KG and LLM

July 21, 2026 ยท View on GitHub

**My personal thoughts**:
1. As a data representation, graph structure is fundamentally an intuitive modeling approach that aligns with human cognition.
Nevertheless, it may not necessarily be scientifically optimal for modeling all types of world knowledge, especially when
compared with LLM which captures complex distributions through neural networks.
2. KGs, however, offer verified factual accuracy (primarily curated through crowdsourcing and human validation), thereby
avoiding the hallucination issues inherent in LLMs.
3. In the era of LLMs, these research directions may be more insightful.
(1) leveraging pre-existing high-precision knowledge graphs to enhance post-training and
fine-tuning of LLMs (particularly domain-specific LLMs) by integrating structured knowledge into their data preparation pipeline.
(2) leveraging pre-existing high-precision knowledge graphs to enhance the quality of inference, either by RAG or prompting.

๐Ÿ“ Overviews, Surveys, Tutorials and Summary

  1. Github Repo: KG-LLM-Papers [Link]
  2. Github Repo: Awesome LLM-KGs [Link]
  3. The Journey to A Knowledgeable Assistant with Retrieval-Augmented Generation (RAG) (WSDM 2024 and SIGMOD 2024 keynote, by Luna Dong in Meta) ๐ŸŒŸ
  4. LLM+KG Workshop, VLDB 2024 [Link]
  • Talk 1 โ€“ Integrating Knowledge Graph with Large Language Model: From the Perspective of Knowledge Engineering (Prof Qi)
  • Introduction of KG and LLM
  • KG for LLM (1) pretraining (2) as prompt (e.g., KAPING) (3) fine-tuning (4) inference (reduce the hallucination) (5) RAG (6) knowledge update or edit (7) knowledge fusion (8) knowledge validation (as benchmark) (9) future: learning of symbolic knowledge, benchmark, improve the interpretable reasoning of llm
  • LLM for KG (1) entity and relation extraction (in-context learning & SFT) (2) triple generation (3) ontology matching (e.g., olala) (4) entity alignment (e.g., chatea) (5) KBQA (e.g., ELLMKGQA framework) (6) ontology reasoning (data construction: subsumption checking + instance checking) (7) KG reasoning (kg embedding is enhanced, e.g., KoPA)
  • KG&LLM integration (1) knowledge service platform based on KG&LLM Integration (2) OpenKG + Tool + Application
  • The summary of the other talks. [My notes]
  1. Research Trends for the Interplay between Large Language Models and Knowledge Graphs (LLM+KG Workshop, VLDB 2024) [Paper] ๐ŸŒŸ
  2. 1st Workshop on Knowledge Graphs and Large Language Models (KaLLM 2024, from ACL) [Link]
  3. Knowledge Graphs Meet Multi-Modal Learning: A Comprehensive Survey [Paper]
  4. OneGraph: A KG constructed by LLM [Website] [Data Download]
  5. Unifying Large Language Models and Knowledge Graphs: A Roadmap (TKDE, July 2024) [Paper] ๐ŸŒŸ
  6. Knowledge-Empowered, Collaborative, and Co-Evolving AI Models: The Post-LLM Roadmap (Engineering 2024) [Paper]
  7. Integrating Large Language Models and Knowledge Graphs for Next-level AGI (WWW 2025 tutorial) [Paper]
  8. Towards Graph Foundation Models: A Survey and Beyond (WWW 2024 tutorial, PHILIP S. YU, CHUAN SHI) [Paper] [Slides of WWW 2024 tutorial] [Github]
  • Discusses (1). GNN-based Papers; (2). LLM-based Papers; (3). GNN+LLM-based Papers.
  1. Advancing Graph Representation Learning with Large Language Models: A Comprehensive Survey of Techniques (Arxiv 2024) [Paper]
  2. Give Us the Facts: Enhancing Large Language Models with Knowledge Graphs for Fact-aware Language Modeling (TKDE 2024) [Paper] ๐ŸŒŸ
  • KG in Before-training/During-training/Post-training enhancement for pretrained LMs. (see Table 2)
  1. Large Language Models on Graphs: A Comprehensive Survey (TKDE 2024, Jiawei Han) [Paper] ๐ŸŒŸ
  2. A Survey of Graph Meets Large Language Model: Progress and Future Directions (IJCAI 2024 Survey Track, Jeffery Xu Yu) [Paper]
  3. Graph learning and its advancements on large language models: A holistic survey (Neurocomputing 2025) [Paper]

๐Ÿ’ฌ Discussions about Knowledge Graphs and LLMs

  1. The Future of Knowledge Graphs in a World of Large Language Models, given by Denny Vrandeฤiฤ‡, Post-conference recording of the keynote for May 11 at the Knowledge Graph Conference 2023 in New York, NY. [Video]
  2. Bridging the Gap: Integrating Knowledge Graphs and Large Language Models (Arango DB) [Video]
  3. Uniting Large Language Models and Knowledge Graphs for Enhanced Knowledge Representation (Big Data LDN 2023) [Video]
  4. ChatGPTไธ‹็š„็Ÿฅ่ฏ†ๅ›พ่ฐฑๅฎก่ง†๏ผšไธ€ๆฌกๅ…ณไบŽๅฟ…็„ถๅฝฑๅ“ใ€ๆœชๆฅๆ–นๅ‘็š„่ฎจ่ฎบๅฎžๅฝ•ไธŽๆ€ป็ป“ [Link]
  5. Using Knowledge Graph data in Large Language Models (Keynote) [Video]

๐Ÿ“ Knowledge Graphs and LLMs

These papers mainly discuss or investigate the relationship between KG and LLM, e.g., how one affects the other one.
  1. Language Models as Knowledge Bases? (EMNLP 2019) [Paper]
  2. Language Models are Open Knowledge Graphs (ICLR 2021, Rejected) [Paper]
  3. JAKET: Joint Pre-training of Knowledge Graph and Language Understanding (AAAI 2022) [Paper]
  4. Language Models sounds the Death Knell of Knowledge Graphs (Arxiv 2023) [Paper]
  5. ChatGPT versus Traditional Question Answering for Knowledge Graphs: Current Status and Future Directions Towards Knowledge Graph Chatbots (Arxiv 2023) [Paper]
  • โญ Interesting!! Key point: ChatGPT is trained on information from different data sources. Thus, it does not generalize to unseen domain information.
  • Advanced KG Chatbots could be developed by incorporating capabilities of language models, such as ChatGPT, in question understanding, robustness, and explainability, with the outstanding capabilities of a QA system, such as KGQAN, in incorporating recent information and generality across different domains.
  1. Towards InnoGraph: A Knowledge Graph for AI Innovation (WWW 2023) [Paper]
  2. Augmented Language Models: a Survey, by MetaAI (Arxiv, Feb 2023) [Paper]
  • it is often necessary to process structured information when interacting with tools like a knowledge base.
  1. Let's Chat to Find the APIs: Connecting Human, LLM and Knowledge Graph through AI Chain (ASE 2023, Top Conf in Software Engineering) [Paper]
  2. Head-to-Tail: How Knowledgeable are Large Language Models (LLM)? A.K.A. Will LLMs Replace Knowledge Graphs? (NAACL 2024) [Paper] - Xin Luna Dong from Meta ๐Ÿ”ฅ
  3. Large Language Models as Reliable Knowledge Bases? (Arxiv, Jul 2024 - from University of Edinburgh and Huawei Edinburgh Research Center) [Paper]

This study defines criteria that a reliable LLM-as-KB should meet, focusing on factuality and consistency, and covering both seen and unseen knowledge.

  1. Llm-tikg: Threat intelligence knowledge graph construction utilizing large language model (Computers & Security, 2024) [Paper]
  2. Large Language Models and Knowledge Graphs: Opportunities and Challenges (Arxiv, Aug 2023) [Paper] [Notes (Chinese)]
  3. MKGL: Mastery of a Three-Word Language (NeurIPS 2024 Spotlight) [Paper]
  1. Main target: LLMs to carry out knowledge graph completion.
  2. They investigate the integration of LLMs with KGs by introducing a specialized KG Language (KGL), where a sentence precisely consists of an entity noun, a relation verb, and ends with another entity noun.
  3. Conclusions: (1) LLMs can achieve fluency in KGL, drastically reducing errors compared to conventional KG embedding methods on KG completion. (2) The enhanced LLM shows exceptional competence in generating accurate three-word sentences from an initial entity and interpreting new unseen terms out of KGs.
  1. Integrating Large Language Models and Knowledge Graphs for Next-level AGI (WWW 2025 tutorial) [Paper]
  2. OpenSPG, a knowledge graph engine developed by Ant Group in collaboration with OpenKG [Website] [Github] ๐Ÿ”ฅ
  • an open engine for knowledge graph designed and implemented on the basis of SPG framework, which provides explicit semantic representations, logical rule definitions, operator frameworks (construction, inference) and other capabilities for the domain knowledge graphs, and supports pluggable adaptation of basic engines and algorithmic services by various vendors to build customized solutions.

๐Ÿ“ LLMs for Knowledge Graphs

These papers discuss how to utilze LLM for the existing topics in KG, such as KG Constuction and Information Retrieval.
You can also find these papers marked with ๐Ÿ”ฅ in the other topic pages of this github repository.
  1. Enhancing Knowledge Graph Construction Using Large Language Models (Arxiv 2023) [Paper]
  2. Complex Logical Reasoning over Knowledge Graphs using Large Language Models (Arxiv 2023) [[Paper(https://arxiv.org/pdf/2305.01157)]
  3. Enhanced Story Comprehension for Large Language Models through Dynamic Document-Based Knowledge Graphs (AAAI 2022) [Paper]
  4. Fusing topology contexts and logical rules in language models for knowledge graph completion (Information Fusion 2022) [Paper]
  5. From Discrimination to Generation: Knowledge Graph Completion with Generative Transformer (WWW 2022) [Paper]
  6. Path Language Modeling over Knowledge Graphsfor Explainable Recommendation (WWW 2022) [Paper]
  7. GraphGPT [Github]
  8. CodeKGC: Code Language Model for Generative Knowledge Graph Construction (Arxiv, April 2023) [Paper]
  9. LLMs for Knowledge Graph Construction and Reasoning: Recent Capabilities and Future Opportunities (Arxiv, May 2023) [Paper]
  10. Think-on-Graph: Deep and Responsible Reasoning of Large Language Model on Knowledge Graph (ICLR 2024 poster) [Paper]
  • A new LLM-KG integrating paradigm which treats the LLM as an agent to interactively explore related entities and relations on KGs and perform reasoning based on the retrieved knowledge.
  • The authors of paper 11 (below) comments on the openreview panel.
  1. Donโ€™t Generate, Discriminate: A Proposal for Grounding Language Models to Real-World Environments (ACL 2023) [Paper]
  • Pangu consists of a symbolic agent and a neural LM working in a concerted fashion: The agent explores the environment to incrementally construct valid plans, and the LM evaluates the plausibility of the candidate plans to guide the search process.
  • The evaluation is conducted on KBQA.
  1. LLM-assisted Knowledge Graph Engineering: Experiments with ChatGPT (AIDRST 2023) [Paper]
  2. Making Large Language Models Perform Better in Knowledge Graph Completion (Arxiv 2023, but with 30+ citations) [Paper]
  3. Assessing the evolution of llm capabilities for knowledge graph engineering in 2023 (ESWC 2024) [Paper]
  4. Language Models-enhanced Semantic Topology Representation Learning For Temporal Knowledge Graph Extrapolation (CIKM 2024) [Paper]
  5. LLMs for knowledge graph construction and reasoning: recent capabilities and future opportunities (WWW journal 2024) [Paper]
  6. Collaborative System Synergizing Human Expertise and Large-scale Language Models for Legal Knowledge Graph Construction (CEUR-WS 2024) [Paper]
  7. SAC-KG: Exploiting Large Language Models as Skilled Automatic Constructors for Domain Knowledge Graph (ACL 2024) [Paper] [Notes (Chinese)]
  8. UrbanKGent: A Unified Large Language Model Agent Framework for Urban Knowledge Graph Construction (NeurIPS 2024) [Paper]
  9. A Prompt-Based Knowledge Graph Foundation Model for Universal In-Context Reasoning (NeurIPS 2024) [Paper]

LLM for KG reasoning

  1. Editing Language Model-based Knowledge Graph Embeddings (AAAI 2024) [Paper]

Update KG embedding for data freshness

  1. Language Models over Large-Scale Knowledge Base: on Capacity, Flexibility and Reasoning for New Facts (COLING 2025)
  2. Construction and Canonicalization of Economic Knowledge Graphs with LLMs (KGSWC 2024) [Link]
  3. Utilizing Language Models For Synthetic Knowledge Graph Generation (ICLR 2025 Workshop Data Problems) [Paper]
  4. SSKG Hub: An Expert-Guided Platform for LLM-Empowered Sustainability Standards Knowledge Graphs (arXiv 2026) [Paper] [Platform]
  • LLM extraction turns sustainability disclosure standards (GRI, SASB, TCFD, IFRS S2) into provenance-linked Draft KGs that domain experts review and promote to Certified KGs under role-based governance.
  1. Finetuning Generative Large Language Models with Discrimination Instructions for Knowledge Graph Completion (ISWC 2024) [Paper]
  2. CypherBench: Towards Precise Retrieval over Full-scale Modern Knowledge Graphs in the LLM Era (ACL 2025)
  3. Generating Domain-Specific Knowledge Graphs from Large Language Models (ACL 2025)
  4. Enriching contextualized language model from knowledge graph for biomedical information extraction (Briefings in Bioinformatics, KGE on BERT based LM) [Paper]
  5. Combining Self-Retrieval-Augmented Generation with Divide-and-Conquer for Language Model-based Knowledge Base Construction (KBC-LM and LM-KBC @ ISWC 2025) [Paper]
  6. KG4ESG: The ESG Knowledge Graph Atlas (Preprints 2026) [Paper]
  • An LLM-empowered knowledge graph atlas for the Environmental, Social, and Governance (ESG) domain, integrating ESG standards, frameworks, and entities.

๐Ÿ“ Knowledge Graphs for LLMs

These papers discuss how to utilze KG in the pre-training, post-training, SFT, inference and other stages of LLM.
Personally, KG for preparing training data (in pre-training, post-training, SFT) will be a more promising direction.

1. KG for Pretraining

1.1 KG for Contextual Word Representations

  1. Knowledge enhanced contextual word representations ((EMNLP-IJCNLP 2019)

1.2 KG as/for Training Corpus

  1. SKILL: Structured Knowledge Infusion for Large Language Models (NAACL 2022, by Google Research) [Paper]
  2. Knowledge Graph Based Synthetic Corpus Generation for Knowledge-Enhanced Language Model Pre-training (NAACL 2021) [Paper]
  3. Knowledge-Enhanced Image Captioning with Adaptive Graph-based Multimodal Alignment and LLM (AAAI 2026) [Paper]

2. KG for SFT

2.1 KG for generation of CoT (Chain of Thoughts) data

  1. MedReason: Eliciting Factual Medical Reasoning Steps in LLMs via Knowledge Graphs (Arxiv 2025, UCSC+UBC+NTU+Stanford+Cornell+NYU) [Paper]
  2. Graph Chain-of-Thought: Augmenting Large Language Models by Reasoning on Graphs (ACL 2024-Findings, UIUC+Amazon+Upenn+UV, Han Jiawei) [Paper]
  3. REASONING-ENHANCED HEALTHCARE PREDICTIONS WITH KNOWLEDGE GRAPH COMMUNITY RETRIEVAL (ICLR 2025, Han Jiawei) [Paper]
  4. Knowledge Graph Finetuning Enhances Knowledge Manipulation in Large Language Models (ICLR 2025 Poster, Jieping Ye) [Paper]
  5. KBLAM: KNOWLEDGE BASE AUGMENTED LANGUAGE MODEL [Paper] (ICLR 2025)

3. KG for Inference

The papers listed in this section may have overlaps with the KGQA papers.

3.1 KG for RAG (Retrieval Augmented Generation)

Since there are too many SOTA works, so we organize the results in a new page. [Link]

3.2 KG in Agent Memory (people may use "Memory Graph" instead of "Knowledge Graph")

Since there are too many SOTA works, so we organize the results in a new page. [Link]

3.3 KG and Prompting

  1. Knowledge Graph Prompting for Multi-Document Question Answering (AAAI 2024) [Paper]
  • KG enhances knowledge retrieval, where at query time an LLM-based agent dynamically traverses a graph with nodes representing document elements (e.g., passages, tables) and edges encoding lexical and semantical similarity or structural relationships.
  1. KG-prompt: Interpretable knowledge graph prompt for pre-trained language models (Knowledge-Based Systems, 2025) [Paper]
  2. KnowGPT: Knowledge Graph based Prompting for Large Language Models (NeurIPS 2024) [Paper]
  • 3 existing challenges for KG-enhanced LLMs๏ผšhuge search space, high API costs, and laborious prompt engineering.
  • KnowGPT enhance LLMs with domain knowledge. It contains a knowledge extraction module to extract the most informative knowledge from KGs, and a context-aware prompt construction module to automatically convert extracted knowledge into effective prompts.
  1. Knowledge-Augmented Language Model Prompting for Zero-Shot Knowledge Graph Question Answering (ACL 2023 Workshop on Matching Entities) [Paper] - RAG and Prompting
  2. G-Retriever: Retrieval-Augmented Generation for Textual Graph Understanding and Question Answering (NeurIPS 2024) [Paper] - RAG and Prompting
  3. Graph-ToolFormer: To Empower LLMs with Graph Reasoning Ability via Prompt Augmented by ChatGPT (Arxiv 2023, hot) [Paper]
  4. An Expert-In-The-Loop Design Utilising Knowledge Graphs to Prompt LLMs in Professional Writing (HCII 2026) [Paper]
  • Domain experts curate knowledge graphs that structure the prompts driving LLMs in professional writing workflows โ€” KG-based prompting with an expert in the loop.

3.4 KG as Grounded Truths/Benchmarks in Inference

Some of these works may be related to Retrieval-augmented language model (RALM), which relies on retrieved external knowledge to generate responses.
  1. Knowledge Graph-augmented Language Models for Complex Question Answering (NLRSE, ACL workshop) [Paper]
  2. FABULA: Intelligence Report Generation Using Retrieval-Augmented Narrative Construction (ASONAM 2023) [Paper]
  3. Mitigating Large Language Model Hallucinations via Autonomous Knowledge Graph-Based Retrofitting (AAAI 2024) [Paper]
  4. Conceptual Diagnostics for Knowledge Graphs and Large Language Models (ACL 2025, emerging industry track submission) [Paper]

3.5 KG for Reasoning/CoT in Inference

  1. GIVE: Structured Reasoning of Large Language Models with Knowledge Graph Inspired Veracity Extrapolation (ICML 2025) [Paper]
  2. LightPROF: A Lightweight Reasoning Framework for Large Language Model on Knowledge Graph (AAAI 2025) [Paper] [Discussion (Chinese)]
  3. Graph-constrained Reasoning: Faithful Reasoning on Knowledge Graphs with Large Language Models (ICML 2025) [Paper] [Discussion (Chinese)]
  4. GIVE: Structured Reasoning of Large Language Models with Knowledge Graph Inspired Veracity Extrapolation (ICML 2025) [Paper]
  5. Graph-constrained Reasoning: Faithful Reasoning on Knowledge Graphs with Large Language Models (ICML 2025) [Paper]

3.6 KG as other roles in Inference

  1. KBLaM: Knowledge Base augmented Language Model (ICLR 2025, Poster) [Paper]
  2. Barack's Wife Hillary: Using Knowledge-Graphs for Fact-Aware Language Modeling (ACL 2019) [Paper]
  • knowledge graph language model (KGLM): a neural language model with mechanisms for selecting and copying facts from a knowledge graph that are relevant to the context.
  1. Tree-of-Traversals: A Zero-Shot Reasoning Algorithm for Augmenting Black-box Language Models with Knowledge Graphs (ACL 2024) [Paper] [Notes]
  • A new dataset: MusicBrainz-x-Wikidata, which contains 109 questions that require reasoning with information from both MusicBrainz and Wikidata.
  1. Aligning Complex Knowledge Graph Question Answering as Knowledge-Aware Constrained Code Generation (COLING 2025) [Paper]

4. KG for Evaluation

  1. Knowledge Graph Guided Evaluation of Abstention Techniques (NAACL 2025) [Paper]

๐Ÿ“ LLMs and Knowledge Graphs for Other Tasks

These papers discuss how to utilze LLM and KG for other tasks, such as Information Extraction and Question generation.
  1. ChatGraph: Interpretable Text Classification by Converting ChatGPT Knowledge to Graphs (Arxiv, May 2023) [Paper]
  2. KG-BART: Knowledge Graph-Augmented BART for Generative Commonsense Reasoning (AAAI 2021) [Paper]
  3. QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering (NAACL 2021) [Paper]
  4. Knowledge Graph-Based Chatbot With GPT-3 and Neo4j (from Neo4j) [Link]
  5. Integrate LLM workflows with Knowledge Graph using Neo4j and APOC [Link] โญ
  • Solution 1: use LLMs to generate a Cypher statement that can be used to retrieve connected information from the database.
  • Solution 2: use the connection information to enrich the text embedding representations. Additionally, the enhanced information can be retrieved at query time to provide additional context to the LLM from which it can base its response. (used in this project)
  1. FLEEK: Factual Error Detection and Correction with Evidence Retrieved from External Knowledge (EMNLP 2023) [Paper]
  2. KGLink: A column type annotation method that combines knowledge graph and pre-trained language model (ICDE 2024) [Paper] โญ
  3. Contextualization Distillation from Large Language Model for Knowledge Graph Completion (ACL 2024) [Paper]
  4. From Superficial to Deep: Integrating External Knowledge for Follow-up Question Generation Using Knowledge Graph and LLM (COLING 2025)
  5. CogMG: Collaborative Augmentation Between Large Language Model and Knowledge Graph (ACL 2024) [Paper] [Notes (Chinese)]
  • KGs can be employed to address the hallusiasion in LLMs. Howevers, KGs may be (1) incomplete, or (2) knowledge inconsistant while applied with information extraction techniques. LLM helps to address this two issues. Therefore, a CogMG framework is proposed for the collaborative augmentation between LLM and KGs.
  1. Advancing Feature Extraction in Healthcare through the Integration of Knowledge Graphs and Large Language Models (AAAI 2025) [Paper]
  2. LLM-based Typed Hyperresolution for Commonsense Reasoning with Knowledge Bases (ICLR 2025, poster) [Paper]
  3. LLM is Knowledge Graph Reasoner: LLMโ€™s Intuition-Aware Knowledge Graph Reasoning for Cold-Start Sequential Recommendation (ECIR 2025) [Paper]
  4. Self-supervised Quantized Representation for Seamlessly Integrating Knowledge Graphs with Large Language Models (ACL 2025)
  5. Synergizing Large Language Models and Knowledge Graphs in Science: A Survey (NeurIPS 2025 workshop)
  6. From Chaos to Clarity: A Knowledge Graph-Driven Audit Dataset Generation Framework for LLM Unlearning (AAAI 2026) [Paper]

๐Ÿ“ VLMs and Knowledge Graphs

The papers in this section may have overlaps with MMKGs.
  1. GraphAdapter: Tuning Vision-Language Models With Dual Knowledge Graph (NeurIPS 2023 poster) [OpenReview][Papers]
  2. Improving Commonsense in Vision-Language Models via Knowledge Graph Riddles (Arxiv 2022) [Paper]
  1. ReGen: Reinforcement Learning for Text and Knowledge Base Generation using Pretrained Language Models (Arxiv 2021) [Paper]

From Graph to Text

  1. Tutorial: Knowledge-Driven Vision-Language Encoding (CVPR 2023) [Link]
  2. How Large Language Models Will Disrupt Data Management (VLDB 2023) [Paper] โญ
  3. Graph Machine Learning in the Era of Large Language Models (LLMs) (ACM Transactions on Intelligent Systems and Technology 2025) [Paper]

๐Ÿ“Š Tools, Datasets, and Benchmarks

  1. Open RAG, an opensource RAG KB. [Link]
  2. A Benchmark to Understand the Role of Knowledge Graphs on Large Language Model's Accuracy for Question Answering on Enterprise SQL Databases (GRADES-NDA 2024) [Paper]
  3. OneEval: A benchmark for LLM+KB tasks. [Details (Chinese)] [Website]
  • 4 types of Knowledge (text, table, KG, code) + 6 key areas (general, medical, politics, science, law, programming)