Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Towards Temporal Knowledge Graph Alignment in the Wild

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that real-world temporal graph alignment can be handled by a multi-scale hypergraph retrieval-augmented generation framework, reaching 0.871 Hits@1 on the new WildBETA benchmark and 0.924 on BETA.

desk verdict New benchmarks are the real contribution; HyDRA's SOTA is conditional on an unverified LLM supplementation step that can act as an oracle—still worth refereeing. read the letter →

arxiv 2507.14475 v1 pith:RPYC376D submitted 2025-07-19 cs.DB

Runhao Zhao , Weixin Zeng , Wentao Zhang , Xiang Zhao , Jiuyang Tang , Lei Chen This is my paper
classification cs.DB
keywords TemporalknowledgegraphalignmentTKGA-WildMulti-scalehypergraphRetrieval-augmentedgenerationEntitygraphsBenchmarkdatasetsLargelanguagemodels
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

Real-world temporal knowledge graphs do not behave like existing benchmarks: they mix day-, month-, and year-scale timestamps, aligned entities often have non-overlapping or nested active intervals, and one graph can be densely timestamped while its counterpart is almost bare. The paper defines these conditions as 'TKGA-Wild' and claims that they break current alignment models, which implicitly assume a single time granularity, matching intervals, and balanced temporal density. To make progress, it introduces two new benchmarks, BETA and WildBETA, that exhibit those conditions, and a framework, HyDRA, that casts alignment as multi-scale hypergraph retrieval-augmented generation. HyDRA decomposes each entity into time- and relation-masked projections, organizes them into a three-layer hypergraph, retrieves relevant projections, and lets an LLM supplement, trim, and reconcile alignment evidence across scales. The paper reports Hits@1 of 0.871 on WildBETA and 0.924 on BETA, an improvement of 43.3% and 21.3% over the best evaluated baseline, and argues this establishes a new paradigm and a fairer benchmark for temporal alignment.

What carries the argument

The load-bearing object is the multi-scale hypergraph built from adaptive entity projections. Each source entity considers the top $k$ most similar target entities in an adaptive similarity matrix; timestamps and relation types not shared with the source are masked away, producing time and relation projections (Eq. 5); a hyperedge joins the source entity to those targets and projections (Eqs. 6-8). A projection memory bank stores target projections, retrieval selects the top $k$ most relevant ones (Eq. 9), and three hypergraph layers organize fine projections, retrieved projections, and abstract target entities. The scale-weave synergy mechanism, intra-scale LLM fact supplementation and trimming plus cross-scale conflict detection, is what the paper credits with coping with temporal incompleteness and density imbalance.

What would settle it

Measure the oracle recall of the top $k$ candidate lists on WildBETA: count how often the true target entity is absent from the time and relation projection set of its source entity. If that miss rate is non-negligible, then no downstream hypergraph or LLM component can fix it. Separately, re-run HyDRA with the LLM's fact supplementation disabled or restricted to facts that appear in the source graphs; a large drop in Hits@1 on the one-to-none and none-to-none scenarios would show that the reported gains depend on unverified LLM-generated facts rather than on retrieved graph evidence.

Watch

Extended reading notes

Core claim

The central claim is that TKGA-Wild is not a noise problem but a structural one, and the right response is to stop aligning entities at a single fixed scale. HyDRA's pipeline breaks each entity into projections along time and relation dimensions, builds a projection hypergraph whose hyperedges connect a source entity to its top $k$ similar targets and their masked projections, stores those projections in a memory bank, then retrieves the most relevant projections to form a multi-scale hypergraph with three layers: a fine projection hypergraph, a retrieved-projection hypergraph, and an abstract target-entity hypergraph. A multi-scale interaction-augmented fusion stage lets an LLM perform intra-scale supplementation and trimming of temporal facts and cross-scale conflict detection, feeding resolved alignments back into the encoders for iterative refinement. The paper's experimental assertion is that this design beats 24 baselines across eight datasets, with the largest gains on the new benchmarks and on previously unstudied alignment scenarios such as one-to-none and none-to-none temporal coverage, while using far fewer tokens than general RAG baselines.

Load-bearing premise

The pipeline assumes that the true match for each entity is almost always inside the small candidate set produced by the embedding similarity matrix, and that supplemental facts the LLM adds do not need to be checked against the original graphs.

Editorial extensions

If this is right

  • The two most widely used TKGA benchmarks are nearly saturated (baseline Hits@1 often exceeds 0.97), so they reward methods that exploit a single dominant temporal relation and fail to expose the structural imbalances that real cross-source temporal graphs exhibit; BETA and WildBETA are designed to restore that discrimination.
  • On the hardest scenario classes introduced, multi-to-none, one-to-none, and none-to-none, the strongest baselines fall below 0.30 Hits@1, while HyDRA reports 0.81-0.84, implying that pure embedding models are fundamentally limited when one side lacks temporal evidence.
  • The framework remains accurate when underlying embeddings are perturbed: HyDRA degrades far less than embedding-only baselines as injected noise rises to 80%, indicating the retrieval and LLM fusion stages compensate for weak representation quality.
  • HyDRA is efficient in LLM usage: on WildBETA it consumes 301 tokens and 2.5 seconds per target entity, versus 34,295 tokens and 30.5 seconds for a self-reflective RAG baseline, while beating it by a large margin; this makes the RAG-based paradigm practical at benchmark scale.
  • Performance improves as the underlying LLM gets stronger, from a smaller local model to larger commercial models, so the framework is positioned to inherit future LLM advances rather than requiring architectural changes.

Reading between the lines

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

  • The framework's ceiling is set before the LLM is invoked: if the true counterpart is not in the top $k$ candidates of the adaptive similarity matrix, neither the hypergraph nor the retrieval stage can recover it; the reported gains therefore rest on the recall of that candidate set, not on the LLM alone.
  • The intra-scale supplementation step lets the LLM add temporal facts that are not verifiable against the original graphs, so part of the accuracy gain on sparse scenarios may come from the LLM's parametric memory rather than from genuine alignment evidence; an audit comparing supplemented facts to the source graphs would separate these contributions.
  • The same projection-hypergraph recipe is portable: swapping time/relation masking for modality or language masking would give a general multi-scale retrieval-augmented paradigm for cross-modal entity alignment, where the same incompleteness and imbalance patterns appear.
  • Because the conflict-detection step only reviews pairs that already emerged per scale, hidden conflicts, where the correct partner never surfaces in any scale's candidate list, are invisible to the mechanism; extending detection to a broader candidate pool would be a direct testable improvement.
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

4 major / 5 minor

Summary. This paper studies temporal knowledge graph alignment in the wild (TKGA-Wild), where multi-scale temporal element entanglement and cross-source temporal structural imbalances are prevalent. The authors propose HyDRA, a multi-scale hypergraph retrieval-augmented generation framework that combines structural/temporal encoders, scale-adaptive projection, multi-scale hypergraph retrieval, and an LLM-based multi-scale fusion module. They also introduce two new benchmarks, BETA and WildBETA, built from Wikidata and YAGO, which reflect multi-granular timestamps, incomplete temporal annotations, and imbalanced temporal densities. Experiments on these and six existing datasets claim consistent state-of-the-art Hits@1 results, e.g., 0.871 on WildBETA and 0.924 on BETA, outperforming the best evaluated baseline Simple-HHEA by 43.3% and 21.3%.

Significance. The paper makes a valuable empirical contribution by identifying and operationalizing a realistic variant of TKGA and by releasing two challenging benchmarks. The proposed framework is complex and the experimental effort is substantial (24 baselines across 8 datasets). If the reported results are robust, HyDRA would be a strong new result in the area. The evaluation design (held-out ground truth at the final stage) avoids blatant circularity, and the paper provides concrete statistics (e.g., temporal completeness and density imbalance) to characterize the new datasets. However, the headline claims rest on several load-bearing points that need strengthening: the strongest LLM baseline (ChatEA) is absent from the new benchmarks, the LLM-based time-fact supplementation is unverified and may allow the model to import external world knowledge rather than solve the graph-based task, and the dataset construction description contains numerical inconsistencies that impede reproducibility. The paper also reports no error bars for stochastic LLM-based runs.

major comments (4)
  1. [Section 5.1, Tables 4 and 5] ChatEA, described in Section 2 as an LLM-based entity alignment method applicable to TKGA, is evaluated on ICEWS-WIKI, ICEWS-YAGO, DBP15K(EN-FR), and DBP-WIKI (Table 5) but not on WildBETA or BETA (Table 4). On the datasets where it appears, ChatEA is the strongest baseline (e.g., Hits@1 of 0.880 on ICEWS-WIKI and 0.935 on ICEWS-YAGO). Its absence on the two new benchmarks undercuts the support for the claim of 'consistently outperforming 24 competitive baselines' and for the specific statement that Simple-HHEA is the 'best evaluated baseline' on WildBETA/BETA. Please add ChatEA results (or justify its exclusion) on the new datasets, including its Hits@1 and MRR, and update the corresponding improvement claims if needed.
  2. [Section 3.6 (Intra-Scale Interaction) and Eqs. (6), (9)] The intra-scale interaction operation lets the LLM 'supplement the missing core temporal information (including time, relations, or facts)' with no verification against the original source TKGs. Because the projection hypergraph (Eq. (6)) and the retrieval memory bank (Eq. (9)) are built from top-k candidates of the adaptive similarity matrix, any supplemented facts enter the evidence pool even if they are hallucinated. This creates a channel through which the LLM's parametric knowledge of real-world entities, rather than the proposed hypergraph-RAG machinery, could drive the reported 0.871/0.924 Hits@1. The ablation in Table 6 removes the whole multi-scale interaction-augmented fusion module (- w/o Multi-Scale Interaction-Augmented Fusion), but not the supplementation operation alone. Please quantify the fraction of supplemented facts that are actually present in the source TKGs, report a variant that disables supplementation while keeping the rest of the fusion module, and/or add a verification step that checks supplemented facts against the input quadruples before they are used as evidence.
  3. [Section 4.1 (Dataset Construction)] The construction pipeline for BETA is numerically inconsistent. The text first states that extracting quadruples associated with the preliminary seeds yields 131,853 quadruples from Wikidata and 89,702 from YAGO 4.5; it then says that after screening 'the entities that exist in the mapping' there are 40,364 pairs, and that the final KG subsets contain 199,879 Wikidata quadruples and 162,320 YAGO quadruples. Since the final numbers are larger than the initial ones, and since the preliminary seed count is 26,594 pairs while the reported final seed count is 40,364 pairs, the pipeline cannot be reproduced as written. Please correct the sequence of numbers and clarify whether 'pairs' refers to aligned entity pairs every time; this is essential for validating the benchmark contribution (Table 3 depends on these counts).
  4. [Section 5.1 and Tables 4, 5, 6, 9] All reported results appear to come from a single run of each condition. Given that the method relies on stochastic LLM sampling (GPT-3.5/GPT-4) and that the headline improvements are large (43.3% and 21.3%), the absence of error bars or multiple seeds makes it impossible to assess statistical reliability. Please provide results over several runs (or bootstrapped confidence intervals) for at least the main tables, and state the variance of the GPT-4-based results in Table 8.
minor comments (5)
  1. [Equation (4)] The symbol \otimes is used where feature concatenation is intended; concatenation is more commonly denoted by \oplus or [·;·]; please clarify the operator.
  2. [Section 4.2] The paragraph about 'Temporal Interval Topological Disparity' states that in WildBETA only 5.27% of entity pairs share consistent intervals, but it does not define how consistency is measured; please include the metric definition or cite the technical report.
  3. [Tables 3, 4, and 5] Some table entries contain stray characters or placeholder symbols (e.g., 'Temp.' column entries are '%'), which look like formatting errors; please clean the tables.
  4. [Section 5.1] Several experimental details (e.g., hyperparameter values, the construction of WildBETA's non-1-to-1 sampling, and metric formulations) are deferred to a 'technical report' that is not included; please provide the report as supplementary material or include the details in the paper.
  5. [Section 3.3] The justification for the scale-adaptive projection module refers to 'the experimental results of previous work MGTEA [17]' but does not reproduce or summarize that evidence; consider adding a sentence describing the specific limitation observed.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the SOTA claim rests on held-out evaluation; the MGTEA self-citation is not load-bearing, and the unverified LLM supplementation is a validity risk, not a circular step.

full rationale

The claimed derivation chain—multi-granular encoders, adaptive similarity, projection hypergraph, multi-scale retrieval, and interaction-augmented fusion—is evaluated on held-out alignment pairs from BETA/WildBETA and six existing benchmarks, so the central SOTA claim does not reduce by construction to a fitted value. The iterative update uses pseudo-pairs generated from the model's own similarity matrix, but final Hits@1/MRR are reported against ground-truth alignments, not against the self-training signal. The paper does extend the authors' prior MGTEA [17] and cites it in Sec. 3.4 to justify that simple multi-granular integration is insufficient; this self-citation is not load-bearing because the same conclusion is independently supported by the paper's own ablations (Table 6: removing the multi-granular encoder, scale-adaptive projection, multi-scale retrieval, or interaction-augmented fusion degrades Hits@1 from 0.859 to between 0.611 and 0.812). The most serious evidentiary concern—Sec. 3.6's LLM 'time fact supplementation' that 'allows the LLM to analyze the entity pair's situation and supplement the missing core temporal information (including time, relations, or facts)' with no verification against the source TKGs—is a potential external-oracle/hallucination risk to construct validity, not a circular derivation: the final alignment is still selected from candidate pairs produced by the learned similarity/hypergraph pipeline and scored against held-out labels. No equation in the paper defines the output in terms of the label it predicts, and no fitted parameter is renamed as a prediction. I therefore find no significant circularity; the minor self-citation to MGTEA and the LLM-supplementation limitation are correctness risks, not circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on dataset construction assumptions (prior mapping equals ground truth), a bootstrap assumption (pseudo-pairs seed the hypergraph), and an unverified LLM assumption (supplemented facts are accepted as true). No new physical entities are introduced.

free parameters (5)
  • Random walk balance beta
    Controls breadth versus depth in the meta structural encoder (Eq. 1); value not reported.
  • Top-k for projection and retrieval
    Number of candidate target entities kept per source entity in Eq. (6) and Eq. (9); value not reported.
  • Temporal granularity set = year, month, date
    Hand-chosen decomposition of time into three scales; other granularities are excluded.
  • Iteration count I_n
    Number of iterative refinement rounds; value not reported.
  • Margin ranking loss margin
    Training objective hyperparameter; value not reported.
assumptions (4)
  • domain assumption The YAGO-to-Wikidata QID mapping used in Section 4.1 is assumed to be a correct and complete source of entity equivalence and non-equivalence.
    Used to generate preliminary seeds and final alignment pairs for BETA and WildBETA.
  • domain assumption Facts whose time interval tau is not none are considered valid temporal facts, and timestamps can be split into year, month, and date.
    Definition 1 and Section 3.3 make this structural assumption about the input TKGs.
  • ad hoc to paper Pseudo-aligned pairs generated from the adaptive similarity matrix are reliable enough to bootstrap the projection hypergraph and iterative self-training.
    Section 3.2 and Section 3.4 use these pairs as the foundation of the multi-scale hypergraph.
  • ad hoc to paper LLM-supplemented temporal facts are treated as valid evidence for alignment without being checked against the original TKGs.
    Section 3.6 intra-scale interaction relies on this unverified premise, which could inflate alignment scores.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Temporal Knowledge Graph Alignment in the Wild." pith.science (2026). https://pith.science/paper/RPYC376D

@misc{pith2026250714475,
  author       = {Pith},
  title        = {Pith review of: Towards Temporal Knowledge Graph Alignment in the Wild},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RPYC376D}},
  note         = {Machine review of arXiv:2507.14475}
}
read the original abstract

Temporal Knowledge Graph Alignment (TKGA) seeks to identify equivalent entities across heterogeneous temporal knowledge graphs (TKGs) for fusion to improve their completeness. Although some approaches have been proposed to tackle this task, most assume unified temporal element standards and simplified temporal structures across different TKGs. They cannot deal with TKGA in the wild (TKGA-Wild), where multi-scale temporal element entanglement and cross-source temporal structural imbalances are common. To bridge this gap, we study the task of TKGA-Wild and propose HyDRA, a new and effective solution. HyDRA is the first to reformulate the task via multi-scale hypergraph retrieval-augmented generation to address the challenges of TKGA-Wild.In addition, we design a new scale-weave synergy mechanism for HyDRA, which incorporates intra-scale interactions and cross-scale conflict detection. This mechanism is designed to alleviate the fragmentation caused by multi-source temporal incompleteness and resolves inconsistencies arising from complex and uneven temporal event density distributions, thereby enhancing the model capacity to handle the intricacies of real-world temporal alignment. Finally, there is no standard benchmark that captures these challenges of TKGA-Wild and effectively evaluates existing methods. To this end, we formally propose to benchmark challenges for TKGA-Wild and validate the effectiveness of the method by establishing two new datasets(BETA and WildBETA). Extensive experiments on the new datasets and six representative benchmarks show that BETA and WildBETA better reflect real-world challenges. Meanwhile, HyDRA proposes a new paradigm for TKGA-Wild, consistently outperforming 24 competitive baselines, while maintaining strong efficiency and scalability.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Panning for Gold: Expanding Domain-Specific Knowledge Graphs with General Knowledge

    cs.AI 2026-01 reject novelty 6.0 of 10

    ExeFuse uses learned 'logic' transformations and density checks to fuse general-graph facts into domain knowledge graphs, but the benchmark labels and baseline comparisons are too underspecified to support the claimed gains.

Reference graph

Works this paper leans on

84 extracted references · 58 canonical work pages · cited by 1 Pith paper

  1. [1]

    RetrievalQA: Assessing Adaptive Retrieval-Augmented Generation for Short-form Open-Domain Question Answering

    Z. Zhang, M. Fang, and L. Chen, “Retrievalqa: Assessing adaptive retrieval-augmented generation for short-form open-domain question answering,” CoRR, vol. abs/2402.16457, 2024. [Online]. Available: https://doi.org/10.48550/arXiv.2402.16457

    work page Pith review arXiv doi:10.48550/arxiv.2402.16457 2024

  2. [2]

    Inductive meta-path learning for schema-complex heterogeneous information networks,

    S. Liu, C. Fan, K. Cheng, Y . Wang, P. Cui, Y . Sun, and Z. Liu, “Inductive meta-path learning for schema-complex heterogeneous information networks,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 46, no. 12, pp. 10 196–10 209, 2024. [Online]. Available: https://doi.org/10.1109/TPAMI.2024.3435055

    work page arXiv 2024

  3. [3]

    INFER: A neural-symbolic model for extrapolation reasoning on temporal knowledge graph,

    N. Li, H. E, T. Yao, T. Hu, Y . Li, H. Luo, M. Song, and Y . Zhu, “INFER: A neural-symbolic model for extrapolation reasoning on temporal knowledge graph,” in The Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025 . OpenReview.net, 2025. [Online]. Available: https://openreview.net/forum?id=ExHUtB2vnz

    2025

  4. [5]

    Time-aware entity alignment using temporal relational attention,

    C. Xu, F. Su, B. Xiong, and J. Lehmann, “Time-aware entity alignment using temporal relational attention,” in WWW ’22: The ACM Web Con- ference 2022, Virtual Event, Lyon, France, April 25 - 29, 2022 . ACM, 2022, pp. 788–797

    2022

  5. [6]

    Time-aware graph neural network for entity alignment between temporal knowledge graphs,

    C. Xu, F. Su, and J. Lehmann, “Time-aware graph neural network for entity alignment between temporal knowledge graphs,” in Proceedings of the 2021 Conference on Empirical Methods in Natural Language Pro- cessing, EMNLP 2021, Virtual Event / Punta Cana, Dominican Republic, 7-11 November, 2021. Association for Computational Linguistics, 2021, pp. 8999–9010

    2021

  6. [7]

    An effective and efficient time-aware entity alignment framework via two-aspect three-view label propagation,

    L. Cai, X. Mao, Y . Xiao, C. Wu, and M. Lan, “An effective and efficient time-aware entity alignment framework via two-aspect three-view label propagation,” in Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, IJCAI 2023, 19th-25th August 2023, Macao, SAR, China. ijcai.org, 2023, pp. 5021–5029

    2023

  7. [8]

    Unsupervised entity alignment for temporal knowledge graphs,

    X. Liu, J. Wu, T. Li, L. Chen, and Y . Gao, “Unsupervised entity alignment for temporal knowledge graphs,” in Proceedings of the ACM Web Conference 2023, WWW 2023, Austin, TX, USA, 30 April 2023 - 4 May 2023 , Y . Ding, J. Tang, J. F. Sequeda, L. Aroyo, C. Castillo, and G. Houben, Eds. ACM, 2023, pp. 2528–2538. [Online]. Available: https://doi.org/10.1145...

    arXiv 2023

  8. [9]

    Toward practical entity alignment method design: Insights from new highly heterogeneous knowledge graph datasets,

    X. Jiang, C. Xu, Y . Shen, Y . Wang, F. Su, Z. Shi, F. Sun, Z. Li, J. Guo, and H. Shen, “Toward practical entity alignment method design: Insights from new highly heterogeneous knowledge graph datasets,” in Proceedings of the ACM on Web Conference 2024, WWW 2024, Singapore, May 13-17, 2024 , T. Chua, C. Ngo, R. Kumar, H. W. Lauw, and R. K. Lee, Eds. ACM, ...

    work page arXiv 2024

Show all 84 references
  1. [10]

    Temporal knowledge graph entity alignment via representation learning,

    X. Song, L. Bai, R. Liu, and H. Zhang, “Temporal knowledge graph entity alignment via representation learning,” in Database Systems for Advanced Applications - 27th International Conference, DASFAA 2022, Virtual Event, April 11-14, 2022, Proceedings, Part II, ser. Lecture Note...

    2022

  2. [11]

    Enhancing knowledge graph attention by temporal modeling for entity alignment with sparse seeds,

    C. Sun, Y . Jin, D. Shen, T. Nie, X. Wang, and Y . Xiao, “Enhancing knowledge graph attention by temporal modeling for entity alignment with sparse seeds,” in Database Systems for Advanced Applications - 28th International Conference, DASFAA 2023, Tianjin, China, April 17- 20,...

    2023

  3. [12]

    A simple temporal information matching mechanism for entity alignment between temporal knowledge graphs,

    L. Cai, X. Mao, M. Ma, H. Yuan, J. Zhu, and M. Lan, “A simple temporal information matching mechanism for entity alignment between temporal knowledge graphs,” in Proceedings of the 29th International Conference on Computational Linguistics, COLING 2022, Gyeongju, Republic of K...

    2022

  4. [13]

    Large language models-guided dynamic adaptation for temporal knowledge graph reasoning,

    J. Wang, K. Sun, L. Luo, W. Wei, Y . Hu, A. W.-C. Liew, S. Pan, and B. Yin, “Large language models-guided dynamic adaptation for temporal knowledge graph reasoning,” in Advances in Neural Information Processing Systems , A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet,...

    2024

  5. [14]

    Y AGO 4.5: A large and clean knowledge base with a rich taxonomy,

    F. M. Suchanek, M. Alam, T. Bonald, L. Chen, P. Paris, and J. Soria, “Y AGO 4.5: A large and clean knowledge base with a rich taxonomy,” in Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2024, Washington D...

    2024

  6. [15]

    Yago: a core of semantic knowledge,

    F. M. Suchanek, G. Kasneci, and G. Weikum, “Yago: a core of semantic knowledge,” in Proceedings of the 16th International Conference on World Wide Web, WWW 2007, Banff, Alberta, Canada, May 8-12, 2007 . ACM, 2007, pp. 697–706

    2007

  7. [16]

    Tackling sparse facts for temporal knowledge graph completion,

    Y . Zhang, X. Kong, K. Ye, G. Shen, and S. Zheng, “Tackling sparse facts for temporal knowledge graph completion,” in Proceedings of the ACM on Web Conference 2025, WWW 2025, Sydney, NSW, Australia, 28 April 2025- 2 May 2025, G. Long, M. Blumestein, Y . Chang, L. Lewin-Eytan, ...

    2025

  8. [17]

    Benchmarking challenges for temporal knowledge graph alignment,

    W. Zeng, J. Zhou, and X. Zhao, “Benchmarking challenges for temporal knowledge graph alignment,”Proceedings of the 33rd ACM International Conference on Information and Knowledge Management, 2024. [Online]. Available: https://api.semanticscholar.org/CorpusID:273501043

    2024

  9. [18]

    Unlocking the power of large language models for entity alignment,

    X. Jiang, Y . Shen, Z. Shi, C. Xu, W. Li, Z. Li, J. Guo, H. Shen, and Y . Wang, “Unlocking the power of large language models for entity alignment,” in Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2024, Ba...

    2024

  10. [19]

    Introducing wikidata to the linked data web,

    F. Erxleben, M. G ¨unther, M. Kr ¨otzsch, J. Mendez, and D. Vrandecic, “Introducing wikidata to the linked data web,” in The Semantic Web - ISWC 2014 - 13th International Semantic Web Conference, Riva del Garda, Italy, October 19-23, 2014. Proceedings, Part I , ser. Lecture No...

    2014

  11. [22]

    Towards semantic consistency: Dirichlet energy driven robust multi-modal entity alignment,

    Y . Wang, H. Sun, J. Wang, J. Wang, W. Tang, Q. Qi, S. Sun, and J. Liao, “Towards semantic consistency: Dirichlet energy driven robust multi-modal entity alignment,” in 40th IEEE International Conference on Data Engineering, ICDE 2024, Utrecht, The Netherlands, May 13-16, 2024...

    2024

  12. [23]

    An experimental study of state-of-the-art entity alignment approaches,

    X. Zhao, W. Zeng, J. Tang, W. Wang, and F. M. Suchanek, “An experimental study of state-of-the-art entity alignment approaches,”IEEE Trans. Knowl. Data Eng., vol. 34, no. 6, pp. 2610–2625, 2022

    2022

  13. [24]

    X. Zhao, W. Zeng, and J. Tang, Entity Alignment - Concepts, Recent Advances and Novel Approaches . Springer, 2023. [Online]. Available: https://doi.org/10.1007/978-981-99-4250-3

    2023 doi

  14. [25]

    A benchmarking study of embedding-based entity alignment for knowledge graphs,

    Z. Sun, Q. Zhang, W. Hu, C. Wang, M. Chen, F. Akrami, and C. Li, “A benchmarking study of embedding-based entity alignment for knowledge graphs,” Proc. VLDB Endow., vol. 13, no. 11, pp. 2326–2340, 2020

    2020

  15. [26]

    Entity and relation matching consensus for entity alignment,

    J. Yang, D. Wang, W. Zhou, W. Qian, X. Wang, J. Han, and S. Hu, “Entity and relation matching consensus for entity alignment,” in CIKM. ACM, 2021, pp. 2331–2341

    2021

  16. [27]

    A benchmark and comprehensive survey on knowledge graph entity alignment via representation learning,

    R. Zhang, B. D. Trisedya, M. Li, Y . Jiang, and J. Qi, “A benchmark and comprehensive survey on knowledge graph entity alignment via representation learning,” VLDB J., vol. 31, no. 5, pp. 1143–1168, 2022

    2022

  17. [28]

    HLMEA: unsupervised entity alignment based on hybrid language models,

    X. Jin, Z. Wang, J. Chen, L. Yang, B. Oh, S. Hwang, and J. Li, “HLMEA: unsupervised entity alignment based on hybrid language models,” in AAAI-25, Sponsored by the Association for the Advancement of Artificial Intelligence, February 25 - March 4, 2025, Philadelphia, PA, USA , ...

    2025 doi

  18. [29]

    SE-GNN: seed expanded-aware graph neural network with iterative optimization for semi-supervised entity alignment,

    T. Meng, S. Shan, H. Shao, Y . Shou, W. Ai, and K. Li, “SE-GNN: seed expanded-aware graph neural network with iterative optimization for semi-supervised entity alignment,” IEEE Trans. Knowl. Data Eng., vol. 37, no. 6, pp. 3700–3713, 2025. [Online]. Available: https://doi.org/1...

    2025

  19. [30]

    Unsupervised robust cross-lingual entity alignment via neighbor triple matching with entity and relation texts,

    S. Yoon, S. Ko, T. Kim, S. Kang, J. Yeo, and D. Lee, “Unsupervised robust cross-lingual entity alignment via neighbor triple matching with entity and relation texts,” in Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining, WSDM 2025, Hannov...

    2025

  20. [31]

    Multilingual knowledge graph embeddings for cross-lingual knowledge alignment,

    M. Chen, Y . Tian, M. Yang, and C. Zaniolo, “Multilingual knowledge graph embeddings for cross-lingual knowledge alignment,” in IJCAI, 2017, pp. 1511–1517

    2017

  21. [32]

    Bootstrapping entity alignment with knowledge graph embedding,

    Z. Sun, W. Hu, Q. Zhang, and Y . Qu, “Bootstrapping entity alignment with knowledge graph embedding,” inProceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence, IJCAI 2018, July 13-19, 2018, Stockholm, Sweden. ijcai.org, 2018, pp. 4396–4402

    2018

  22. [33]

    Translating embeddings for modeling multi-relational data,

    A. Bordes, N. Usunier, A. Garcia-Duran, J. Weston, and O. Yakhnenko, “Translating embeddings for modeling multi-relational data,”Adv Neural Inf Process Syst, vol. 26, 2013

    2013

  23. [34]

    Hybea: Hybrid attention models for entity alignment,

    N. Fanourakis, F. Lekbour, V . Efthymiou, G. Renton, and V . Christophides, “Hybea: Hybrid attention models for entity alignment,” CoRR, vol. abs/2407.02862, 2024. [Online]. Available: https://doi.org/10.48550/arXiv.2407.02862

    2024 doi arXiv

  24. [35]

    A benchmarking study of embedding-based entity alignment for knowledge graphs,

    Z. Sun, Q. Zhang, W. Hu, C. Wang, M. Chen, F. Akrami, and C. Li, “A benchmarking study of embedding-based entity alignment for knowledge graphs,” Proc. VLDB Endow. , vol. 13, no. 11, pp. 2326–2340, 2020. [Online]. Available: http://www.vldb.org/pvldb/vol13/p2326-sun.pdf

    2020

  25. [36]

    Cross-lingual entity alignment via joint attribute-preserving embedding,

    Z. Sun, W. Hu, and C. Li, “Cross-lingual entity alignment via joint attribute-preserving embedding,” in ISWC, 2017, pp. 628–644

    2017

  26. [37]

    Make it easy: An effective end-to-end entity alignment framework,

    C. Ge, X. Liu, L. Chen, B. Zheng, and Y . Gao, “Make it easy: An effective end-to-end entity alignment framework,” in SIGIR. ACM, 2021, pp. 777–786

    2021

  27. [38]

    Cross-lingual knowledge graph alignment via graph convolutional networks,

    Z. Wang, Q. Lv, X. Lan, and Y . Zhang, “Cross-lingual knowledge graph alignment via graph convolutional networks,” in Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Belgium, October 31 - November 4, 2018 . Association for Comp...

    2018

  28. [39]

    Integrating manifold knowledge for global entity linking with heterogeneous graphs,

    Z. Chen, Y . Wu, Y . Feng, and D. Zhao, “Integrating manifold knowledge for global entity linking with heterogeneous graphs,” Data Intelligence, vol. 4, no. 1, pp. 20–40, 2022

    2022

  29. [40]

    Boosting the speed of entity alignment 10 ×: Dual attention matching network with normalized hard sample mining,

    X. Mao, W. Wang, Y . Wu, and M. Lan, “Boosting the speed of entity alignment 10 ×: Dual attention matching network with normalized hard sample mining,” in WWW ’21: The Web Conference 2021, Virtual Event / Ljubljana, Slovenia, April 19-23, 2021 . ACM / IW3C2, 2021, pp. 821–832

    2021

  30. [41]

    MRAEA: an efficient and robust entity alignment approach for cross-lingual knowledge graph,

    X. Mao, W. Wang, H. Xu, M. Lan, and Y . Wu, “MRAEA: an efficient and robust entity alignment approach for cross-lingual knowledge graph,” in WSDM ’20: The Thirteenth ACM International Conference on Web Search and Data Mining, Houston, TX, USA, February 3-7, 2020. ACM, 2020, pp...

    2020

  31. [42]

    Relational reflection entity alignment,

    X. Mao, W. Wang, H. Xu, Y . Wu, and M. Lan, “Relational reflection entity alignment,” in CIKM ’20: The 29th ACM International Conference on Information and Knowledge Management, Virtual Event, Ireland, October 19-23, 2020. ACM, 2020, pp. 1095–1104

    2020

  32. [43]

    FuAlign: Cross-lingual entity alignment via multi-view representation learning of fused knowledge graphs,

    C. Wang, Z. Huang, Y . Wan, J. Wei, J. Zhao, and P. Wang, “FuAlign: Cross-lingual entity alignment via multi-view representation learning of fused knowledge graphs,” Inform. Fusion, vol. 89, pp. 41–52, Jan. 2023. [Online]. Available: https://doi.org/10.1016/j.inffus.2022.08.002

    2023 doi

  33. [44]

    BERT-INT: A bert-based interaction model for knowledge graph alignment,

    X. Tang, J. Zhang, B. Chen, Y . Yang, H. Chen, and C. Li, “BERT-INT: A bert-based interaction model for knowledge graph alignment,” in Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI 2020, C. Bessiere, Ed. ijcai.org, 2020, pp. 3...

    2020 doi

  34. [45]

    Entity alignment with noisy annotations from large language models,

    S. Chen, Q. Zhang, J. Dong, W. Hua, Q. Li, and X. Huang, “Entity alignment with noisy annotations from large language models,” in Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC,...

    2024

  35. [46]

    Two heads are better than one: Integrating knowledge from knowledge graphs and large language models for entity alignment,

    L. Yang, H. Chen, X. Wang, J. Yang, F.-Y . Wang, and H. Liu, “Two heads are better than one: Integrating knowledge from knowledge graphs and large language models for entity alignment,” ArXiv, vol. abs/2401.16960, 2024. [Online]. Available: https://api.semanticscholar.org/Corp...

    2024 arXiv

  36. [47]

    BERT: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of deep bidirectional transformers for language understanding,” in Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologi...

    2019

  37. [48]

    Paris: Probabilistic alignment of relations, instances, and schema,

    F. M. Suchanek, S. Abiteboul, and P. Senellart, “Paris: Probabilistic alignment of relations, instances, and schema,” Proceedings of the VLDB Endowment, vol. 5, no. 3, 2011

    2011

  38. [49]

    Matching knowledge graphs in entity embedding spaces: An experimental study,

    W. Zeng, X. Zhao, Z. Tan, J. Tang, and X. Cheng, “Matching knowledge graphs in entity embedding spaces: An experimental study,” IEEE Trans. Knowl. Data Eng., vol. 35, no. 12, pp. 12 770–12 784, 2023

    2023

  39. [50]

    Collective entity alignment via adaptive features,

    W. Zeng, X. Zhao, J. Tang, and X. Lin, “Collective entity alignment via adaptive features,” in 36th IEEE International Conference on Data Engineering, ICDE 2020, Dallas, TX, USA, April 20-24, 2020 . IEEE, 2020, pp. 1870–1873. [Online]. Available: https://doi.org/10.1109/ICDE48...

    2020

  40. [51]

    High-quality task division for large-scale entity alignment,

    B. Liu, W. Hua, G. Zuccon, G. Zhao, and X. Zhang, “High-quality task division for large-scale entity alignment,” in CIKM. ACM, 2022, pp. 1258–1268

    2022

  41. [52]

    Clusterea: Scalable entity alignment with stochastic training and normalized mini-batch similarities,

    Y . Gao, X. Liu, J. Wu, T. Li, P. Wang, and L. Chen, “Clusterea: Scalable entity alignment with stochastic training and normalized mini-batch similarities,” in KDD. ACM, 2022, pp. 421–431

    2022

  42. [53]

    On entity alignment at scale,

    W. Zeng, X. Zhao, X. Li, J. Tang, and W. Wang, “On entity alignment at scale,” VLDB J. , vol. 31, no. 5, pp. 1009–1033, 2022. [Online]. Available: https://doi.org/10.1007/s00778-021-00703-3

    2022 doi

  43. [54]

    Reinforced active entity alignment,

    W. Zeng, X. Zhao, J. Tang, and C. Fan, “Reinforced active entity alignment,” in CIKM ’21: The 30th ACM International Conference on Information and Knowledge Management, Virtual Event, Queensland, Australia, November 1 - 5, 2021. ACM, 2021, pp. 2477–2486. SHELL zhao et al.: TEM...

    2021

  44. [55]

    Interactive contrastive learning for self-supervised entity alignment,

    K. Zeng, Z. Dong, L. Hou, Y . Cao, M. Hu, J. Yu, X. Lv, L. Cao, X. Wang, H. Liu, Y . Huang, J. Feng, J. Wan, J. Li, and L. Feng, “Interactive contrastive learning for self-supervised entity alignment,” in CIKM. ACM, 2022, pp. 2465–2475

    2022

  45. [56]

    Leveraging neighborhood distance awareness for entity alignment in temporal knowledge graphs,

    L. Zhu, G. Li, and L. Bai, “Leveraging neighborhood distance awareness for entity alignment in temporal knowledge graphs,” Neural Networks , vol. 185, p. 107181, 2025. [Online]. Available: https://doi.org/10.1016/j.neunet.2025.107181

    2025

  46. [57]

    Embedding- based entity alignment of cross-lingual temporal knowledge graphs,

    L. Bai, N. Li, G. Li, Z. Zhang, and L. Zhu, “Embedding- based entity alignment of cross-lingual temporal knowledge graphs,” Neural Networks , vol. 172, p. 106143, 2024. [Online]. Available: https://doi.org/10.1016/j.neunet.2024.106143

    2024

  47. [58]

    A relation enhanced model for temporal knowledge graph alignment,

    Z. Wang, X. You, and X. Lv, “A relation enhanced model for temporal knowledge graph alignment,” J. Supercomput. , vol. 80, no. 5, pp. 5733–5755, 2024. [Online]. Available: https://doi.org/10.1007/ s11227-023-05670-w

    2024

  48. [59]

    TEA: time-aware entity alignment in knowledge graphs,

    Y . Liu, W. Hua, K. Xin, S. Hosseini, and X. Zhou, “TEA: time-aware entity alignment in knowledge graphs,” in Proceedings of the ACM Web Conference 2023, WWW 2023, Austin, TX, USA, 30 April 2023 - 4 May

    2023

  49. [60]

    T2TD: text-3d generation model based on prior knowledge guidance,

    W. Nie, R. Chen, W. Wang, B. Lepri, and N. Sebe, “T2TD: text-3d generation model based on prior knowledge guidance,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 47, no. 1, pp. 172–189, 2025. [Online]. Available: https://doi.org/10.1109/TPAMI.2024.3463753

    2025

  50. [61]

    Self-rag: Learning to retrieve, generate, and critique through self-reflection,

    A. Asai, Z. Wu, Y . Wang, A. Sil, and H. Hajishirzi, “Self-rag: Learning to retrieve, generate, and critique through self-reflection,” in The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024 . OpenReview.net, 2024. [Onlin...

    2024

  51. [62]

    A survey of graph retrieval-augmented generation for customized large language models,

    Q. Zhang, S. Chen, Y . Bei, Z. Yuan, H. Zhou, Z. Hong, J. Dong, H. Chen, Y . Chang, and X. Huang, “A survey of graph retrieval-augmented generation for customized large language models,” CoRR, vol. abs/2501.13958, 2025. [Online]. Available: https://doi.org/10.48550/arXiv.2501.13958

    2025 doi

  52. [63]

    Retrieval augmented language model pre-training,

    K. Guu, K. Lee, Z. Tung, P. Pasupat, and M. Chang, “Retrieval augmented language model pre-training,” in Proceedings of the 37th International Conference on Machine Learning, ICML 2020, 13-18 July 2020, Virtual Event, ser. Proceedings of Machine Learning Research, vol

    2020

  53. [64]

    Unsupervised dense information retrieval with contrastive learning,

    G. Izacard, M. Caron, L. Hosseini, S. Riedel, P. Bojanowski, A. Joulin, and E. Grave, “Unsupervised dense information retrieval with contrastive learning,” Trans. Mach. Learn. Res. , vol. 2022, 2022. [Online]. Available: https://openreview.net/forum?id=jKN1pXi7b0

    2022

  54. [65]

    Radio: Real-time hallucination detection with contextual index optimized query formulation for dynamic retrieval augmented generation,

    J. Zhu, H. Guo, W. Shi, Z. Chen, and P. D. Meo, “Radio: Real-time hallucination detection with contextual index optimized query formulation for dynamic retrieval augmented generation,” in AAAI-25, Sponsored by the Association for the Advancement of Artificial Intelligence, Feb...

    2025 doi

  55. [66]

    Understand what LLM needs: Dual preference alignment for retrieval-augmented generation,

    G. Dong, Y . Zhu, C. Zhang, Z. Wang, J. Wen, and Z. Dou, “Understand what LLM needs: Dual preference alignment for retrieval-augmented generation,” in Proceedings of the ACM on Web Conference 2025, WWW 2025, Sydney, NSW, Australia, 28 April 2025- 2 May 2025 , G. Long, M. Blume...

    2025

  56. [67]

    Agentic retrieval-augmented generation for time series analysis,

    C. Ravuru, S. S. Srinivas, and V . Runkana, “Agentic retrieval-augmented generation for time series analysis,” CoRR, vol. abs/2408.14484, 2024. [Online]. Available: https://doi.org/10.48550/arXiv.2408.14484

    2024 doi arXiv

  57. [68]

    Rakg: Document-level retrieval augmented knowledge graph construc- tion,

    H. Zhang, J. Si, G. Yan, B. Qi, P. Cai, S. Mao, D. Wang, and B. Shi, “Rakg: Document-level retrieval augmented knowledge graph construc- tion,” arXiv preprint arXiv:2504.09823, 2025

    2025 arXiv

  58. [69]

    Learning to cut via hierarchical sequence/set model for efficient mixed-integer programming,

    J. Wang, Z. Wang, X. Li, Y . Kuang, Z. Shi, F. Zhu, M. Yuan, J. Zeng, Y . Zhang, and F. Wu, “Learning to cut via hierarchical sequence/set model for efficient mixed-integer programming,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 46, no. 12, pp. 9697–9713, 2024. [Online]. ...

    2024

  59. [70]

    Retrieval-augmented generation for large language model based few-shot chinese spell checking,

    M. Dong, Z. Cheng, C. Luo, and T. He, “Retrieval-augmented generation for large language model based few-shot chinese spell checking,” in Proceedings of the 31st International Conference on Computational Linguistics, COLING 2025, Abu Dhabi, UAE, January 19-24, 2025 , O. Rambow...

    2025

  60. [71]

    Self-consistency improves chain of thought reasoning in language models,

    X. Wang, J. Wei, D. Schuurmans, Q. V . Le, E. H. Chi, S. Narang, A. Chowdhery, and D. Zhou, “Self-consistency improves chain of thought reasoning in language models,” in The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023...

    2023

  61. [72]

    Query rewriting in retrieval-augmented large language models,

    X. Ma, Y . Gong, P. He, H. Zhao, and N. Duan, “Query rewriting in retrieval-augmented large language models,” in Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023 , H. Bouamor, J. Pino, and K. Bal...

    2023 doi

  62. [73]

    Speculative RAG: enhancing retrieval augmented generation through drafting,

    Z. Wang, Z. Wang, L. T. Le, H. S. Zheng, S. Mishra, V . Perot, Y . Zhang, A. Mattapalli, A. Taly, J. Shang, C. Lee, and T. Pfister, “Speculative RAG: enhancing retrieval augmented generation through drafting,” in The Thirteenth International Conference on Learning Representati...

    2025

  63. [74]

    Retrieval-augmented hypergraph for multimodal social media popularity prediction,

    Z. Cheng, J. Zhang, X. Xu, G. Trajcevski, T. Zhong, and F. Zhou, “Retrieval-augmented hypergraph for multimodal social media popularity prediction,” in Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2024, Barcelona, Spain, August 25-2...

    2024

  64. [75]

    Tensor decompositions for temporal knowledge base completion,

    T. Lacroix, G. Obozinski, and N. Usunier, “Tensor decompositions for temporal knowledge base completion,” in 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020. OpenReview.net, 2020

    2020

  65. [76]

    Diachronic embedding for temporal knowledge graph completion,

    R. Goel, S. M. Kazemi, M. Brubaker, and P. Poupart, “Diachronic embedding for temporal knowledge graph completion,” Proceedings of the AAAI Conference on Artificial Intelligence , vol. 34, no. 04, pp. 3988–3995, Apr. 2020. [Online]. Available: https: //doi.org/10.1609/aaai.v34i04.5815

    2020 doi

  66. [77]

    Word translation without parallel data,

    G. Lample, A. Conneau, M. Ranzato, L. Denoyer, and H. J ´egou, “Word translation without parallel data,” in 6th International Conference on Learning Representations, ICLR 2018, Vancouver, BC, Canada, April 30 - May 3, 2018, Conference Track Proceedings. OpenReview.net, 2018. [...

    2018

  67. [78]

    Temporal knowledge question answering via abstract reasoning induction,

    Z. Chen, D. Li, X. Zhao, B. Hu, and M. Zhang, “Temporal knowledge question answering via abstract reasoning induction,” in Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2024, Bangkok, Thailand, August 11-16...

    2024 doi

  68. [79]

    Timemixer: Decomposable multiscale mixing for time series forecasting,

    S. Wang, H. Wu, X. Shi, T. Hu, H. Luo, L. Ma, J. Y . Zhang, and J. Zhou, “Timemixer: Decomposable multiscale mixing for time series forecasting,” in The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024 . OpenReview.net, 2...

    2024

  69. [80]

    Ada-mshyper: Adaptive multi- scale hypergraph transformer for time series forecasting,

    Z. Shang, L. Chen, B. Wu, and D. Cui, “Ada-mshyper: Adaptive multi- scale hypergraph transformer for time series forecasting,” in Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC,...

    2024

  70. [81]

    Self-alignment for factuality: Mitigating hallucinations in llms via self-evaluation,

    X. Zhang, B. Peng, Y . Tian, J. Zhou, L. Jin, L. Song, H. Mi, and H. Meng, “Self-alignment for factuality: Mitigating hallucinations in llms via self-evaluation,” in Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)...

    2024 doi

  71. [82]

    Unifying large language models and knowledge graphs: A roadmap,

    S. Pan, L. Luo, Y . Wang, C. Chen, J. Wang, and X. Wu, “Unifying large language models and knowledge graphs: A roadmap,” IEEE Trans. Knowl. Data Eng. , vol. 36, no. 7, pp. 3580–3599, 2024. [Online]. Available: https://doi.org/10.1109/TKDE.2024.3352100

    2024

  72. [83]

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

    L. Yang, H. Chen, Z. Li, X. Ding, and X. Wu, “Give us the facts: Enhancing large language models with knowledge graphs for fact-aware language modeling,” IEEE Trans. Knowl. Data Eng., vol. 36, no. 7, pp. 3091–3110, 2024. [Online]. Available: https://doi.org/10.1109/TKDE.2024.3...

    2024

  73. [84]

    VTQA: visual text question answering via entity alignment and cross-media reasoning,

    K. Chen and X. Wu, “VTQA: visual text question answering via entity alignment and cross-media reasoning,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024 . IEEE, 2024, pp. 27 208–27 217. [Online]. Available: http...

    2024

  74. [119]

    3929–3938

    PMLR, 2020, pp. 3929–3938

    2020

  75. [2023]

    2591–2599

    ACM, 2023, pp. 2591–2599

    2023

  76. [2024]

    Available: http://papers.nips.cc/paper files/paper/2024/ hash/3a6935d11910d6f9142b0a1e36fc6753-Abstract-Conference.html

    [Online]. Available: http://papers.nips.cc/paper files/paper/2024/ hash/3a6935d11910d6f9142b0a1e36fc6753-Abstract-Conference.html

    2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.