Pith. sign in

REVIEW 5 major objections 9 minor 134 references

Enhancing Large Language Models with Reliable Knowledge Graphs

T0 review · 5 major / 9 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Knowledge-graph prompting can raise a weaker LLM above a stronger one on factual QA, the thesis argues.

desk verdict A thesis that repackages four good papers but doesn't test its advertised pipeline. read the letter →

arxiv 2506.13178 v1 pith:G3TTWUZS submitted 2025-06-16 cs.CL

classification cs.CL
keywords knowledgegraphrefinementerrordetectioncompletionLLMpromptingcontrastivelearningreinforcementmulti-armedbanditquestionanswering
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 PhD thesis is trying to establish that the reason large language models hallucinate is not only missing parameters but missing verified structure: if you first clean a knowledge graph, removing wrong triples and adding missing relations, and then feed the relevant subgraph to a frozen LLM as a prompt, the model answers factual questions more accurately than the same model alone and even than a much larger model. The evidence is a pipeline of four connected methods: contrastive error detection (CAGED), attribute-aware error-aware embedding (AEKE), inductive completion (NORAN), and graph-based prompting (KnowGPT). The headline empirical result is that KnowGPT, built on GPT-3.5, averages 23.7% higher accuracy than GPT-3.5 and 2.9% higher than GPT-4 across CommonsenseQA, OpenBookQA, and MedQA, reaching 92.6% on OpenBookQA. If correct, this means structured, cleaned KG knowledge can be injected into black-box LLMs through prompts to improve factual QA without retraining.

What carries the argument

The load-bearing object is the triple-as-node view of a KG. CAGED builds two congruent triple graphs by linking triples that share a head entity versus a tail entity, then scores confidence by cross-view consistency and intra-triple coherence. AEKE replaces these with a relational hypergraph and an attribute hypergraph, adding a structure-attribute homogeneity signal. NORAN reuses the same triple-as-node idea as a relation network and defines inductive completion as k-hop logical evidence over it, trained with mutual-information maximization (LEIM) so the model captures entity-independent relational patterns. KnowGPT is the integration mechanism: a deep RL policy extracts concise reasoning paths, and a multi-armed bandit chooses among six extraction-and-format pairs, so the final prompt is question-adaptive.

What would settle it

Run the KnowGPT pipeline twice on the same QA benchmarks, once with raw ConceptNet and USMLE and once with the same graphs passed through CAGED, AEKE, and NORAN so that detected errors are removed and completed triples are added. If the refined-graph version does not match or beat the raw version on CommonsenseQA, OpenBookQA, and MedQA accuracy, the thesis's lifecycle claim is falsified as stated even though the individual components may work in isolation.

Watch

Extended reading notes

Core claim

The central claim is that a lifecycle of KG refinement, detecting erroneous triples, exploiting entity attributes, and completing missing relations, produces knowledge that a black-box LLM can consume through prompt construction, and that this pipeline materially reduces hallucination. The thesis states this as four interconnected contributions: CAGED detects errors by contrasting triple representations built from shared-head and shared-tail views; AEKE adds entity attributes as a second view so errors invisible to structure alone become visible; NORAN completes KGs inductively by turning each triple into a node in a relation network and reasoning over k-hop logical evidence without retraining entity embeddings; KnowGPT then retrieves question-specific reasoning chains with reinforcement learning and selects prompt format with a multi-armed bandit. Together, the thesis argues, these form a systematic pipeline demonstrating that reliable KGs significantly enhance the robustness, interpretability, and adaptability of LLMs.

Load-bearing premise

The pipeline claim assumes that cleaning and completing a knowledge graph can only help the later prompting stage, yet the experiments never feed the refined graphs into KnowGPT; if a refined ConceptNet misled the path extractor or the LLM, the central reliable-KGs-enhance-LLMs claim would not hold as stated.

Editorial extensions

If this is right

  • If reliable KGs are injected as prompts, a mid-size model can beat a frontier model on factual QA without training or fine-tuning: KnowGPT, built on GPT-3.5, beats GPT-4 by 2.9% on average across three datasets.
  • Error detection can be framed as self-supervised contrastive consistency: on FB15K, WN18RR, and NELL-995, CAGED ranks injected errors above embedding baselines, and AEKE improves further by unifying structural and attribute signals.
  • Inductive KG completion can operate entity-independently: NORAN predicts relations for unseen entities by message passing over a relation network, outperforming embedding-based and message-passing baselines on five inductive benchmarks.
  • Prompt construction is a decision problem worth optimizing: the multi-armed bandit in KnowGPT shows the same knowledge can be serialized in different formats with 2.2-3.3% accuracy swings, so format selection is part of the gain.
  • The full pipeline is only as good as its weakest KG: the thesis claims the four components form a lifecycle, so refining first and prompting second is the intended deployment path.

Reading between the lines

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

  • Editorial inference: the paper never actually runs KnowGPT on KGs refined by CAGED, AEKE, or NORAN; its experiments use raw ConceptNet and USMLE. A direct extension would be to apply error removal and completion to those graphs and re-run the Chapter 6 experiments, and if accuracy does not rise, the reliable-KGs-enhance-LLMs claim is not yet demonstrated end-to-end.
  • Editorial inference: the refined-then-prompt pipeline assumes monotonicity, that cleaning and completing a KG cannot hurt later prompting. Since completion can add misleading facts and refinement can change the paths the RL extractor finds, this assumption is testable but currently untested.
  • Editorial inference: the multi-armed bandit result suggests a general principle that knowledge grounding and prompt format are coupled hyperparameters for black-box LLMs, and adaptive selection of both may transfer to other retrieval-augmented systems beyond the three QA benchmarks.
  • Editorial inference: the triple-as-node construction recurs across CAGED, AEKE, and NORAN, so the thesis's deeper contribution may be a unified triple-level view of knowledge graphs rather than four separate models.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 9 minor

Summary. This PhD thesis compiles four previously published papers into a claimed lifecycle solution for grounding large language models in reliable knowledge graphs: CAGED (contrastive error detection, Chapter 3), AEKE (attribute-aware error-aware embedding, Chapter 4), NORAN (inductive KG completion, Chapter 5), and KnowGPT (KG-based prompting of black-box LLMs, Chapter 6). The thesis argues that these components form a systematic pipeline from error detection to LLM integration. Each chapter reports benchmark evaluations: error-detection Precision@K/Recall@K on FB15K, WN18RR, NELL-995, FB15K-237, DB15K, and YAGO15K; inductive completion results on five KG benchmarks; and QA accuracy on CommonsenseQA, OpenBookQA, and MedQA-USMLE, including a submitted OpenBookQA leaderboard result of 0.926. The central thesis-level claim is that refining KGs and then using them to prompt LLMs significantly improves robustness, interpretability, and adaptability.

Significance. If the full pipeline claim were supported, the thesis would make a meaningful integrative contribution, combining structured KG refinement with opaque, API-only LLMs. The individual components are valuable: they are well-specified, benchmarked against reasonable baselines, and several have appeared in peer-reviewed venues (CIKM, TKDE, KDD, NeurIPS), which lends credibility to the per-chapter results. The thesis also provides useful ablations, hyperparameter analyses, and a public leaderboard submission for KnowGPT. However, the significance of the thesis as a single work rests on the pipeline claim, and that claim is not tested: no experiment connects Chapters 3-5 to Chapter 6, and the evaluation of error detection uses synthetic corruptions of the same form as the training negatives. The thesis is therefore best characterized as a collection of strong components with an unsupported integrative conclusion, rather than a demonstrated end-to-end system.

major comments (5)
  1. [§3.2.3] The confidence score that the experiments rank by is never defined. The subsection titled 'Joint Confidence Estimation' presents only the KG embedding loss L_kge and the contrastive loss L_con; Figure 3.1's caption indicates that confidence should combine cross-view consistency sim(x_A, z_A) and self-consistency ||e_h + e_r - e_t||, but no closed-form scoring function appears. Without this formula, the Precision@K and Recall@K results in Tables 3.2-3.5 cannot be reproduced or checked, and the claimed 'joint confidence estimation' component of CAGED is unspecified.
  2. [Abstract and §7.1] The thesis's central claim is a systematic pipeline from error detection to LLM integration, but no experiment connects Chapters 3-5 to Chapter 6. KnowGPT is evaluated on the raw ConceptNet and USMLE graphs (§6.4.1), not on graphs refined by CAGED, AEKE, or NORAN, and Chapters 3-5 evaluate their methods on benchmark KGs in isolation. Section 7.1 states that the four papers 'form a lifecycle solution,' which assumes monotonicity: refining and completing a KG cannot degrade later prompting. Section 6.4.3's claim that MAB adapts to KG noise refers to raw KG noise, not to the structural changes induced by the refinement methods. An end-to-end experiment is needed, or the abstract and conclusion should be revised to claim independent components rather than a pipeline.
  3. [§3.3.1 and §4.3.1] Error-detection evaluation uses test errors generated by the same corruption operator used to create training negatives. Chapter 3 injects noise by replacing head or tail entities in the same way as Eq. (3.10), and Chapter 4 explicitly generates noisy triples by replacing the head or tail entity with an entity that has appeared in the same position with the same relation (§4.3.1), matching the negative-sampling distribution. Section 3.1 itself argues that random corruption 'does not reflect the nuanced errors encountered in practice,' so the benchmark measures detection of the synthetic corruption family rather than of the complex, attribute-level errors that motivate AEKE. Additional evaluation on naturally noisy triples, or at least on a different corruption generator, is needed to support the reliability claim.
  4. [§6.4.2, Table 6.1] The headline 'average improvement of 23.7% over GPT-3.5' is the mean of per-dataset percentage-point accuracy differences, not a relative improvement. For example, the CommonsenseQA IHdev entry is 0.827 versus 0.735, a 9.2 percentage-point difference, which the table labels '+9.2%'; the relative improvement is 12.5%. Likewise, the GPT-4 comparison '2.9% (Avg.)' is an average of absolute differences (5.1, 3.3, 2.2, 1.4, 3.7, 1.8). The text should state that these are percentage-point gains, or recompute relative gains; as written, the metric is ambiguous and appears to overstate the improvements.
  5. [§3.3.1 and §5.3.1] No measures of variability are reported. Section 3.3.1 says results are averaged over ten runs using a fixed random seed, which cannot provide variance, and Section 5.3.1 averages three runs with a fixed seed. Several headline claims rest on margins of about 1-3% (e.g., Tables 4.1, 5.2, 6.1), so the absence of standard deviations or multiple-seed results makes it impossible to judge whether the observed differences are significant. Report variance or justify why it is omitted.
minor comments (9)
  1. [§6.1-6.2] Chapter 6 contains two consecutive 'Introduction' sections (§6.1 and §6.2), with §6.2's content actually being a literature review; the sections should be renumbered and retitled.
  2. [§3.3.2] The text refers to 'Table 3.3' where the main 5%-anomaly results are shown, but Table 3.2 is the relevant table; the discussion of other noise levels also cites 'Table 3.3' ambiguously.
  3. [§6.4.2 and Table 6.1] Table 6.1 reports OpenBookQA Test-Acc. of 0.924 for KnowGPT, while the text and Table 6.2 report 0.926 on the leaderboard; clarify whether these are different evaluation settings or correct the inconsistency.
  4. [§5.3.2, Table 5.2] Observation 1 states that NORAN outperforms the best embedding-based and MP-based baselines by 11.2% and 2.2%, respectively, but the margin rows in Table 5.2 report +2.2% versus embedding-based methods and +11.2% versus GNN/MP-based methods; the two values are reversed in the text.
  5. [Tables 6.3-6.4] The captions of Tables 6.3 and 6.4 are identical ('Ablation study on the effectiveness of two knowledge extraction methods'), although Table 6.4 reports prompt-format ablations; the captions should be corrected.
  6. [§7.1] The conclusion lists 'Inductive Knowledge Graph Completion (Paper 3)' twice verbatim; the duplicate paragraph should be removed.
  7. [References] The reference list contains duplicates: [9] and [10] are both the TransE paper, and [53] and [54] are both the UnifiedQA paper; in [4], 'Clude' should be 'Claude.'
  8. [§6.4.1] The text says dataset statistics are in 'Table 3.1 in the Appendix,' but the thesis has no appendix; provide the table or fix the reference.
  9. [Eq. (5.2)] The notation C^{(k)}X^{(k)} \circ f^{(k)} leaves unclear whether the feature transformation f is applied before or after the Hadamard product; specify the operation order.

Circularity Check

1 steps flagged · score 4.0 of 10

Error-detection evaluation is partially circular: test noise is generated by the same head/tail corruption used to build training negatives; LLM results remain externally grounded.

  1. fitted input called prediction [Sec. 3.2.3 (Eq. 3.10); Sec. 3.3.1; Sec. 4.3.1; Sec. 3.4]
    "Gˆ = {(hˆ, r, t)|hˆ∈ G} ∪ {(h, r, tˆ)|tˆ∈ G}. (3.10) ... Each dataset is constructed by injecting noise at levels of 5%, 10%, and 15% of the total triples. ... The noisy triples are generated by replacing the head or tail entity in a given triple (h, r, t) with an entity that has appeared in the same position with the same relation in the dataset. ... Traditional KG error detection methods typically rely on synthetically generated false triples—created by randomly replacing head or tail entities—which fail to capture the nuanced errors found in practice."

    Eq. (3.10) defines the negative triples used to train CAGED's margin-based scorer by replacing the head or tail entity of each true triple. Sec. 3.3.1 builds the error-detection test sets by injecting noise, and Sec. 4.3.1 specifies that noisy triples are generated by the same head/tail-replacement operation. The reported Precision@K/Recall@K is therefore a measure of how well the model re-detects the very corruption distribution it was optimized to down-weight; the test errors are not independent of the training signal. The paper itself acknowledges that random head/tail replacement fails to capture the nuanced errors found in practice (Sec. 3.4), so the synthetic benchmark does not independently validate detection of realistic KG noise.

full rationale

The only concrete circular step is confined to Chapters 3 and 4: the negative triples used to train the error detector (Eq. 3.10) and the noisy triples injected into the evaluation benchmarks (Secs. 3.3.1 and 4.3.1) are both produced by replacing the head or tail entity of true triples, so the reported error-detection accuracy partially measures recognition of the training corruption pattern rather than independent detection of real-world KG errors. This does not compromise the KnowGPT results, which are evaluated on external QA benchmarks using raw ConceptNet and USMLE; those results are independent of the fitted corruption distribution. The thesis-level pipeline claim that refined KGs from Chapters 3 to 5 feed KnowGPT is not demonstrated because KnowGPT retrieves directly from raw KGs (Sec. 6.4.1), but that is a composition and monotonicity gap, not circularity. No load-bearing uniqueness theorem or self-citation chain is invoked, so the overall circularity burden is modest.

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

The central pipeline claim adds no new free parameters, but each component relies on tuned hyperparameters and unverified domain assumptions. No invented entities are introduced; the relation network and logical evidence are representational constructs rather than new particles, forces, or dimensions. The most consequential assumptions are the equivalence of synthetic corruption to real KG noise and the untested composability of refinement and prompting.

free parameters (6)
  • margin gamma = 0.1 to 1.0 grid; stable range reported
    CAGED and AEKE margin losses (Eq. 3.9, Eq. 4.1); selected on validation and reported stable.
  • attention threshold mu = 0.005 for FB15K and NELL-995, 0.01 for WN18RR
    EaGNN neighbor pruning in Eq. 3.5; tuned via grid search from 0.001 to 0.2.
  • loss trade-off lambda = about 10 for WN18RR and NELL-995, about 0.1 for FB15K
    Balances contrastive and embedding losses, e.g., Chapter 4 objective Eq. 4.17; grid search over 0.001 to 1000.
  • contrastive temperature tau = not reported
    Used in InfoNCE losses Eq. 3.11 and Eq. 4.10; no tuning details supplied, hurting reproducibility.
  • MAB exploration constant and ridge lambda_i = not reported
    Equations 6.7 and 6.8 define the bandit update; concrete values and the feedback schedule are absent.
  • RL auxiliary reward weights = not reported
    Context-relatedness and conciseness rewards in Eqs. 6.3 and 6.4 are combined with reachability, but the combination weights are not given.
assumptions (5)
  • domain assumption Synthetically injected corruptions represent real KG errors.
    Chapters 3 and 4 create test errors by replacing an entity or relation (Sec. 3.3.1, Sec. 4.3.1) and then claim real-world error detection. If real errors are subtler, reported Precision and Recall overstate field performance.
  • domain assumption Entity attributes in benchmark KGs are complete enough to expose relational errors.
    AEKE's attribute hypergraph (Def. 3, Sec. 4.2.2) assumes attribute types correlate with the relation of a triple; the paper provides no analysis of attribute coverage or noise.
  • domain assumption Removing entities from the training subgraph and re-adding their edges at test time is a valid inductive KGC protocol.
    Sec. 5.3.1 constructs inductive datasets by sampling disjoint subgraphs; this assumes the sampled split preserves the logical patterns needed for inference.
  • domain assumption LLM API feedback used to train the multi-armed bandit is stable enough to learn from.
    Section 6.3.2 treats each prompt as correct or incorrect and updates alpha weights; no repeated sampling or confidence intervals are reported, so stochastic API outputs could skew selection.
  • domain assumption GNN message passing over the relation network captures logical evidence.
    Definitions 4 and 5 identify k-hop ego graphs with logical evidence; this is a modeling choice, not a proven equivalence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Large Language Models with Reliable Knowledge Graphs." pith.science (2026). https://pith.science/paper/G3TTWUZS

@misc{pith2026250613178,
  author       = {Pith},
  title        = {Pith review of: Enhancing Large Language Models with Reliable Knowledge Graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G3TTWUZS}},
  note         = {Machine review of arXiv:2506.13178}
}
read the original abstract

Large Language Models (LLMs) have demonstrated remarkable capabilities in text generation and understanding, yet their reliance on implicit, unstructured knowledge often leads to factual inaccuracies and limited interpretability. Knowledge Graphs (KGs), with their structured, relational representations, offer a promising solution to ground LLMs in verified knowledge. However, their potential remains constrained by inherent noise, incompleteness, and the complexity of integrating their rigid structure with the flexible reasoning of LLMs. This thesis presents a systematic framework to address these limitations, advancing the reliability of KGs and their synergistic integration with LLMs through five interconnected contributions. This thesis addresses these challenges through a cohesive framework that enhances LLMs by refining and leveraging reliable KGs. First, we introduce contrastive error detection, a structure-based method to identify incorrect facts in KGs. This approach is extended by an attribute-aware framework that unifies structural and semantic signals for error correction. Next, we propose an inductive completion model that further refines KGs by completing the missing relationships in evolving KGs. Building on these refined KGs, KnowGPT integrates structured graph reasoning into LLMs through dynamic prompting, improving factual grounding. These contributions form a systematic pipeline (from error detection to LLM integration), demonstrating that reliable KGs significantly enhance the robustness, interpretability, and adaptability of LLMs.

Figures

Figures reproduced from arXiv: 2506.13178 by the authors.

Figure 3.1
Figure 3.1. Two separate augmentation operators are applied to the original KG, [PITH_FULL_IMAGE:figures/full_fig_p029_3_1.png] view at source ↗
Figure 3.2
Figure 3.2. Impact of hyperparameters on the three datasets. [PITH_FULL_IMAGE:figures/full_fig_p040_3_2.png] view at source ↗
Figure 4.1
Figure 4.1. Running examples of complex errors in real-world KGs. (a) presents a KG [PITH_FULL_IMAGE:figures/full_fig_p043_4_1.png] view at source ↗
Figures from the paper (6 more)
Figure 4.2
Figure 4.2. Figure 4.2: We perform a relation-induced construction process to build the [PITH_FULL_IMAGE:figures/full_fig_p046_4_2.png]
Figure 4.3
Figure 4.3. Figure 4.3: KG completion results of AEKE variants based on the three datasets [PITH_FULL_IMAGE:figures/full_fig_p059_4_3.png]
Figure 5.1
Figure 5.1. Figure 5.1: (i) A toy example of inductive knowledge graph completion, i.e. predict￾ing the relationship (“?”) for unseen entities, e.g. “Martin Eberhard”, provided with a few links, e.g. (Martin, :WorkAt, TESLA); (ii) Illustration of the corresponding relation network for knowl…
Figure 5
Figure 5. Figure 5: , NORAN is composed of three core components: ( [PITH_FULL_IMAGE:figures/full_fig_p065_5.png]
Figure 6.1
Figure 6.1. Figure 6.1: The overall architecture of our proposed knowledge graph prompting framework, i.e., KnowGPT. Given the question context with multiple choices, we first retrieve a question-specific subgraph from the real-world KG. Knowledge Extraction is first dedicated to searching …
Figure 6.2
Figure 6.2. Figure 6.2: A case study on exploring the effectiveness of different prompt formats for [PITH_FULL_IMAGE:figures/full_fig_p093_6_2.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

134 extracted references · 57 canonical work pages

  1. [1]

    IEEE Transactions on Neural Networks and Learning Systems , 2021

    A survey on knowledge graphs: Representation, acquisition, and applications. IEEE Transactions on Neural Networks and Learning Systems , 2021

  2. [2]

    Rakesh Agrawal, Tomasz Imielinski, and Arun N. Swami. Mining association rules between sets of items in large databases. In Peter Buneman and Sushil Jajodia, editors, SIGMOD , pages 207--216, 1999

  3. [3]

    Publicly available clinical bert embeddings

    Emily Alsentzer, John Murphy, William Boag, Wei-Hung Weng, Di Jindi, Tristan Naumann, and Matthew McDermott. Publicly available clinical bert embeddings. In Proceedings of the 2nd Clinical Natural Language Processing Workshop , pages 72--78, 2019

  4. [4]

    The claude 3 model family: Opus, sonnet, haiku

    AI Anthropic. The claude 3 model family: Opus, sonnet, haiku. Claude-3 Model Card , 2024

  5. [5]

    What is normal, what is strange, and what is missing in a knowledge graph: Unified characterization via inductive summarization

    Caleb Belth, Xinyi Zheng, Jilles Vreeken, and Danai Koutra. What is normal, what is strange, and what is missing in a knowledge graph: Unified characterization via inductive summarization. In WWW , 2020

  6. [6]

    The Unified Medical Language System (UMLS): integrating biomedical terminology

    Olivier Bodenreider. The Unified Medical Language System (UMLS): integrating biomedical terminology . Nucleic Acids Research , 32:D267--D270, 01 2004

  7. [7]

    Freebase: a collaboratively created graph database for structuring human knowledge

    Kurt Bollacker, Colin Evans, Praveen Paritosh, Tim Sturge, and Jamie Taylor. 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. ACM, 2008

  8. [8]

    Bollacker, Robert P

    Kurt D. Bollacker, Robert P. Cook, and Patrick Tufts. Freebase: A shared database of structured general human knowledge. In AAAI , 2007

Show all 134 references
  1. [9]

    Translating embeddings for modeling multi-relational data

    Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and Oksana Yakhnenko. Translating embeddings for modeling multi-relational data. NeurIPS , 26, 2013

  2. [10]

    Translating embeddings for modeling multi-relational data

    Antoine Bordes, Nicolas Usunier, Alberto Garc \' a - Dur \' a n, Jason Weston, and Oksana Yakhnenko. Translating embeddings for modeling multi-relational data. In NeurIPS , 2013

  3. [11]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, 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, Jeffr...

  4. [12]

    Hruschka Jr., and Tom M

    Andrew Carlson, Justin Betteridge, Bryan Kisiel, Burr Settles, Estevam R. Hruschka Jr., and Tom M. Mitchell. Toward an architecture for never-ending language learning. In AAAI 2010 , 2010

  5. [13]

    Neural legal judgment prediction in english

    Ilias Chalkidis, Ion Androutsopoulos, and Nikolaos Aletras. Neural legal judgment prediction in english. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics , pages 4317--4323, 2019

  6. [14]

    Lexglue: A benchmark dataset for legal language understanding in english

    Ilias Chalkidis, Abhik Jana, Dirk Hartung, Michael Bommarito, Ion Androutsopoulos, Daniel Martin Katz, and Nikolaos Aletras. Lexglue: A benchmark dataset for legal language understanding in english. arXiv preprint arXiv:2110.00976 , 2021

  7. [15]

    Novelty detection via contrastive learning with negative data augmentation

    Chengwei Chen, Yuan Xie, Shaohui Lin, Ruizhi Qiao, Jian Zhou, Xin Tan, Yi Zhang, and Lizhuang Ma. Novelty detection via contrastive learning with negative data augmentation. arXiv preprint arXiv:2106.09958 , 2021

  8. [16]

    Neighbor enhanced graph convolutional networks for node classification and recommendation

    Hao Chen, Zhong Huang, Yue Xu, Zengde Deng, Feiran Huang, Peng He, and Zhoujun Li. Neighbor enhanced graph convolutional networks for node classification and recommendation. Knowledge-Based Systems , 246:108594, 2022

  9. [17]

    Label-aware graph convolutional networks

    Hao Chen, Yue Xu, Feiran Huang, Zengde Deng, Wenbing Huang, Senzhang Wang, Peng He, and Zhoujun Li. Label-aware graph convolutional networks. In Proceedings of the 29th ACM International Conference on Information & Knowledge Management , page 1977–1980, 2020

  10. [18]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey E. Hinton. A simple framework for contrastive learning of visual representations. In ICML , 2020

  11. [19]

    Noigan: Noise aware knowledge graph embedding with gan

    Kewei Cheng, Yikai Zhu, Ming Zhang, and Yizhou Sun. Noigan: Noise aware knowledge graph embedding with gan. 2019

  12. [20]

    Rule-based graph repairing: Semantic and efficient repairing methods

    Yurong Cheng, Lei Chen, Ye Yuan, and Guoren Wang. Rule-based graph repairing: Semantic and efficient repairing methods. In ICDE , 2018

  13. [21]

    Palm: Scaling language modeling with pathways

    Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, and et al. Palm: Scaling language modeling with pathways. Journal of Machine Learning Research (JMLR) , 2023

  14. [22]

    From ‘f’to ‘a’on the ny regents science exams: An overview of the aristo project

    Peter Clark, Oren Etzioni, Tushar Khot, Daniel Khashabi, Bhavana Mishra, Kyle Richardson, Ashish Sabharwal, Carissa Schoenick, Oyvind Tafjord, Niket Tandon, et al. From ‘f’to ‘a’on the ny regents science exams: An overview of the aristo project. AI Magazine , 41(4):39--53, 2020

  15. [23]

    Hierarchy-aware multi-hop question answering over knowledge graphs

    Junnan Dong, Qinggang Zhang, Xiao Huang, Keyu Duan, Qiaoyu Tan, and Zhimeng Jiang. Hierarchy-aware multi-hop question answering over knowledge graphs. In WWW , pages 2519--2527, 2023

  16. [24]

    Active ensemble learning for knowledge graph error detection

    Junnan Dong, Qinggang Zhang, Xiao Huang, Qiaoyu Tan, Daochen Zha, and Zhao Zihao. Active ensemble learning for knowledge graph error detection. In WSDM , pages 877--885, 2023

  17. [25]

    Cost-efficient knowledge-based question answering with large language models

    Junnan Dong, Qinggang Zhang, Chuang Zhou, Hao Chen, Daochen Zha, and Xiao Huang. Cost-efficient knowledge-based question answering with large language models. arXiv preprint arXiv:2405.17337 , 2024

  18. [26]

    Modality-aware integration with large language models for knowledge-based visual question answering

    Junnan Dong, Qinggang Zhang, Huachi Zhou, and et al. Modality-aware integration with large language models for knowledge-based visual question answering. arXiv preprint arXiv:2402.12728 , 2024

  19. [27]

    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 Proceedings of the 20th ACM SIGKDD international conference on Knowledge dis...

  20. [28]

    Snomed-ct: The advanced terminology and coding system for ehealth

    Kevin Donnelly et al. Snomed-ct: The advanced terminology and coding system for ehealth. Studies in health technology and informatics , 121:279, 2006

  21. [29]

    Scalable multi-hop relational reasoning for knowledge-aware question answering

    Yanlin Feng, Xinyue Chen, Bill Yuchen Lin, Peifeng Wang, Jun Yan, and Xiang Ren. Scalable multi-hop relational reasoning for knowledge-aware question answering. In EMNLP , pages 1295--1309, 2020

  22. [30]

    Suchanek

    Luis Antonio Gal \' a rraga, Christina Teflioudi, Katja Hose, and Fabian M. Suchanek. AMIE: association rule mining under incomplete evidence in ontological knowledge bases. In WWW , 2013

  23. [31]

    Kgclean: An embedding powered knowledge graph cleaning framework

    Congcong Ge, Yunjun Gao, Honghui Weng, Chong Zhang, Xiaoye Miao, and Baihua Zheng. Kgclean: An embedding powered knowledge graph cleaning framework. arXiv preprint arXiv:2004.14478 , 2020

  24. [32]

    Openagi: When llm meets domain experts

    Yingqiang Ge, Wenyue Hua, Kai Mei, Juntao Tan, Shuyuan Xu, Zelong Li, Yongfeng Zhang, et al. Openagi: When llm meets domain experts. Advances in Neural Information Processing Systems , 36, 2024

  25. [33]

    Mitigating large language model hallucinations via autonomous knowledge graph-based retrofitting

    Xinyan Guan, Yanjiang Liu, Hongyu Lin, Yaojie Lu, Ben He, Xianpei Han, and Le Sun. Mitigating large language model hallucinations via autonomous knowledge graph-based retrofitting. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 38, pages 18126--18134, 2024

  26. [34]

    Knowledge graph embedding with iterative guidance from soft rules

    Shu Guo, Quan Wang, Lihong Wang, Bin Wang, and Li Guo. Knowledge graph embedding with iterative guidance from soft rules. In AAAI , 2018

  27. [35]

    Don’t stop pretraining: Adapt language models to domains and tasks

    Suchin Gururangan, Ana Marasovi \'c , Swabha Swayamdipta, Kyle Lo, Iz Beltagy, Doug Downey, and Noah A Smith. Don’t stop pretraining: Adapt language models to domains and tasks. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , pages ...

  28. [36]

    Contrastive multi-view representation learning on graphs

    Kaveh Hassani and Amir Hosein Khas Ahmadi. Contrastive multi-view representation learning on graphs. In ICML , 2020

  29. [37]

    Vandalism detection in wikidata

    Stefan Heindorf, Martin Potthast, Benno Stein, and Gregor Engels. Vandalism detection in wikidata. In CIKM , 2016

  30. [38]

    H \' e naff

    Olivier J. H \' e naff. Data-efficient image recognition with contrastive predictive coding. In ICML , 2020

  31. [39]

    Devon Hjelm, Alex Fedorov, Samuel Lavoie - Marchildon, Karan Grewal, Philip Bachman, Adam Trischler, and Yoshua Bengio

    R. Devon Hjelm, Alex Fedorov, Samuel Lavoie - Marchildon, Karan Grewal, Philip Bachman, Adam Trischler, and Yoshua Bengio. Learning deep representations by mutual information estimation and maximization. In ICLR , 2019

  32. [40]

    Knowledge-to-sql: Enhancing sql generation with data expert llm

    Zijin Hong, Zheng Yuan, Hao Chen, Qinggang Zhang, Feiran Huang, and Xiao Huang. Knowledge-to-sql: Enhancing sql generation with data expert llm. arXiv preprint arXiv:2402.11517 , 2024

  33. [41]

    Next-generation database interfaces: A survey of llm-based text-to-sql

    Zijin Hong, Zheng Yuan, Qinggang Zhang, Hao Chen, Junnan Dong, Feiran Huang, and Xiao Huang. Next-generation database interfaces: A survey of llm-based text-to-sql. arXiv preprint arXiv:2406.08426 , 2024

  34. [42]

    A survey of knowledge enhanced pre-trained language models

    Linmei Hu, Zeyi Liu, Ziwang Zhao, Lei Hou, Liqiang Nie, and Juanzi Li. A survey of knowledge enhanced pre-trained language models. IEEE Transactions on Knowledge and Data Engineering , 2023

  35. [43]

    Aligning distillation for cold-start item recommendation

    Feiran Huang, Zefan Wang, Xiao Huang, Yufeng Qian, Zhetao Li, and Hao Chen. Aligning distillation for cold-start item recommendation. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval , page 1147–1157, 2023

  36. [44]

    Large language model interaction simulator for cold-start item recommendation

    Feiran Huang, Zhenghang Yang, Junyi Jiang, Yuanchen Bei, Yijie Zhang, and Hao Chen. Large language model interaction simulator for cold-start item recommendation. arXiv preprint arXiv:2402.09176 , 2024

  37. [45]

    Knowledge graph embedding based question answering

    Xiao Huang, Jingyuan Zhang, Dingcheng Li, and Ping Li. Knowledge graph embedding based question answering. In WSDM , 2019

  38. [46]

    Knowledge graph embedding based question answering

    Xiao Huang, Jingyuan Zhang, Dingcheng Li, and Ping Li. Knowledge graph embedding based question answering. In WSDM , pages 105--113, 2019

  39. [47]

    Mvp-tuning: Multi-view knowledge retrieval with prompt tuning for commonsense reasoning

    Yongfeng Huang, Yanyang Li, Yichong Xu, Lin Zhang, Ruyi Gan, Jiaxing Zhang, and Liwei Wang. Mvp-tuning: Multi-view knowledge retrieval with prompt tuning for commonsense reasoning. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volu...

  40. [48]

    Clues before answers: Generation-enhanced multiple-choice qa

    Zixian Huang, Ao Wu, Jiaying Zhou, Yu Gu, Yue Zhao, and Gong Cheng. Clues before answers: Generation-enhanced multiple-choice qa. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies ,...

  41. [49]

    Triple trustworthiness measurement for knowledge graph

    Shengbin Jia, Yang Xiang, Xiaojun Chen, Kun Wang, and Shijia E. Triple trustworthiness measurement for knowledge graph. In WWW , 2019

  42. [50]

    Uni KGQA : Unified retrieval and reasoning for solving multi-hop question answering over knowledge graph

    Jinhao Jiang, Kun Zhou, Xin Zhao, and Ji-Rong Wen. Uni KGQA : Unified retrieval and reasoning for solving multi-hop question answering over knowledge graph. In The Eleventh International Conference on Learning Representations , 2023

  43. [51]

    What disease does this patient have? a large-scale open domain question answering dataset from medical exams

    Di Jin, Eileen Pan, Nassim Oufattole, Wei-Hung Weng, Hanyi Fang, and Peter Szolovits. What disease does this patient have? a large-scale open domain question answering dataset from medical exams. Applied Sciences , 11(14), 2021

  44. [52]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin Ming-Wei Chang Kenton and Lee Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of NAACL-HLT , pages 4171--4186, 2019

  45. [53]

    UnifiedQA : Crossing format boundaries with a single qa system

    Daniel Khashabi, Sewon Min, Tushar Khot, Ashish Sabharwal, Oyvind Tafjord, Peter Clark, and Hannaneh Hajishirzi. UnifiedQA : Crossing format boundaries with a single qa system. In Findings of the Association for Computational Linguistics: EMNLP 2020 , pages 1896--1907, 2020

  46. [54]

    Unifiedqa: Crossing format boundaries with a single qa system

    Daniel Khashabi, Sewon Min, Tushar Khot, Ashish Sabharwal, Oyvind Tafjord, Peter Clark, and Hannaneh Hajishirzi. Unifiedqa: Crossing format boundaries with a single qa system. In EMNLP 2020 , pages 1896--1907, 2020

  47. [55]

    Kipf, Elise van der Pol, and Max Welling

    Thomas N. Kipf, Elise van der Pol, and Max Welling. Contrastive learning of structured world models. In ICLR , 2020

  48. [56]

    Patent classification by fine-tuning bert language model

    Jieh-Sheng Lee and Jieh Hsiang. Patent classification by fine-tuning bert language model. World Patent Information , 61:101965, 2020

  49. [57]

    Biobert: a pre-trained biomedical language representation model for biomedical text mining

    Jinhyuk Lee, Wonjin Yoon, Sungdong Kim, Donghyeon Kim, Sunkyu Kim, Chan Ho So, and Jaewoo Kang. Biobert: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics , 36(4):1234--1240, 2020

  50. [58]

    Mendes, Sebastian Hellmann, Mohamed Morsey, Patrick van Kleef, S \" o ren Auer, and Christian Bizer

    Jens Lehmann, Robert Isele, Max Jakob, Anja Jentzsch, Dimitris Kontokostas, Pablo N. Mendes, Sebastian Hellmann, Mohamed Morsey, Patrick van Kleef, S \" o ren Auer, and Christian Bizer. Dbpedia - A large-scale, multilingual knowledge base extracted from wikipedia. Semantic Web...

  51. [59]

    Retrieval-augmented generation for knowledge-intensive nlp tasks

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, and et al. Retrieval-augmented generation for knowledge-intensive nlp tasks. In Advances in Neural Information Processing Systems (NeurIPS) , 2020

  52. [60]

    A survey on retrieval-augmented text generation

    Huayang Li, Yixuan Su, Deng Cai, Yan Wang, and Lemao Liu. A survey on retrieval-augmented text generation. arXiv preprint arXiv:2202.01110 , 2022

  53. [61]

    Kcube: A knowledge graph university curriculum framework for student advising and career planning

    Qing Li, George Baciu, Jiannong Cao, Xiao Huang, Richard Chen Li, Peter HF Ng, Junnan Dong, Qinggang Zhang, Zackary PT Sin, and Yaowei Wang. Kcube: A knowledge graph university curriculum framework for student advising and career planning. In International Conference on Blende...

  54. [62]

    Constructing low-redundant and high-accuracy knowledge graphs for education

    Wentao Li, Huachi Zhou, Junnan Dong, and et al. Constructing low-redundant and high-accuracy knowledge graphs for education. In International Conference on Web-Based Learning (ICWL) , 2022

  55. [63]

    Learning entity and relation embeddings for knowledge graph completion

    Yankai Lin, Zhiyuan Liu, Maosong Sun, Yang Liu, and Xuan Zhu. Learning entity and relation embeddings for knowledge graph completion. In AAAI , 2015

  56. [64]

    Self-alignment pretraining for biomedical entity representations

    Fangyu Liu, Ehsan Shareghi, Zaiqiao Meng, Marco Basaldella, and Nigel Collier. Self-alignment pretraining for biomedical entity representations. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language...

  57. [65]

    Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing

    Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hiroaki Hayashi, and Graham Neubig. Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Computing Surveys , 55(9):1--35, 2023

  58. [66]

    K-bert: Enabling language representation with knowledge graph

    Weijie Liu, Peng Zhou, Zhe Zhao, Zhiruo Wang, Qi Ju, Haotang Deng, and Ping Wang. K-bert: Enabling language representation with knowledge graph. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 34, pages 2901--2908, 2020

  59. [67]

    Error detection on knowledge graphs with triple embedding

    Yezi Liu, Qinggang Zhang, Mengnan Du, Xiao Huang, and Xia Hu. Error detection on knowledge graphs with triple embedding. In 2023 31st European Signal Processing Conference (EUSIPCO) , pages 1604--1608. IEEE, 2023

  60. [68]

    Roberta: A robustly optimized bert pretraining approach

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692 , 2019

  61. [69]

    Reasoning on graphs: Faithful and interpretable large language model reasoning

    LINHAO LUO, Yuan-Fang Li, Reza Haf, and et al. Reasoning on graphs: Faithful and interpretable large language model reasoning. In International Conference on Learning Representations (ICLR) , 2024

  62. [70]

    Reasoning on graphs: Faithful and interpretable large language model reasoning

    Linhao Luo, Yuan-Fang Li, Gholamreza Haffari, and Shirui Pan. Reasoning on graphs: Faithful and interpretable large language model reasoning. arXiv preprint arXiv:2310.01061 , 2023

  63. [71]

    Suchanek

    Farzaneh Mahdisoltani, Joanna Biega, and Fabian M. Suchanek. YAGO3 : A knowledge base from multilingual wikipedias. In CIDR , 2015

  64. [72]

    Detection of relation assertion errors in knowledge graphs

    Andr \' e Melo and Heiko Paulheim. Detection of relation assertion errors in knowledge graphs. In Knowledge Capture Conference , 2017

  65. [73]

    Can a suit of armor conduct electricity? a new dataset for open book question answering

    Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish Sabharwal. Can a suit of armor conduct electricity? a new dataset for open book question answering. In EMNLP , pages 2381--2391, 2018

  66. [74]

    Never-ending learning

    Tom Mitchell, William Cohen, Estevam Hruschka, Partha Talukdar, Bishan Yang, Justin Betteridge, Andrew Carlson, Bhavana Dalvi, Matt Gardner, Bryan Kisiel, et al. Never-ending learning. Communications of the ACM , 61(5):103--115, 2018

  67. [75]

    Learning attention-based embeddings for relation prediction in knowledge graphs

    Deepak Nathani, Jatin Chauhan, Charu Sharma, and Manohar Kaul. Learning attention-based embeddings for relation prediction in knowledge graphs. In ACL , 2019

  68. [76]

    Pattern-aware and noise-resilient embedding models

    Mojtaba Nayyeri, Sahar Vahdati, Emanuel Sallinger, Mirza Mohtashim Alam, Hamed Shariat Yazdi, and Jens Lehmann. Pattern-aware and noise-resilient embedding models. In European Conference on Information Retrieval , pages 483--496. Springer, 2021

  69. [77]

    Gpt-4 technical report

    OpenAI. Gpt-4 technical report. arXiv preprint arXiv:2303.08774 , 2023

  70. [78]

    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 , 2024

  71. [79]

    Knowledge graph refinement: A survey of approaches and evaluation methods

    Heiko Paulheim. Knowledge graph refinement: A survey of approaches and evaluation methods. Semantic Web , 2017

  72. [80]

    Serving dbpedia with DOLCE - more than just adding a cherry on top

    Heiko Paulheim and Aldo Gangemi. Serving dbpedia with DOLCE - more than just adding a cherry on top. In ISWC , 2015

  73. [81]

    Knowledge enhanced contextual word representations

    Matthew E Peters, Mark Neumann, Robert Logan, Roy Schwartz, Vidur Joshi, Sameer Singh, and Noah A Smith. Knowledge enhanced contextual word representations. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Join...

  74. [82]

    Modeling relational data with graph convolutional networks

    Michael Schlichtkrull, Thomas N Kipf, Peter Bloem, Rianne van den Berg, Ivan Titov, and Max Welling. Modeling relational data with graph convolutional networks. In European semantic web conference , pages 593--607. Springer, 2018

  75. [83]

    Confidence-aware negative sampling method for noisy knowledge graph embedding

    Yingchun Shan, Chenyang Bu, Xiaojian Liu, Shengwei Ji, and Lei Li. Confidence-aware negative sampling method for noisy knowledge graph embedding. In 2018 IEEE International Conference on Big Knowledge (ICBK) , pages 33--40. IEEE, 2018

  76. [84]

    Differentiable neuro-symbolic reasoning on large-scale knowledge graphs

    Chen Shengyuan, Yunfeng Cai, Huang Fang, Xiao Huang, and Mingming Sun. Differentiable neuro-symbolic reasoning on large-scale knowledge graphs. Advances in Neural Information Processing Systems , 36, 2024

  77. [85]

    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 Proceedings of the AAAI conference on artificial intelligence , volume 31, 2017

  78. [86]

    Ernie 3.0: Large-scale knowledge enhanced pre-training for language understanding and generation

    Yu Sun, Shuohuan Wang, Shikun Feng, Siyu Ding, Chao Pang, Junyuan Shang, Jiaxiang Liu, Xuyi Chen, Yanbin Zhao, Yuxiang Lu, et al. Ernie 3.0: Large-scale knowledge enhanced pre-training for language understanding and generation. arXiv preprint arXiv:2107.02137 , 2021

  79. [87]

    Jointlk: Joint reasoning with language models and knowledge graphs for commonsense question answering

    Yueqing Sun, Qi Shi, Le Qi, and Yu Zhang. Jointlk: Joint reasoning with language models and knowledge graphs for commonsense question answering. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language...

  80. [88]

    Rotate: Knowledge graph embedding by relational rotation in complex space

    Zhiqing Sun, Zhi-Hong Deng, Jian-Yun Nie, and et al. Rotate: Knowledge graph embedding by relational rotation in complex space. arXiv preprint arXiv:1902.10197 , 2019

  81. [89]

    Csi: Novelty detection via contrastive learning on distributionally shifted instances

    Jihoon Tack, Sangwoo Mo, Jongheon Jeong, and Jinwoo Shin. Csi: Novelty detection via contrastive learning on distributionally shifted instances. Advances in neural information processing systems , 33:11839--11852, 2020

  82. [90]

    Commonsenseqa: A question answering challenge targeting commonsense knowledge

    Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. Commonsenseqa: A question answering challenge targeting commonsense knowledge. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua...

  83. [91]

    Completeness-aware rule learning from knowledge graphs

    Thomas Pellissier Tanon, Daria Stepanova, Simon Razniewski, Paramita Mirza, and Gerhard Weikum. Completeness-aware rule learning from knowledge graphs. In IJCAI , 2018

  84. [92]

    Grapeqa: Graph augmentation and pruning to enhance question-answering

    Dhaval Taunk, Lakshya Khanna, Siri Venkata Pavan Kumar Kandru, Vasudeva Varma, Charu Sharma, and Makarand Tapaswi. Grapeqa: Graph augmentation and pruning to enhance question-answering. In Companion Proceedings of the ACM Web Conference 2023 , pages 1138--1144, 2023

  85. [93]

    Inductive relation prediction by subgraph reasoning

    Komal Teru, Etienne Denis, and Will Hamilton. Inductive relation prediction by subgraph reasoning. In International Conference on Machine Learning , pages 9448--9457. PMLR, 2020

  86. [94]

    Complex embeddings for simple link prediction

    Th \' e o Trouillon, Johannes Welbl, Sebastian Riedel, \' E ric Gaussier, and Guillaume Bouchard. Complex embeddings for simple link prediction. In ICML , 2016

  87. [95]

    The ai doctor is in: A survey of task-oriented dialogue systems for healthcare applications

    Mina Valizadeh and Natalie Parde. The ai doctor is in: A survey of task-oriented dialogue systems for healthcare applications. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages 6638--6660, 2022

  88. [96]

    Relational message passing for knowledge graph completion

    Hongwei Wang, Hongyu Ren, and Jure Leskovec. Relational message passing for knowledge graph completion. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , pages 1697--1707, 2021

  89. [97]

    Knowledge graph convolutional networks for recommender systems

    Hongwei Wang, Miao Zhao, Xing Xie, Wenjie Li, and Minyi Guo. Knowledge graph convolutional networks for recommender systems. In WWW , 2019

  90. [98]

    Knowledge graph convolutional networks for recommender systems

    Hongwei Wang, Miao Zhao, Xing Xie, Wenjie Li, and Minyi Guo. Knowledge graph convolutional networks for recommender systems. In The world wide web conference , pages 3307--3313, 2019

  91. [99]

    Boosting language models reasoning with chain-of-knowledge prompting

    Jianing Wang, Qiushi Sun, Nuo Chen, and et al. Boosting language models reasoning with chain-of-knowledge prompting. arXiv preprint arXiv:2306.06427 , 2023

  92. [100]

    Knowledge-driven cot: Exploring faithful reasoning in llms for knowledge-intensive question answering

    Keheng Wang, Feiyu Duan, Sirui Wang, Peiguang Li, Yunsen Xian, Chuantao Yin, Wenge Rong, and Zhang Xiong. Knowledge-driven cot: Exploring faithful reasoning in llms for knowledge-intensive question answering. arXiv preprint arXiv:2308.13259 , 2023

  93. [101]

    Gnn is a counter? revisiting gnn for question answering, 2021

    Kuan Wang, Yuyu Zhang, Diyi Yang, Le Song, and Tao Qin. Gnn is a counter? revisiting gnn for question answering, 2021

  94. [102]

    Logic attention based neighborhood aggregation for inductive knowledge graph embedding

    Peifeng Wang, Jialong Han, Chenliang Li, and Rong Pan. Logic attention based neighborhood aggregation for inductive knowledge graph embedding. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 33, pages 7152--7159, 2019

  95. [103]

    Kepler: A unified model for knowledge embedding and pre-trained language representation

    Xiaozhi Wang, Tianyu Gao, Zhaocheng Zhu, Zhengyan Zhang, Zhiyuan Liu, Juanzi Li, and Jian Tang. Kepler: A unified model for knowledge embedding and pre-trained language representation. Transactions of the Association for Computational Linguistics , 9:176--194, 2021

  96. [104]

    Multi-view graph contrastive representation learning for drug-drug interaction prediction

    Yingheng Wang, Yaosen Min, Xin Chen, and Ji Wu. Multi-view graph contrastive representation learning for drug-drug interaction prediction. In WWW , 2021

  97. [105]

    Mindmap: Knowledge graph prompting sparks graph of thoughts in large language models

    Yilin Wen, Zifeng Wang, and Jimeng Sun. Mindmap: Knowledge graph prompting sparks graph of thoughts in large language models. arXiv preprint arXiv:2308.09729 , 2023

  98. [106]

    DrugBank 5.0: a major update to the DrugBank database for 2018

    David S Wishart, Yannick D Feunang, An C Guo, Elvis J Lo, Ana Marcu, Jason R Grant, Tanvir Sajed, Daniel Johnson, Carin Li, Zinat Sayeeda, Nazanin Assempour, Ithayavani Iynkkaran, Yifeng Liu, Adam Maciejewski, Nicola Gale, Alex Wilson, Lucy Chin, Ryan Cummings, Diana Le, Allis...

  99. [107]

    Self-supervised hypergraph convolutional networks for session-based recommendation

    Xin Xia, Hongzhi Yin, Junliang Yu, Qinyong Wang, Lizhen Cui, and Xiangliang Zhang. Self-supervised hypergraph convolutional networks for session-based recommendation. In AAAI , pages 4503--4511, 2021

  100. [108]

    When to use graphs in rag: A comprehensive analysis for graph retrieval-augmented generation

    Zhishang Xiang, Chuanjie Wu, Qinggang Zhang, Shengyuan Chen, Zijin Hong, Xiao Huang, and Jinsong Su. When to use graphs in rag: A comprehensive analysis for graph retrieval-augmented generation. arXiv preprint arXiv:2506.05690 , 2025

  101. [109]

    Reliable reasoning path: Distilling effective guidance for llm reasoning with knowledge graphs

    Yilin Xiao, Chuang Zhou, Qinggang Zhang, Bo Li, Qing Li, and Xiao Huang. Reliable reasoning path: Distilling effective guidance for llm reasoning with knowledge graphs. arXiv preprint arXiv:2506.10508 , 2025

  102. [110]

    Does william shakespeare really write hamlet? knowledge representation learning with confidence

    Ruobing Xie, Zhiyuan Liu, Fen Lin, and Leyu Lin. Does william shakespeare really write hamlet? knowledge representation learning with confidence. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 32, 2018

  103. [111]

    Deeppath: A reinforcement learning method for knowledge graph reasoning

    Wenhan Xiong, Thien Hoang, and William Yang Wang. Deeppath: A reinforcement learning method for knowledge graph reasoning. In EMNLP , pages 564--573, 2017

  104. [112]

    Contrastive attributed network anomaly detection with data augmentation

    Zhiming Xu, Xiao Huang, Yue Zhao, Yushun Dong, and Jundong Li. Contrastive attributed network anomaly detection with data augmentation. In Pacific-Asia Conference on Knowledge Discovery and Data Mining , pages 444--457. Springer, 2022

  105. [113]

    Embedding entities and relations for learning and inference in knowledge bases

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

  106. [114]

    Evaluating world models with llm for decision making

    Chang Yang, Xinrun Wang, Junzhe Jiang, Qinggang Zhang, and Xiao Huang. Evaluating world models with llm for decision making. arXiv preprint arXiv:2411.08794 , 2024

  107. [115]

    Differentiable learning of logical rules for knowledge base completion

    Fan Yang, Zhilin Yang, and William W Cohen. Differentiable learning of logical rules for knowledge base completion. CoRR, abs/1702.08367 , 2017

  108. [116]

    Give us the facts: Enhancing large language models with knowledge graphs for fact-aware language modeling

    Linyao Yang, Hongyang Chen, Zhao Li, Xiao Ding, and Xindong Wu. Give us the facts: Enhancing large language models with knowledge graphs for fact-aware language modeling. IEEE Transactions on Knowledge and Data Engineering , 2024

  109. [117]

    Deep bidirectional language-knowledge graph pretraining

    Michihiro Yasunaga, Antoine Bosselut, Hongyu Ren, Xikun Zhang, Christopher D Manning, Percy S Liang, and Jure Leskovec. Deep bidirectional language-knowledge graph pretraining. Advances in Neural Information Processing Systems , 35:37309--37323, 2022

  110. [118]

    Qa-gnn: Reasoning with language models and knowledge graphs for question answering

    Michihiro Yasunaga, Hongyu Ren, Antoine Bosselut, Percy Liang, and Jure Leskovec. Qa-gnn: Reasoning with language models and knowledge graphs for question answering. NAACL , 2021

  111. [119]

    Graph contrastive learning with augmentations

    Yuning You, Tianlong Chen, Yongduo Sui, Ting Chen, Zhangyang Wang, and Yang Shen. Graph contrastive learning with augmentations. In NeurIPS , 2020

  112. [120]

    Knapsack optimization-based schema linking for llm-based text-to-sql generation

    Zheng Yuan, Hao Chen, Zijin Hong, Qinggang Zhang, Feiran Huang, and Xiao Huang. Knapsack optimization-based schema linking for llm-based text-to-sql generation. arXiv preprint arXiv:2502.12911 , 2025

  113. [121]

    Contrastive self-supervised learning for graph classification

    Jiaqi Zeng and Pengtao Xie. Contrastive self-supervised learning for graph classification. In AAAI , 2021

  114. [122]

    Link prediction based on graph neural networks

    Muhan Zhang and Yixin Chen. Link prediction based on graph neural networks. In Advances in Neural Information Processing Systems (NeurIPS) , 2018

  115. [123]

    Knowgpt: Knowledge graph based prompting for large language models

    Qinggang Zhang, Junnan Dong, Hao Chen, and et al. Knowgpt: Knowledge graph based prompting for large language models. In Advances in Neural Information Processing Systems (NeurIPS) , 2024

  116. [124]

    Structure guided large language model for sql generation

    Qinggang Zhang, Junnan Dong, Hao Chen, Wentao Li, Feiran Huang, and Xiao Huang. Structure guided large language model for sql generation. arXiv preprint arXiv:2402.13284 , 2024

  117. [125]

    Contrastive knowledge graph error detection

    Qinggang Zhang, Junnan Dong, Keyu Duan, Xiao Huang, Yezi Liu, and Linchuan Xu. Contrastive knowledge graph error detection. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management , pages 2590--2599, 2022

  118. [126]

    Integrating entity attributes for error-aware knowledge graph embedding

    Qinggang Zhang, Junnan Dong, Qiaoyu Tan, and Xiao Huang. Integrating entity attributes for error-aware knowledge graph embedding. IEEE Transactions on Knowledge and Data Engineering , 2023

  119. [127]

    Logical reasoning with relation network for inductive knowledge graph completion

    Qinggang Zhang, Keyu Duan, Junnan Dong, Pai Zheng, and Xiao Huang. Logical reasoning with relation network for inductive knowledge graph completion. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages 4268--4277, 2024

  120. [128]

    Faithfulrag: Fact-level conflict modeling for context-faithful retrieval-augmented generation

    Qinggang Zhang, Zhishang Xiang, Yilin Xiao, Le Wang, Junhui Li, Xinrun Wang, and Jinsong Su. Faithfulrag: Fact-level conflict modeling for context-faithful retrieval-augmented generation. arXiv preprint arXiv:2506.08938 , 2025

  121. [129]

    Greaselm: Graph reasoning enhanced language models for question answering

    Xikun Zhang, Antoine Bosselut, Michihiro Yasunaga, Hongyu Ren, Percy Liang, Christopher D Manning, and Jure Leskovec. Greaselm: Graph reasoning enhanced language models for question answering. arXiv preprint arXiv:2201.08860 , 2022

  122. [130]

    Kg-cot: Chain-of-thought prompting of large language models over knowledge graphs for knowledge-aware question answering

    Ruilin Zhao, Feng Zhao, Long Wang, Xianzhi Wang, and Guandong Xu. Kg-cot: Chain-of-thought prompting of large language models over knowledge graphs for knowledge-aware question answering. In Kate Larson, editor, Proceedings of the Thirty-Third International Joint Conference on...

  123. [131]

    How does nlp benefit legal system: A summary of legal artificial intelligence

    Haoxi Zhong, Chaojun Xiao, Cunchao Tu, Tianyang Zhang, Zhiyuan Liu, and Maosong Sun. How does nlp benefit legal system: A summary of legal artificial intelligence. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , pages 5218--5230, 2020

  124. [132]

    Adaptive popularity debiasing aggregator for graph collaborative filtering

    Huachi Zhou, Hao Chen, Junnan Dong, Daochen Zha, Chuang Zhou, and Xiao Huang. Adaptive popularity debiasing aggregator for graph collaborative filtering. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages ...

  125. [133]

    Retrieving and reading: A comprehensive survey on open-domain question answering

    Fengbin Zhu, Wenqiang Lei, Chao Wang, Jianming Zheng, Soujanya Poria, and Tat-Seng Chua. Retrieving and reading: A comprehensive survey on open-domain question answering. arXiv preprint arXiv:2101.00774 , 2021

  126. [134]

    Deep graph contrastive representation learning

    Yanqiao Zhu, Yichen Xu, Feng Yu, Qiang Liu, Shu Wu, and Liang Wang. Deep graph contrastive representation learning. arXiv preprint arXiv:2006.04131 , 2020

Pith tools

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