Augmented Generation. InCompanion Proceedings of Web Conference. ACM, 334–

    2025

  25. [26]

    doi:10.1145/3701716.3715240

    doi:10.1145/3701716.3715240

  26. [27]

    Junnan Liu, Qianren Mao, Weifeng Jiang, and Jianxin Li. 2024. KNOWFORMER: revisiting transformers for knowledge graph reasoning. InProceedings of ICML (Vienna, Austria)(ICML’24). Article 1280, 22 pages

    2024

  27. [28]

    Lihui Liu, Yuzhong Chen, Mahashweta Das, Hao Yang, and Hanghang Tong. 2023. Knowledge Graph Question Answering with Ambiguous Query. InProceedings of Web Conference. ACM, 2477–2486. doi:10.1145/3543507.3583316

    doi:10.1145/3543507.3583316 2023

  28. [29]

    Yang Liu, Xiaobin Tian, Zequn Sun, and Wei Hu. 2024. Finetuning Generative Large Language Models with Discrimination Instructions for Knowledge Graph Completion. InProceedings of ISWC. 199–217

    2024

  29. [30]

    Xiao Long, Liansheng Zhuang, Aodi Li, Houqiang Li, and Shafei Wang. 2024. Fact Embedding through Diffusion Model for Knowledge Graph Completion. In Proceedings of WWW. ACM, 2020–2029. doi:10.1145/3589334.3645451

    doi:10.1145/3589334.3645451 2024

  30. [31]

    Xiao Long, Liansheng Zhuang, Aodi Li, Jiuchang Wei, Houqiang Li, and Shafei Wang. 2024. KGDM: A Diffusion Model to Capture Multiple Relation Semantics for Knowledge Graph Embedding. InProceedings of AAAI. AAAI Press, 8850–8858. doi:10.1609/AAAI.V38I8.28732

    work page doi:10.1609/aaai.v38i8.28732 2024

  31. [32]

    Weinberger

    Justin Lovelace, Varsha Kishore, Chao Wan, Eliot Shekhtman, and Kilian Q. Weinberger. 2023. Latent Diffusion for Language Generation. InProceedings of NeurIPS

    2023

  32. [33]

    Pengfei Luo, Xi Zhu, Tong Xu, Yi Zheng, and Enhong Chen. 2024. Semantic Interaction Matching Network for Few-Shot Knowledge Graph Completion.ACM Trans. Web18, 2 (2024), 20:1–20:19. doi:10.1145/3589557

    work page doi:10.1145/3589557 2024

  33. [34]

    Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, and Xindong Wu

  34. [35]

    IEEE Transactions on Knowledge and Data Engineering36, 7 (2024), 3580–3599

    Unifying Large Language Models and Knowledge Graphs: A Roadmap. IEEE Transactions on Knowledge and Data Engineering36, 7 (2024), 3580–3599. doi:10.1109/TKDE.2024.3352100

    doi:10.1109/tkde.2024.3352100 2024

  35. [36]

    Duo Peng, Qiuhong Ke, Arulmurugan Ambikapathi, Yasin Yazici, Yinjie Lei, and Jun Liu. 2024. Unsupervised Domain Adaptation via Domain-Adaptive Diffusion. IEEE Trans. Image Process.33 (2024), 4245–4260. doi:10.1109/TIP.2024.3424985

    work page doi:10.1109/tip.2024.3424985 2024

  36. [37]

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe Penna, and Robin Rombach. 2024. SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis. InProceedings of ICLR

    2024

  37. [38]

    Aimon Rahman, Jeya Maria Jose Valanarasu, Ilker Hacihaliloglu, and Vishal M. Patel. 2023. Ambiguous Medical Image Segmentation Using Diffusion Models. In Proceedings of CVPR. IEEE, 11536–11546. doi:10.1109/CVPR52729.2023.01110

    doi:10.1109/cvpr52729.2023.01110 2023

  38. [39]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-Resolution Image Synthesis with Latent Diffusion Models. InProceedings of CVPR. 10674–10685. doi:10.1109/CVPR52688.2022.01042

    doi:10.1109/cvpr52688.2022.01042 2022

  39. [40]

    Michael Schlichtkrull, Thomas N Kipf, Peter Bloem, Rianne Van Den Berg, Ivan Titov, and Max Welling. 2018. Modeling relational data with graph convolutional networks. InEuropean semantic web conference. Springer, 593–607

    2018

  40. [41]

    Chao Shang, Yun Tang, Jing Huang, Jinbo Bi, Xiaodong He, and Bowen Zhou

  41. [42]

    InProceedings of AAAI, Vol

    End-to-end structure-aware convolutional networks for knowledge base completion. InProceedings of AAAI, Vol. 33. 3060–3067

  42. [43]

    Jiawei Sheng, Shu Guo, Zhenyu Chen, Juwei Yue, Lihong Wang, Tingwen Liu, and Hongbo Xu. 2020. Adaptive Attentional Network for Few-Shot Knowledge Graph Completion. InProceedings of EMNLP. Association for Computational Linguistics, 1681–1691. doi:10.18653/V1/2020.EMNLP-MAIN.131

    work page doi:10.18653/v1/2020.emnlp-main.131 2020

  43. [44]

    Jiawei Sheng, Taoyu Su, Weiyi Yang, Linghui Wang, Yongxiu Xu, and Tingwen Liu

  44. [45]

    InProceedings of AAAI

    Information-Theoretic Minimal Sufficient Representation for Multi-Domain Knowledge Graph Completion. InProceedings of AAAI

  45. [46]

    Harkanwar Singh, Prachi Jain, Soumen Chakrabarti, et al . 2021. Multilingual Knowledge Graph Completion with Joint Relation and Entity Alignment.arXiv preprint arXiv:2104.08804(2021)

    Pith/arXiv arXiv 2021

  46. [47]

    Weiss, Niru Maheswaranathan, and Surya Ganguli

    Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, and Surya Ganguli

  47. [48]

    In Proceedings of ICML, Vol

    Deep Unsupervised Learning using Nonequilibrium Thermodynamics. In Proceedings of ICML, Vol. 37. 2256–2265

  48. [49]

    Zhiqing Sun, Zhi-Hong Deng, Jian-Yun Nie, and Jian Tang. 2019. RotatE: Knowl- edge Graph Embedding by Relational Rotation in Complex Space. InProceedings of ICLR

    2019

  49. [50]

    Zequn Sun, Wei Hu, Qingheng Zhang, and Yuzhong Qu. 2018. Bootstrapping Entity Alignment with Knowledge Graph Embedding. InProceedings of IJCAI. 4396–4402. doi:10.24963/IJCAI.2018/611

    doi:10.24963/ijcai.2018/611 2018

  50. [51]

    Zequn Sun, Jiacheng Huang, Xiaozhou Xu, Qijin Chen, Weijun Ren, and Wei Hu

  51. [52]

    InProceedings of IMCL

    What Makes Entities Similar? A Similarity Flooding Perspective for Multi- sourced Knowledge Graph Embeddings. InProceedings of IMCL. 32875–32885. https://proceedings.mlr.press/v202/sun23d.html

  52. [53]

    Rongchuan Tang, Yang Zhao, Chengqing Zong, and Yu Zhou. 2023. Multilingual Knowledge Graph Completion with Language-Sensitive Multi-Graph Attention. InProceedings of ACL. 10508–10519. doi:10.18653/V1/2023.ACL-LONG.586

    doi:10.18653/v1/2023.acl-long.586 2023

  53. [54]

    Théo Trouillon, Johannes Welbl, Sebastian Riedel, Éric Gaussier, and Guillaume Bouchard. 2016. Complex embeddings for simple link prediction. InProceedings WWW ’26, April 13–17, 2026, Dubai, United Arab Emirates Jiawei Sheng et al. of ICML. PMLR, 2071–2080

    2016

  54. [55]

    Quan Wang, Zhendong Mao, Bin Wang, and Li Guo. 2017. Knowledge Graph Embedding: A Survey of Approaches and Applications.IEEE Transactions on Knowledge and Data Engineering29, 12 (2017), 2724–2743. doi:10.1109/TKDE. 2017.2754499

    doi:10.1109/tkde 2017

  55. [56]

    Wenjie Wang, Yiyan Xu, Fuli Feng, Xinyu Lin, Xiangnan He, and Tat-Seng Chua

  56. [57]

    InProceedings of SIGIR

    Diffusion Recommender Model. InProceedings of SIGIR. ACM, 832–841

  57. [58]

    Xiaozhi Wang, Tianyu Gao, Zhaocheng Zhu, Zhengyan Zhang, Zhiyuan Liu, Juanzi Li, and Jian Tang. 2021. KEPLER: A Unified Model for Knowledge Em- bedding and Pre-trained Language Representation.TACL9 (2021), 176–194. doi:10.1162/TACL_A_00360

    doi:10.1162/tacl_a_00360 2021

  58. [59]

    Yanbin Wei, Qiushi Huang, Yu Zhang, and James T. Kwok. 2023. KICGPT: Large Language Model with Knowledge in Context for Knowledge Graph Completion. InFindings of EMNLP. 8667–8683. doi:10.18653/V1/2023.FINDINGS-EMNLP.580

    doi:10.18653/v1/2023.findings-emnlp.580 2023

  59. [60]

    Ruobing Xie, Zhiyuan Liu, Jia Jia, Huanbo Luan, and Maosong Sun. 2016. Rep- resentation Learning of Knowledge Graphs with Entity Descriptions. InPro- ceedings of AAAI, Dale Schuurmans and Michael P. Wellman (Eds.). 2659–2665. doi:10.1609/AAAI.V30I1.10329

    work page doi:10.1609/aaai.v30i1.10329 2016

  60. [61]

    Bishan Yang, Wen-tau Yih, Xiaodong He, Jianfeng Gao, and Li Deng. 2015. Em- bedding entities and relations for learning and inference in knowledge bases. In Proceedings of ICLR

    2015

  61. [62]

    Linyan Yang, Shiqiao Zhou, Jingwei Cheng, Fu Zhang, Jizheng Wan, Shuo Wang, and Mark Lee. 2025. DAEA: Enhancing Entity Alignment in Real-World Knowl- edge Graphs Through Multi-Source Domain Adaptation. InProceedings of COL- ING. 5890–5901. https://aclanthology.org/2025.coling-main.393/

    2025

  62. [63]

    Yaming Yang, Zhuofeng Luo, Zhe Wang, Weigang Lu, Yiheng Lu, Ziyu Guan, Wei Zhao, and Yuanhai Lv. 2025. A Translation-Based Heterogeneous Graph Neural Network for Multiple Knowledge Graphs Alignment. InProceedings of ICDE. 2215–2226

    2025

  63. [64]

    Zhengyi Yang, Jiancan Wu, Zhicai Wang, Xiang Wang, Yancheng Yuan, and Xiangnan He. 2023. Generate What You Prefer: Reshap- ing Sequential Recommendation via Guided Diffusion. InProceed- ings of NeurIPS. http://papers.nips.cc/paper_files/paper/2023/hash/ 4c5e2bcbf21bdf40d75fddad0bd43dc9-Abstract-Conference.html

    2023

  64. [65]

    Liang Yao, Chengsheng Mao, and Yuan Luo. 2020. KG-BERT: BERT for Knowledge Graph Completion.Proceedings of AAAI(2020)

    2020

  65. [66]

    Liang Yao, Jiazhen Peng, Chengsheng Mao, and Yuan Luo. 2025. Exploring Large Language Models for Knowledge Graph Completion. InProceedings of IEEE ICASSP. IEEE, 1–5. doi:10.1109/ICASSP49660.2025.10889242

    doi:10.1109/icassp49660.2025.10889242 2025

  66. [67]

    Huanyu Zhang, Xiaoxuan Shen, Baolin Yi, Jianfang Liu, and Yinao Xie. 2025. A Plug-in Critiquing Approach for Knowledge Graph Recommendation Systems via Representative Sampling. InProceedings of Web Conference. ACM, 322–333. doi:10.1145/3696410.3714808

    doi:10.1145/3696410.3714808 2025

  67. [68]

    Qingheng Zhang, Zequn Sun, Wei Hu, Muhao Chen, Lingbing Guo, and Yuzhong Qu. 2019. Multi-view Knowledge Graph Embedding for Entity Alignment. In Proceedings of IJCAI, Sarit Kraus (Ed.). 5429–5435. doi:10.24963/IJCAI.2019/754

    work page doi:10.24963/ijcai.2019/754 2019

  68. [69]

    Yichi Zhang, Zhuo Chen, Lingbing Guo, Yajing Xu, Wen Zhang, and Huajun Chen. 2024. Making Large Language Models Perform Better in Knowledge Graph Completion. InProceedings of ACM MM. 233–242. doi:10.1145/3664647.3681327

    work page doi:10.1145/3664647.3681327 2024

  69. [70]

    Yichi Zhang, Binbin Hu, Zhuo Chen, Lingbing Guo, Ziqi Liu, Zhiqiang Zhang, Lei Liang, Huajun Chen, and Wen Zhang. 2024. Multi-domain Knowledge Graph Collaborative Pre-training and Prompt Tuning for Diverse Downstream Tasks. CoRRabs/2405.13085 (2024). arXiv:2405.13085 doi:10.48550/ARXIV.2405.13085

    Pith/arXiv arXiv doi:10.48550/arxiv.2405.13085 2024

  70. [71]

    Qi Zhu, Hao Wei, Bunyamin Sisman, Da Zheng, Christos Faloutsos, Xin Luna Dong, and Jiawei Han. 2020. Collective Multi-Type Entity Alignment Between Knowledge Graphs. InProceedings of Web Conference

    2020

  71. [72]

    Yao Zhu, Hongzhi Liu, Zhonghai Wu, and Yingpeng Du. 2021. Relation-Aware Neighborhood Matching Model for Entity Alignment. InProceedings of AAAI. 4749–4756. doi:10.1609/AAAI.V35I5.16606 A Appendix In this appendix, we provide: (i) the dataset details, (ii) baselines, (iii) the evaluation protocol, and (iv) optimal hyper-parameters. A.1 Dataset Details For...

    work page doi:10.1609/aaai.v35i5.16606 2021