Pith. sign in

REVIEW 3 major objections 5 minor 69 references

From Symbolic to Neural and Back: Exploring Knowledge Graph-Large Language Model Synergies

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read The paper maps knowledge graph–LLM work into three integration categories and argues that scalability, efficiency, and data quality are the under-served dimensions.

desk verdict Competent but incremental survey whose headline comparative claim is unsupported by its own table and a placeholder reference. read the letter →

arxiv 2506.09566 v1 pith:3E5CPUAB submitted 2025-06-11 cs.CL cs.AIcs.LG

classification cs.CLcs.AIcs.LG
keywords knowledgegraphslargelanguagemodelsKG-enhancedLLMsLLM-augmentedKGsneuro-symbolicintegrationhallucinationmitigationsurveyscalability
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

This survey tries to give the growing field of knowledge graph (KG) and large language model (LLM) integration a usable map. It argues that the two technologies are complementary: KGs can ground LLM outputs in verifiable facts and reduce hallucination, while LLMs can automate the construction, completion, and querying of KGs. The paper's organizing claim is that existing work falls into three categories—KG-enhanced LLMs, LLM-augmented KGs, and joint bidirectional frameworks—and that previous surveys have underplayed scalability, computational efficiency, and data quality. A sympathetic reader would care because the map makes it possible to see where the field's actual open problems lie: dynamic knowledge updating, neuro-symbolic reasoning, and trustworthy automated fact extraction.

What carries the argument

The organizing device is a three-way taxonomy: KG-enhanced LLMs, LLM-augmented KGs, and joint LLM–KG synergy, each defined by the direction of information flow between graph and model. A coverage table comparing four earlier surveys across ten topics supplies the machinery for the gap claim, since the table is what lets the paper assert that scalability, efficiency, and data quality have been under-emphasized. The taxonomy does the work of grouping roughly seventy surveyed references into a coherent picture and turning the field's fragmentation into named open problems.

What would settle it

A systematic reading of the four cited surveys that finds any of them treating scalability, computational efficiency, or data quality in a substantive section would undercut the paper's distinctiveness claim.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central finding is that KG–LLM research is best understood as a three-part landscape rather than a single technique. In the first part, knowledge graphs serve language models as external sources of grounded facts, improving reasoning and reducing hallucinations; in the second, language models serve graphs as extraction, completion, and interface tools; in the third and most recent part, the two are coupled in bidirectional loops where each side constrains and updates the other. The paper further claims that this landscape has been surveyed before, but never with the same emphasis on how these systems scale, how much computation they cost, and whether the knowledge they produce is of high quality. Its distinctive contribution, as the paper states it, is to foreground those three dimensions and to translate them into an open-problem list for future research.

Load-bearing premise

The survey's claim to a unique emphasis depends on its Table 3 checklist accurately and fairly capturing what the four earlier surveys cover.

Editorial extensions

If this is right

  • The three-way taxonomy gives the field a shared vocabulary for positioning new work as KG-enhanced, LLM-augmented, or jointly coupled.
  • If the gap analysis is right, future evaluations of KG–LLM systems should measure scalability, computational cost, and knowledge quality alongside task accuracy.
  • The open-problem list makes dynamic knowledge updating and human-in-the-loop validation central targets rather than side concerns.
  • The joint frameworks reviewed in the paper imply that the next generation of systems will treat the knowledge graph as part of the reasoning process, not merely as a data source.

Reading between the lines

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

  • Extending the paper's coverage comparison into a full coding study, with explicit rules for what counts as covering a topic and two independent raters, would test whether the claimed 'unique emphasis' survives a more rigorous measurement.
  • The paper's concerns connect two of its open problems: hallucination mitigation and dynamic KG updating are largely the same problem of keeping a graph trustworthy while it changes over time.
  • A practical heuristic that follows from the survey is that LLM-built knowledge graphs should be audited for reliability separately from the quality of the LLM's language output, since errors can propagate silently into the graph.
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

3 major / 5 minor

Summary. This paper surveys the integration of Knowledge Graphs (KGs) and Large Language Models (LLMs), organizing the field into three categories: KG-enhanced LLMs, LLM-augmented KGs, and joint LLM–KG frameworks. It reviews background material on KGs and LLMs, provides a comparative table of coverage in four prior surveys (Table 3), and concludes with a list of open problems. The paper's stated contribution is a 'systematic' examination that 'uniquely emphasizes scalability, computational efficiency, and data quality' relative to prior surveys.

Significance. The descriptive survey sections and the proposed taxonomy are useful as an entry point for newcomers: the separation into KG-enhanced LLMs, LLM-augmented KGs, and joint frameworks is sensible and broadly consistent with the cited literature, and the list of open problems in Section 6 is reasonable. The paper also states its scope clearly and gives concrete examples (e.g., ConceptNet triplets, open-source LLM table). However, the central comparative claim—the 'unique emphasis' on scalability, efficiency, and data quality—is not actually supported by the evidence presented, and one of the four comparison surveys is an unresolved placeholder. Because the distinctiveness claim is the load-bearing part of the abstract and the gap analysis, the paper currently does not substantiate its main contribution.

major comments (3)
  1. [Abstract and §4, Table 3] The claim that this survey 'uniquely emphasizes scalability, computational efficiency, and data quality' is not supported by Table 3. The table's ten rows are entity linking/alignment, relation/attribute extraction, KG embedding, KG completion, graph-to-text, KG QA, KG-enhanced LLMs, LLM-augmented KGs, synergized LLM+KG, and hallucination/factual accuracy. None of the three claimed differentiators appear as rows, so the table cannot show that Pan et al. (2024), Pan et al. (2023), Hu et al. (2023), or Yang et al. (2024) omit those dimensions. To make the claim load-bearing, the table must include rows for scalability, computational efficiency, and data quality, and the coverage judgments must be based on a documented coding procedure.
  2. [Bibliography, reference [62]] Reference [62] is an unresolved placeholder: it is listed as '[First Name] Yang and Others. Fact-aware generation in large language models' with the URL https://example.org/yang2024factaware. This reference is used as the fourth column of Table 3 and is cited in §4.1 for KG-enhanced generation, so the coverage comparison for that column cannot be checked against an identifiable published work. The authors must cite the actual survey (with complete author list, venue, year, and a verifiable DOI/arXiv identifier) or remove it from the comparison.
  3. [§4 (methodology for Table 3)] The paper does not describe a systematic literature search or selection protocol: there is no statement of databases queried, inclusion/exclusion criteria, time window, or how the four surveys in Table 3 were chosen. There is also no coding protocol for assigning ✓, ×, or –. The sentence 'We consider as non-applicable the topics which either weren't yet relevant at the time of writing a paper, or intentionally focus on a different topic' is a post-hoc rationalization rather than a reproducible rule. Without such a protocol, the coverage judgments in Table 3 are subjective, and the 'systematic' characterization in the abstract is not justified. A short methodology subsection should be added.
minor comments (5)
  1. [§4.2] The sentence 'However, one needs to be considered when using LLMs to populate KGs' is incomplete; it should read 'one needs to be careful' or similar.
  2. [§4.2] The sentence 'Furthermore,KG completiontasks (predicting missing links or attributes in an existing graph), LLMs can leverage their broad world knowledge' has a grammatical break; the parenthetical is not integrated into the sentence structure.
  3. [Bibliography] References [8] and [9] are identical (both Brown et al., 'Language models are few-shot learners', NeurIPS 2020). One duplicate should be removed and citations renumbered.
  4. [Table 2 and reference [16]] The reference for DeepSeek-V3 [16] is a GitHub URL with the description 'Scaling open-source llms via efficient retrieval', which does not match the model's actual technical report. Verify the reference and update it to the official citation.
  5. [General] The phrase 'random sample of triplets' in Table 1 should clarify the sampling procedure; as written, it is not reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the survey's taxonomy and comparative claims are external summaries, with one non-load-bearing self-citation.

full rationale

This paper is a survey and taxonomy of existing work on Knowledge Graph and Large Language Model integration. It contains no fitted parameters, no equations, and no predictive claim that could reduce to its own inputs by construction. The central organizational claim, namely the division into KG-enhanced LLMs, LLM-augmented KGs, and joint LLM-KG frameworks, is a classification of externally cited systems rather than a derivation from those systems. The abstract's claim of unique emphasis on scalability, computational efficiency, and data quality rests on Table 3, which is difficult to verify because reference [62] is an unresolved placeholder with an example.org URL and the table does not include rows for the three claimed dimensions. That is a support and verifiability weakness, not circularity, because the claimed emphasis is not defined in terms of the table or the cited surveys. The only author-overlapping citation, reference [29], is used illustratively in the sentence about knowledge graph embeddings enabling tasks such as text classification, and it is not load-bearing for any conclusion of the survey. No step exhibits self-definition, fitted-input-called-prediction, a load-bearing self-citation chain, an imported uniqueness theorem, an ansatz smuggled in via citation, or a renamed known result. The survey content is self-contained in the sense that its statements about specific systems are grounded in the cited literature, and none of those statements is forced by the paper's own definitions.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The survey contributes no fitted values or new entities. Its conclusions rest on three domain assumptions: the adequacy of the taxonomy, the accuracy of the coverage table, and the reliability of the cited benefits. The most fragile is the Table 3 comparison, because the paper's claimed uniqueness and gap analysis follow from it.

assumptions (3)
  • domain assumption The three-category taxonomy (KG-enhanced LLMs, LLM-augmented KGs, joint LLM-KG frameworks) is a faithful and useful organization of the field.
    Adopted in Section 4 and Figure 2 without a formal derivation or a systematic comparison to alternative taxonomies; the paper's own Table 3 shows overlap with existing surveys.
  • domain assumption The coverage judgments in Table 3, which mark which topics are covered by four prior surveys, are accurate.
    The table is asserted without a coding protocol or evidence; the paper's 'unique emphasis' claim depends on this comparison.
  • domain assumption Cited benefits such as hallucination reduction and improved reasoning from KG-LLM integration are taken as established by the cited literature.
    Repeated in the abstract and Section 4.1 as motivation; the survey performs no independent evaluation of these benefits.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Symbolic to Neural and Back: Exploring Knowledge Graph-Large Language Model Synergies." pith.science (2026). https://pith.science/paper/3E5CPUAB

@misc{pith2026250609566,
  author       = {Pith},
  title        = {Pith review of: From Symbolic to Neural and Back: Exploring Knowledge Graph-Large Language Model Synergies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3E5CPUAB}},
  note         = {Machine review of arXiv:2506.09566}
}
read the original abstract

Integrating structured knowledge from Knowledge Graphs (KGs) into Large Language Models (LLMs) enhances factual grounding and reasoning capabilities. This survey paper systematically examines the synergy between KGs and LLMs, categorizing existing approaches into two main groups: KG-enhanced LLMs, which improve reasoning, reduce hallucinations, and enable complex question answering; and LLM-augmented KGs, which facilitate KG construction, completion, and querying. Through comprehensive analysis, we identify critical gaps and highlight the mutual benefits of structured knowledge integration. Compared to existing surveys, our study uniquely emphasizes scalability, computational efficiency, and data quality. Finally, we propose future research directions, including neuro-symbolic integration, dynamic KG updating, data reliability, and ethical considerations, paving the way for intelligent systems capable of managing more complex real-world knowledge tasks.

Figures

Figures reproduced from arXiv: 2506.09566 by the authors.

Figure 1
Figure 1. Scope of this paper, illustrating the evolution from Knowledge Graphs to LLMs, their synergy, and directions for future work. random sample of triplets related to the concept “mountain”, taken from the ConceptNet [54] knowledge graph. Relationship extraction focuses on identifying and categorizing the rela￾tionships between entities, such as “works at” or “located in”. This process is essential for building the conn… view at source ↗
Figure 2
Figure 2. The interplay between Large Language Models (LLMs) and Knowledge Graphs (KGs) [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

69 extracted references · 55 canonical work pages

  1. [62]

    Fact-aware generation in large language models.https://example.org/yang2024factaware, 2024

    [First Name] Yang and Others. Fact-aware generation in large language models.https://example.org/yang2024factaware, 2024. Extended ver- sion available online

  2. [1]

    Llama3.1:A405b-parameteropen-sourcelanguagemodel.https: //github.com/facebookresearch/llama, 2024

    MetaAI. Llama3.1:A405b-parameteropen-sourcelanguagemodel.https: //github.com/facebookresearch/llama, 2024. Community license

  3. [2]

    Mixtral 8x22b: Sparse mixture-of-experts model with 141b parameters.https://github.com/mistralai/Mixtral, 2024

    Mistral AI. Mixtral 8x22b: Sparse mixture-of-experts model with 141b parameters.https://github.com/mistralai/Mixtral, 2024. Apache 2.0 License

  4. [3]

    Language model guided knowledge graph embeddings.IEEE Access, 10:76008–76020, 2022

    Mirza Mohtashim Alam, Md Rashad Al Hasan Rony, Mojtaba Nayy- eri, Karishma Mohiuddin, MST Mahfuja Akter, Sahar Vahdati, and Jens Lehmann. Language model guided knowledge graph embeddings.IEEE Access, 10:76008–76020, 2022

  5. [4]

    Neural machine translation by jointly learning to align and translate

    Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. Neural machine translation by jointly learning to align and translate. In Yoshua Bengio and Yann LeCun, editors,3rd International Conference on Learning Repre- sentations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, 2015

  6. [5]

    SciBERT: A pretrained language model for scientific text

    Iz Beltagy, Kyle Lo, and Arman Cohan. SciBERT: A pretrained language model for scientific text. In Kentaro Inui, Jing Jiang, Vincent Ng, and Xi- aojun Wan, editors,Proceedings of the 2019 Conference on Empirical Meth- ods in Natural Language Processing and the 9th International Joint Confer- ence on Natural Language Processing (EMNLP-IJCNLP), pages 3615–3...

  7. [6]

    Courville, and Pascal Vincent

    Yoshua Bengio, Aaron C. Courville, and Pascal Vincent. Representation learning: A review and new perspectives.IEEE Transactions on Pattern Analysis and Machine Intelligence, 35(8):1798–1828, 2013

  8. [7]

    On the opportunities and risks of foundation models

    RishiBommasani,DrewHudson,EhsanAdeli,RussAltman,SanjeevArora, Simone von Arx, Michael S Bernstein, Jeannette Bohg, Rob Brekelmans, Kyle Burns, et al. On the opportunities and risks of foundation models. ArXiv preprint, abs/2108.07258, 2021

Show all 69 references
  1. [8]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Ka- plan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jef...

  2. [9]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Ka- plan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Exploring KG–LLM Synergies 15 Tom Henighan, Rewon Child, Aditya ...

  3. [10]

    PairRE: Knowl- edge graph embeddings via paired relation vectors

    Linlin Chao, Jianshan He, Taifeng Wang, and Wei Chu. PairRE: Knowl- edge graph embeddings via paired relation vectors. In Chengqing Zong, Fei Xia, Wenjie Li, and Roberto Navigli, editors,Proceedings of the 59th An- nual Meeting of the Association for Computational Linguistics ...

  4. [11]

    SAC-KG: Exploiting large language models as skilled automatic construc- torsfordomainknowledgegraph

    Hanzhu Chen, Xu Shen, Qitan Lv, Jie Wang, Xiaoqi Ni, and Jieping Ye. SAC-KG: Exploiting large language models as skilled automatic construc- torsfordomainknowledgegraph. InLun-WeiKu,AndreMartins,andVivek Srikumar, editors,Proceedings of the 62nd Annual Meeting of the Associa- ...

  5. [12]

    Augmenting multimodal llms with self-reflective tokens for knowledge-based visual question answering, 2025

    Federico Cocchi, Nicholas Moratelli, Marcella Cornia, Lorenzo Baraldi, and Rita Cucchiara. Augmenting multimodal llms with self-reflective tokens for knowledge-based visual question answering, 2025

  6. [13]

    Command r: A retrieval-augmented open source language model

    Cohere. Command r: A retrieval-augmented open source language model. https://github.com/cohere-ai/command-r, 2024. Open source for re- search

  7. [14]

    Command r+: Enhanced version for large-scale tasks.https: //github.com/cohere-ai/command-r-plus, 2024

    Cohere. Command r+: Enhanced version for large-scale tasks.https: //github.com/cohere-ai/command-r-plus, 2024. Open source, extended capabilities

  8. [15]

    Gemma: Lightweight llms in 2b and 7b sizes.https: //github.com/google-research/gemma, 2024

    Google DeepMind. Gemma: Lightweight llms in 2b and 7b sizes.https: //github.com/google-research/gemma, 2024. Open source

  9. [16]

    Deepseek-v3: Scaling open-source llms via efficient retrieval

    DeepSeek. Deepseek-v3: Scaling open-source llms via efficient retrieval. https://github.com/deepseek-ai/DeepSeek-V3, 2024. Open source

  10. [17]

    BERT: Pre-training of deep bidirectional transformers for language under- standing

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of deep bidirectional transformers for language under- standing. In Jill Burstein, Christy Doran, and Thamar Solorio, editors, Proceedings of the 2019 Conference of the North American Chapter ...

  11. [18]

    Core techniques of question answering systems over knowledge bases: A survey

    Dennis Diefenbach, Vanessa López, Kamal Singh, and Pierre Maret. Core techniques of question answering systems over knowledge bases: A survey. Knowledge and Information Systems, 55(3):529–569, 2018. 16 Škrlj et al

  12. [19]

    Knowledge vault: a web-scale approach to probabilistic knowledge fusion

    Xin Dong, Evgeniy Gabrilovich, Geremy Heitz, Wilko Horn, Ni Lao, Kevin Murphy, Thomas Strohmann, Shaohua Sun, and Wei Zhang. Knowledge vault: a web-scale approach to probabilistic knowledge fusion. In Sofus A. Macskassy, Claudia Perlich, Jure Leskovec, Wei Wang, and Rayid Ghan...

  13. [20]

    Enrich- ing contextualized language model from knowledge graph for biomedical information extraction.Briefings in bioinformatics, 22(3):bbaa110, 2021

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

  14. [21]

    Comprehensive analysis of knowledge graph embedding techniques benchmarked on link prediction.Electronics, 11(23):3866, 2022

    Ilaria Ferrari, Giacomo Frisoni, Paolo Italiani, Gianluca Moro, and Claudio Sartori. Comprehensive analysis of knowledge graph embedding techniques benchmarked on link prediction.Electronics, 11(23):3866, 2022

  15. [22]

    A survey on knowledge graph-based recommender sys- tems.IEEE Transactions on Knowledge and Data Engineering, 34(8):3549– 3568, 2022

    Qingyu Guo, Fuzhen Zhuang, Chuan Qin, Hengshu Zhu, Xing Xie, Hui Xiong, and Qing He. A survey on knowledge graph-based recommender sys- tems.IEEE Transactions on Knowledge and Data Engineering, 34(8):3549– 3568, 2022

  16. [23]

    Rashid, Anisa Rula, Lukas Schmelzeisen, Juan Sequeda, Steffen Staab, and Antoine Zimmermann.Knowledge Graphs

    Aidan Hogan, Eva Blomqvist, Michael Cochez, Claudia d’Amato, Gerard de Melo, Claudio Gutierrez, Jose Emilio Labra Gayo, Sabrina Kirrane, Sebastian Neumaier, Axel Polleres, Roberto Navigli, Axel-Cyrille Ngonga Ngomo, Sabbir M. Rashid, Anisa Rula, Lukas Schmelzeisen, Juan Sequed...

  17. [24]

    Knowledge graphs.ACM Computing Surveys (Csur), 54(4):1–37, 2021

    AidanHogan,EvaBlomqvist,MichaelCochez,Claudiad’Amato,GerardDe Melo, Claudio Gutierrez, Sabrina Kirrane, José Emilio Labra Gayo, Roberto Navigli, Sebastian Neumaier, et al. Knowledge graphs.ACM Computing Surveys (Csur), 54(4):1–37, 2021

  18. [25]

    A survey of knowledge-enhanced pre-trained language models.IEEE Trans- actions on Knowledge and Data Engineering, 36:1413–1430, 2023

    Linmei Hu, Zeyi Liu, Ziwang Zhao, Lei Hou, Liqiang Nie, and Juanzi Li. A survey of knowledge-enhanced pre-trained language models.IEEE Trans- actions on Knowledge and Data Engineering, 36:1413–1430, 2023

  19. [26]

    Can llms be good graph judger for knowledge graph construction?, 2025

    HaoyuHuang,ChongChen,ConghuiHe,YangLi,JiaweiJiang,andWentao Zhang. Can llms be good graph judger for knowledge graph construction?, 2025

  20. [27]

    Improving knowledge graph em- bedding using locally and globally attentive relation paths

    Ningning Jia, Xiang Cheng, and Sen Su. Improving knowledge graph em- bedding using locally and globally attentive relation paths. InAdvances in Information Retrieval (ECIR 2020), LNCS 12035, pages 17–32. Springer, 2020

  21. [28]

    Carte: pretraining and transfer for tabular learning

    Myung Jun Kim, Léo Grinsztajn, and Gaël Varoquaux. Carte: pretraining and transfer for tabular learning. InProceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org, 2024

  22. [29]

    Knowledge graph informed fake news classification via heterogeneous representation ensembles.Neurocomputing, 496:208–226, 2022

    Boshko Koloski, Timen Stepišnik Perdih, Marko Robnik-Šikonja, Senja Pol- lak, and Blaž Škrlj. Knowledge graph informed fake news classification via heterogeneous representation ensembles.Neurocomputing, 496:208–226, 2022. Exploring KG–LLM Synergies 17

  23. [30]

    Jamba: A production-grade hybrid llm.https://github.com/ ai21labs/Jamba, 2024

    AI21 Labs. Jamba: A production-grade hybrid llm.https://github.com/ ai21labs/Jamba, 2024. Open source under Apache 2.0

  24. [31]

    Decoding on Graphs: Faithful and Sound Reasoning on Knowl- edge Graphs through Generation of Well-Formed Chains.arXiv preprint arXiv:2410.18415, 2024

    Kun Li, Tianhua Zhang, Xixin Wu, Hongyin Luo, James Glass, and He- len Meng. Decoding on Graphs: Faithful and Sound Reasoning on Knowl- edge Graphs through Generation of Well-Formed Chains.arXiv preprint arXiv:2410.18415, 2024

  25. [32]

    TransE-MTP: A new representation learn- ing method for knowledge graph embedding with multi-translation princi- ples and TransE.Electronics, 13(16):3171, 2024

    Yongfang Li and Chunhua Zhu. TransE-MTP: A new representation learn- ing method for knowledge graph embedding with multi-translation princi- ples and TransE.Electronics, 13(16):3171, 2024

  26. [33]

    Dual Reason- ing: A GNN-LLM Collaborative Framework for Knowledge Graph Question Answering

    Guangyi Liu, Yongqi Zhang, Yong Li, and Quanming Yao. Dual Reason- ing: A GNN-LLM Collaborative Framework for Knowledge Graph Question Answering. InConference on Parsimony and Learning (CPAL), 2025

  27. [34]

    K-BERT: enabling language representation with knowl- edge graph

    Weijie Liu, Peng Zhou, Zhe Zhao, Zhiruo Wang, Qi Ju, Haotang Deng, and Ping Wang. K-BERT: enabling language representation with knowl- edge graph. InThe Thirty-Fourth AAAI Conference on Artificial Intelli- gence, AAAI 2020, The Thirty-Second Innovative Applications of Artifici...

  28. [35]

    Think-on-Graph 2.0: Deep and Interpretable Large Language Model Reasoning with Knowledge Graph-guided Retrieval.arXiv preprint arXiv:2407.10805, 2023

    Shengjie Ma, Chengjin Xu, Xuhui Jiang, Muzhi Li, Huaren Qu, and Jian Guo. Think-on-Graph 2.0: Deep and Interpretable Large Language Model Reasoning with Knowledge Graph-guided Retrieval.arXiv preprint arXiv:2407.10805, 2023

  29. [36]

    Tree-of-Traversals: A Zero-Shot Reasoning Algorithm for Augmenting Black-box Language Models with Knowledge Graphs.arXiv preprint arXiv:2407.21358, 2024

    Elan Markowitz, Anil Ramakrishna, Jwala Dhamala, Ninareh Mehrabi, Charith Peris, Rahul Gupta, Kai-Wei Chang, and Aram Galstyan. Tree-of-Traversals: A Zero-Shot Reasoning Algorithm for Augmenting Black-box Language Models with Knowledge Graphs.arXiv preprint arXiv:2407.21358, 2024

  30. [37]

    Phi-3 medium: A 14b-parameter language model.https:// github.com/microsoft/phi, 2024

    Microsoft. Phi-3 medium: A 14b-parameter language model.https:// github.com/microsoft/phi, 2024. Open source under MIT License

  31. [38]

    Corrado, and Jeffrey Dean

    Tomás Mikolov, Ilya Sutskever, Kai Chen, Gregory S. Corrado, and Jeffrey Dean. Distributed representations of words and phrases and their composi- tionality. In Christopher J. C. Burges, Léon Bottou, Zoubin Ghahramani, and Kilian Q. Weinberger, editors,Advances in Neural Infor...

  32. [39]

    Dbrx: An open-source mixture-of-experts language model.https://github.com/databricks/DBRX, 2024

    Databricks Mosaic ML. Dbrx: An open-source mixture-of-experts language model.https://github.com/databricks/DBRX, 2024. Open source

  33. [40]

    Nemotron-4: A 340b-parameter open source language model

    Nvidia. Nemotron-4: A 340b-parameter open source language model. https://github.com/nvidia/Nemotron-4, 2024. Apache 2.0 License

  34. [41]

    Jeff Z. Pan, Simon Razniewski, Jan-Christoph Kalo, Sneha Singhania, Jiaoyan Chen, Stefan Dietze, Hajira Jabeen, Janna Omeliyanenko, Wen Zhang, Matteo Lissandrini, Russa Biswas, Gerard de Melo, Angela Boni- fati, Edlira Vakaj, Mauro Dragoni, and Damien Graux. Large language 18 ...

  35. [42]

    Unifying Large Language Models and Knowledge Graphs: A Roadmap

    Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, and Xindong Wu. Unifying Large Language Models and Knowledge Graphs: A Roadmap. arXiv preprint arXiv:2306.08302, 2023

  36. [43]

    Unifying large language models and knowledge graphs: A roadmap

    Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, and Xindong Wu. Unifying large language models and knowledge graphs: A roadmap. IEEE Transactions on Knowledge and Data Engineering, 36(7):3580–3599, 2024

  37. [44]

    Knowl- edge graphs: Opportunities and challenges.Artificial Intelligence Review, 56(11):13071–13102, 2023

    CiyuanPeng,FengXia,MehdiNaseriparsa,andFrancescoOsborne. Knowl- edge graphs: Opportunities and challenges.Artificial Intelligence Review, 56(11):13071–13102, 2023

  38. [45]

    GloVe: Global vectors for word representation

    Jeffrey Pennington, Richard Socher, and Christopher Manning. GloVe: Global vectors for word representation. In Alessandro Moschitti, Bo Pang, and Walter Daelemans, editors,Proceedings of the 2014 Conference on Em- pirical Methods in Natural Language Processing (EMNLP), pages 1...

  39. [46]

    Peters, M

    Matthew E. Peters, M. Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. Knowbert: Knowledge-enhanced contextualized word representations. InProceedings of EMNLP 2019, 2019

  40. [47]

    Fabio Petroni, Tim Rocktäschel, Sebastian Riedel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and Alexander Miller. Language models as knowl- edge bases? In Kentaro Inui, Jing Jiang, Vincent Ng, and Xiaojun Wan, editors,Proceedings of the 2019 Conference on Empirical Methods in ...

  41. [48]

    Ad-hoc object retrieval in the web of data

    Jeffrey Pound, Peter Mika, and Hugo Zaragoza. Ad-hoc object retrieval in the web of data. In Michael Rappa, Paul Jones, Juliana Freire, and Soumen Chakrabarti, editors,Proceedings of the 19th International Conference on World Wide Web, WWW 2010, Raleigh, North Carolina, USA, A...

  42. [49]

    Language models are unsupervised multitask learners

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. Language models are unsupervised multitask learners. Tech- nical report, OpenAI Technical Report, 2019

  43. [50]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text transformer.J. Mach. Learn. Res., 21:140:1–140:67, 2020

  44. [51]

    You CAN teach an old dog new tricks! on training knowledge graph embeddings

    Daniel Ruffinelli, Samuel Broscheit, and Rainer Gemulla. You CAN teach an old dog new tricks! on training knowledge graph embeddings. In8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020. OpenReview.net, 2020

  45. [52]

    Synthesis Lectures on Data, Semantics, and Knowledge

    Juan Sequeda and Ora Lassila.Designing and Building Enterprise Knowl- edge Graphs. Synthesis Lectures on Data, Semantics, and Knowledge. Mor- gan & Claypool Publishers, 2021. Exploring KG–LLM Synergies 19

  46. [53]

    Neural entity linking: A survey of models based on deep learning.Semantic Web, 13(3):527–570, 2022

    Özge Sevgili, Artem Shelmanov, Mikhail Arkhipov, Alexander Panchenko, and Chris Biemann. Neural entity linking: A survey of models based on deep learning.Semantic Web, 13(3):527–570, 2022

  47. [54]

    Conceptnet 5.5: An open multilingual graph of general knowledge

    Robyn Speer, Joshua Chin, and Catherine Havasi. Conceptnet 5.5: An open multilingual graph of general knowledge. In Satinder P. Singh and Shaul Markovitch, editors,Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence, February 4-9, 2017, San Francisco, Ca...

  48. [55]

    Ernie: Enhanced representation through knowledge integration

    Yu Sun, Shuohuan Wang, Yukun Li, et al. Ernie: Enhanced representation through knowledge integration. InAAAI, 2019

  49. [56]

    Lamda: Language models for dialog applications

    Romal Thoppilan et al. Lamda: Language models for dialog applications. ArXiv preprint, abs/2201.08239, 2022

  50. [57]

    Knowledge graph validation by integrating llms and human-in-the-loop.In- formation Processing & Management, 62(5):104145, 2025

    Stefani Tsaneva, Danilo Dessì, Francesco Osborne, and Marta Sabou. Knowledge graph validation by integrating llms and human-in-the-loop.In- formation Processing & Management, 62(5):104145, 2025

  51. [58]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need. In Isabelle Guyon, Ulrike von Luxburg, Samy Bengio, Hanna M. Wallach, Rob Fergus, S. V. N. Vishwanathan, and Roman Garnett, ...

  52. [59]

    Pan, Wen Zhang, and Huajun Chen

    Junjie Wang, Mingyang Chen, Binbin Hu, Dan Yang, Ziqi Liu, Yue Shen, Peng Wei, Zhiqiang Zhang, Jinjie Gu, Jun Zhou, Jeff Z. Pan, Wen Zhang, and Huajun Chen. Learning to Plan for Retrieval-Augmented Large Lan- guage Models from Knowledge Graphs.arXiv preprint arXiv:2406.14282, 2024

  53. [60]

    A survey on knowledge graph embeddings for link prediction.Symmetry, 13(3):485, 2021

    Meihong Wang, Linling Qiu, and Xiaoli Wang. A survey on knowledge graph embeddings for link prediction.Symmetry, 13(3):485, 2021

  54. [61]

    Mindmap: Prompting large language models with knowledge graphs for graph-of- thoughtreasoning

    Haoran Wen, Muhao Chen, Kai-Wei Chang, and Dan Roth. Mindmap: Prompting large language models with knowledge graphs for graph-of- thoughtreasoning. InProceedings of the 62nd Annual Meeting of the Associ- ation for Computational Linguistics (ACL). Association for Computational ...

  55. [63]

    Toward better drug discovery with knowledge graph.Current opinion in structural biology, 72:114–126, 2022

    Xiangxiang Zeng, Xinqi Tu, Yuansheng Liu, Xiangzheng Fu, and Yansen Su. Toward better drug discovery with knowledge graph.Current opinion in structural biology, 72:114–126, 2022

  56. [64]

    Kg-enhanced reasoning in large language models

    Hangfan Zhang, Zhimeng Guo, and Others. Kg-enhanced reasoning in large language models. InProceedings of ACL 2024, 2024

  57. [65]

    Making large language models perform better in knowledge graph completion.arXiv preprint arXiv:2310.06671, 2023

    Yichi Zhang, Zhuo Chen, Lingbing Guo, Yajing Xu, Wen Zhang, and Hua- jun Chen. Making large language models perform better in knowledge graph completion.arXiv preprint arXiv:2310.06671, 2023. 20 Škrlj et al

  58. [66]

    Chain-of-Knowledge: Integrating Knowledge Reasoning into Large Language Models by Learning from Knowledge Graphs.arXiv preprint arXiv:2407.00653, 2024

    Yifei Zhang, Xintao Wang, Jiaqing Liang, Sirui Xia, Lida Chen, and Yanghua Xiao. Chain-of-Knowledge: Integrating Knowledge Reasoning into Large Language Models by Learning from Knowledge Graphs.arXiv preprint arXiv:2407.00653, 2024

  59. [67]

    A survey of large language models.ArXiv preprint, abs/2303.18223, 2023

    Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, et al. A survey of large language models.ArXiv preprint, abs/2303.18223, 2023

  60. [68]

    A survey of knowl- edge graph construction using machine learning.Computer Modeling in Engineering & Sciences, 139(1):225–257, 2024

    Zhigang Zhao, Xiong Luo, Maojian Chen, and Ling Ma. A survey of knowl- edge graph construction using machine learning.Computer Modeling in Engineering & Sciences, 139(1):225–257, 2024

  61. [69]

    Llms for knowledge graph construction and reasoning: Recent capabilities and future opportunities

    Yuqi Zhu, Xiaohan Wang, Jing Chen, Shuofei Qiao, Yixin Ou, Yunzhi Yao, Shumin Deng, Huajun Chen, and Ningyu Zhang. Llms for knowledge graph construction and reasoning: Recent capabilities and future opportunities. arXiv preprint arXiv:2305.13168, 2023

Pith tools

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