Pith. sign in

REVIEW 4 major objections 5 minor 56 references

TrumorGPT: Graph-Based Retrieval-Augmented Large Language Model for Fact-Checking

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read TrumorGPT claims that adding graph-based retrieval over semantic health knowledge graphs to GPT-4 fact-checks health statements with 88.5% binary accuracy, beating six general-purpose LLMs on PolitiFact health-care and coronavirus claims.

desk verdict The topic-specific TextRank proof is solid, but the GraphRAG accuracy claim is untested and the retrieval rule is internally contradictory. read the letter →

arxiv 2505.07891 v2 pith:DI3OZ6VD submitted 2025-05-11 cs.CL cs.AI

classification cs.CLcs.AI
keywords fact-checkingGraphRAGsemantichealthknowledgetopic-specificTextRanktopic-enhancedsentencecentralitylargelanguagemodelsmisinformationPolitiFact
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

TrumorGPT claims that a fact-checker can be made more accurate by giving GPT-4 access to a structured, up-to-date health knowledge graph and retrieving evidence from that graph before answering. On 600 PolitiFact 'Health Care' and 'Coronavirus' statements, it reports 88.5% binary accuracy, above GPT-4's 83.3% and five other LLMs. The paper's contribution is a graph-construction pipeline, topic-enhanced sentence centrality plus topic-specific TextRank with few-shot learning, feeding a GraphRAG layer that is meant to let the model verify claims against recent health facts instead of its static training data. A sympathetic reader would care because health misinformation is high-stakes and the reported gain is concrete; the authors also report that the graph layer keeps verdicts concise, averaging 2.8 sentences per response.

What carries the argument

The load-bearing object is the semantic health knowledge graph, a directed graph $G = \{E,R,F\}$ whose vertices are entities, edges are relations, and facts are triples $(h,r,t)$. To build these graphs, the paper uses topic-enhanced sentence centrality (BERT embeddings weighted with an LDA topic vector, $\eta = 0.7$) to pick key sentences and topic-specific TextRank, a PageRank variant with a topic-relevance teleportation vector and a health boost factor $\alpha = 1.5$, to rank the sentences; GPT-4 with few-shot examples then constructs the graph. GraphRAG retrieves stored graphs by comparing a query graph to each candidate with a weighted Jaccard score over consecutive triples, with subgraph isomorphism as the match criterion, and feeds the retrieved evidence into GPT-4's semantic reasoning, which outputs True, False, or Undetermined. The Markov-chain convergence theorems guarantee the ranking iteration converges, but the fact-checking verdict itself is carried by the match between the query graph and the knowledge-base graph.

What would settle it

Run the same 600 PolitiFact claims through TrumorGPT with the graph-retrieval step disabled or with randomly selected health graphs; if accuracy stays near 88.5%, the graphs are not load-bearing. The paper reports no such ablation, no value for the threshold $\theta$ used to declare a match, and no account of how DBpedia triples enter the prompt, so the mechanism is untested as stated.

Watch

Extended reading notes

Core claim

The central claim is that TrumorGPT, a GPT-4-based framework augmented by graph-based retrieval-augmented generation over semantic health knowledge graphs, can separate true from false health-related statements with 88.5% accuracy on 600 PolitiFact claims (300 true, 300 false), beating GPT-3.5, GPT-4, LLaMA 3.2, PaLM 2, Claude 3.5 Sonnet, and Gemini 1.5 on accuracy, precision, recall, and F1. The knowledge graphs are built from DBpedia health triples, using few-shot GPT-4 with the proposed topic-enhanced sentence centrality and topic-specific TextRank; retrieval matches the query graph to stored graphs through Jaccard/subgraph-isomorphism scoring, and the retrieved evidence grounds the final true/false decision. The paper also reports that TrumorGPT is concise (2.8 sentences on average) and that its six-way PolitiFact classification is only 49.3% accurate, so its strength is specifically the binary true/false decision.

Load-bearing premise

The claimed accuracy gain comes from the retrieved semantic health graphs rather than from GPT-4's own parametric knowledge or the prompt wording.

Editorial extensions

If this is right

  • A periodically updated health knowledge base lets a fact-checker answer claims that postdate the base LLM's December 2023 training cutoff.
  • Graph-grounded verdicts are concise: TrumorGPT averages 2.8 sentences per response, shorter than every compared LLM.
  • The binary true/false formulation is where the system works; on PolitiFact's full six-category scale accuracy drops to 49.3%, so the framework is a two-way verifier rather than a fine-grained rater.
  • Retrieval from a curated health-only graph filters off-topic information, which the authors credit for precision (91.4%) exceeding recall (85.0%).

Reading between the lines

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

  • Because the paper reports no ablation disabling GraphRAG, the cleanest test of its causal story is TrumorGPT with and without graph retrieval; until that run exists, part of the 5.2-point gap over GPT-4 could come from prompt design or few-shot example choice rather than from the graphs.
  • If the graph layer is genuinely causal, its advantage should be largest on claims about events after GPT-4's December 2023 cutoff; the paper's use of post-2023 DBpedia triples makes that an empirical check.
  • The 49.3% six-category accuracy suggests the knowledge graphs encode enough evidence for a coarse true/false split but not for graded distinctions such as 'Half True' versus 'Mostly False'; a fact-checker aimed at nuanced verdicts would need a finer relation or evidence representation.
  • The reported 2.8-sentence average suggests the graph retrieval acts as an evidence filter; an extension could measure whether conciseness and accuracy are separately attributable to the graphs versus to the few-shot instruction style.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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. The paper proposes TrumorGPT, a fact-checking framework in which GPT-4 constructs semantic health knowledge graphs from input documents using topic-enhanced sentence centrality and topic-specific TextRank, and then uses graph-based retrieval-augmented generation (GraphRAG) over a DBpedia-derived graph database to verify health-related claims. The authors report 88.5% binary accuracy on 600 PolitiFact "Health Care" and "Coronavirus" statements, compared with 72.7–83.3% for six LLM baselines, and they also report 49.3% accuracy on the original six-category PolitiFact scale. The manuscript includes formal convergence statements for the proposed topic-specific TextRank and a discussion of related RAG methods.

Significance. If the central claim were established, a 5.2-point improvement over GPT-4 with shorter responses could be practically valuable for automated fact-checking, and the convergence analysis of the modified TextRank would be a useful supporting contribution. The paper also addresses a timely problem and compares several contemporary LLMs. However, the current evidence does not support the central claim: the evaluation cannot separate the effect of graph retrieval from the underlying GPT-4 model, key components of the retrieval rule are unspecified, and the graph construction pipeline can encode the answer into the very graph that is later used for reasoning. At best, this is a preliminary proof-of-concept, not a validated contribution as stated.

major comments (4)
  1. [Section IV-C, Table I] The central claim that GraphRAG over semantic health knowledge graphs causes the accuracy gain is not supported by the experiments. TrumorGPT is compared with raw LLM baselines, but there is no ablation that removes the graph retrieval component, no prompt-matched baseline with the same few-shot instructions, and no report of error bars or significance tests. The 5.2-point gap over GPT-4 could be due to the added instructions, the extra DBpedia data, or the binary label construction rather than to the graph retrieval mechanism. An ablation such as TrumorGPT without retrieval, or GPT-4 with the same external triples in plain text, is essential before Table I can be interpreted.
  2. [Section III-E, Eq. (5)] The retrieval rule is underspecified and the stated condition is internally questionable. The threshold theta in the rule "if there exists a knowledge graph Gi such that S(Gx,Gi) <= theta, then output True" is never reported, and if S is the Jaccard similarity defined in Eq. (5), a low similarity score indicates low overlap, not evidence of truth. The paper also does not specify the graph embeddings, the subgraph matcher, or the triple weighting function f(t) used in practice, nor does it show a single retrieved graph for the 600 test claims. In addition, the function definition in Section III-E allows the output "Undetermined", but the binary evaluation in Section IV-C gives no mapping from Undetermined cases to True/False.
  3. [Section III-D and Table II] The evidence source and the reasoner are the same model, which creates a circularity concern. GPT-4 with Advanced Data Analysis constructs the knowledge graphs from the query text, and GPT-4 then produces the verdict by reasoning over those graphs. Table II illustrates the problem: the first example's semantic health knowledge graph contains the conclusion "More Deaths in 2021 than 2020" as a node connected by "implies", meaning the graph can encode the answer during construction. Without a counterfactual test (for example, retrieving a fixed external graph or a randomly corrupted graph), the reported 88.5% accuracy may reflect GPT-4's parametric memory or post-hoc graph construction rather than the contribution of graph-based retrieval.
  4. [Section IV-A and Section IV-D] The experimental setup is not described at a level that permits replication or causal attribution. The manuscript does not state how many DBpedia triples remain after filtering, how those triples are assembled into knowledge graphs, how the triples enter the GPT-4 prompt, or what retrieval database size N is used. The multi-class result further weakens the binary claim: the six-category accuracy drops to 49.3%, and the authors acknowledge that the model tends to select middle categories such as "Half True" or "Mostly False". Because the binary labels are obtained by collapsing three coarse categories into True and three into False, the high binary accuracy could be an artifact of coarse label construction rather than of graph content. Per-category cell counts for the six-class confusion matrix are not provided.
minor comments (5)
  1. [Section III-C] The text says "d is the damping vector" but d is a scalar damping factor; this should be corrected. The chosen values eta = 0.7 and alpha = 1.5 are given without sensitivity analysis or justification beyond a brief sentence.
  2. [Section III-C, Theorem 2] The proof of Theorem 2 invokes eigenvalue perturbation theory [49] but does not verify the conditions under which the rank-one update leaves all eigenvalues except the Perron eigenvalue unchanged. The displayed convergence bound also uses a strict inequality without stating the required assumptions on the initial distribution and the spectrum of P.
  3. [Figure 8 and Section IV-D] The six-category confusion matrix is shown only as a normalized heatmap; reporting numeric cell counts would allow readers to see how errors are distributed and whether the "middle category" tendency dominates the binary accuracy.
  4. [Throughout] No code, data, prompt templates, or API access dates are provided, which makes the proprietary-model comparisons difficult to reproduce. At minimum, the exact prompts used for the baselines and for TrumorGPT should be included in an appendix.
  5. [References] Reference [22] is cited for TextRank but has the title "An introduction to information retrieval", which appears mismatched; please check the citation. The reference to [50] should identify the specific theorem used for the convergence bound.

Circularity Check

1 steps flagged · score 6.0 of 10

Fact-checking verdict is defined as presence in a knowledge graph that the same GPT-4 model constructs, so the claimed GraphRAG evidence is self-generated rather than independently retrieved.

  1. self definitional [Section III-A; implementation in Section III-D; illustrative output in Table II (Section IV-D)]
    "The truthfulness of a new statement is evaluated based on its presence as an edge in this graph or the existence of a short path in the graph linking its subject to its object. Otherwise, the absence of such edges or short paths usually indicates that the statement is untrue."

    Section III-A defines the fact-checking verdict as membership in a knowledge graph G. But G is not an external benchmark: Section III-D says GPT-4 is instructed to construct the graphs ('we instruct GPT-4 to use Python to implement the proposed algorithms for constructing knowledge graphs'), and Section IV-D says 'we instruct TrumorGPT to construct a semantic health knowledge graph and perform fact-checking.' The same model therefore generates the evidence and then reads it off. Table II makes the reduction concrete: for the COVID-19 deaths query, the graph already contains the conclusion node "More Deaths in 2021 than 2020" with supporting counts, and the final semantic reasoning simply restates that graph.

full rationale

The paper's central claimed contribution is that graph-based retrieval over semantic health knowledge graphs improves fact-checking accuracy over the base LLM (Table I, Section IV-C). However, the knowledge graphs are themselves built by GPT-4 (Section III-D), and the fact-checking verdict is defined as matching a triple or short path in that graph (Section III-A). Because the same model constructs the evidence and then reasons over it, the accuracy gain is not causally attributable to the retrieval component. The paper provides no ablation without GraphRAG, no operational value for $ heta$ in Eq. (5), and no details on how DBpedia triples enter the prompt, so the retrieved graph could simply be a paraphrase of GPT-4's parametric knowledge. Table II illustrates the reduction: the graph for the COVID-19 death query already contains the final conclusion node, so the reasoning step is redundant. This is a partial circularity: the claim of a GraphRAG advantage is not supported by an independent test, and the verdict by construction depends on a graph that the tested model itself generated. The multi-class accuracy of 49.3% further suggests the binary numbers are an artifact of collapsing categories rather than of graph content. No self-citation is load-bearing here; the circularity is in the experimental design.

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

The framework relies on several hand-set parameters (eta, alpha, theta, f(t), LDA topics) that are not justified by sensitivity analysis. The central causal assumption, that GPT-4-constructed graphs from DBpedia triples drive the fact-checking improvement, is untested. The mathematical theorems, while standard, are peripheral to the empirical claim.

free parameters (6)
  • eta (topic/semantic weight) = 0.7
    Chosen by hand to balance BERT embeddings and LDA topic vectors in sentence embeddings (Section III-B1).
  • alpha (health topic boost) = 1.5
    Chosen by hand to give health-related LDA topics 50% more influence in topic relevance scores (Section III-C).
  • damping factor d = 0.85
    Standard PageRank default; used in TST power iteration and convergence analysis (Section III-C).
  • similarity threshold theta = not specified
    The GraphRAG decision rule declares True if S(Gx,Gi) <= theta, but theta is never given (Section III-E).
  • weighting function f(t) = not specified; presumably f(t)=1
    Eq. (5) defines graph similarity with a weighting function f(t); the paper never specifies it, and with f(t)=1 the score reduces to standard Jaccard similarity (Section III-E).
  • number of LDA topics K and health topic set = not specified
    The LDA topic model requires K and the set of health topics {k1,...,km}; neither is reported (Section III-B and III-C).
assumptions (5)
  • domain assumption A statement is true if it appears as a triple or via a short path in the knowledge graph; otherwise it is usually false.
    This is the premise of the graph-based fact-checking approach stated in Section III-A.
  • domain assumption GPT-4 with few-shot examples can construct accurate semantic health knowledge graphs from articles and queries.
    The entire pipeline relies on GPT-4 correctly implementing TST and sentence centrality via Python ADA and producing valid graphs (Section III-D).
  • domain assumption The health-filtered DBpedia triples contain up-to-date facts sufficient to judge the 600 PolitiFact claims.
    The knowledge base is assumed to cover the entities and relations needed for the test claims; no coverage analysis is provided (Section IV-A).
  • standard math The eigenvalues of dP + (1-d)E are 1, d*lambda2(P), ..., d*lambdan(P).
    Used to prove Theorem 2; cited from [49], but the conditions under which this rank-one update eigenvalue relation holds are not discussed (Section III-C, Theorem 2).
  • domain assumption PolitiFact's six-category ratings can be collapsed to True/False without introducing bias.
    The evaluation uses the collapsed labels as ground truth (Section IV-B).

how reviews work

0 comments
Cite this review

Pith. "Pith review of TrumorGPT: Graph-Based Retrieval-Augmented Large Language Model for Fact-Checking." pith.science (2026). https://pith.science/paper/DI3OZ6VD

@misc{pith2026250507891,
  author       = {Pith},
  title        = {Pith review of: TrumorGPT: Graph-Based Retrieval-Augmented Large Language Model for Fact-Checking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DI3OZ6VD}},
  note         = {Machine review of arXiv:2505.07891}
}
read the original abstract

In the age of social media, the rapid spread of misinformation and rumors has led to the emergence of infodemics, where false information poses a significant threat to society. To combat this issue, we introduce TrumorGPT, a novel generative artificial intelligence solution designed for fact-checking in the health domain. TrumorGPT aims to distinguish "trumors", which are health-related rumors that turn out to be true, providing a crucial tool in differentiating between mere speculation and verified facts. This framework leverages a large language model (LLM) with few-shot learning for semantic health knowledge graph construction and semantic reasoning. TrumorGPT incorporates graph-based retrieval-augmented generation (GraphRAG) to address the hallucination issue common in LLMs and the limitations of static training data. GraphRAG involves accessing and utilizing information from regularly updated semantic health knowledge graphs that consist of the latest medical news and health information, ensuring that fact-checking by TrumorGPT is based on the most recent data. Evaluating with extensive healthcare datasets, TrumorGPT demonstrates superior performance in fact-checking for public health claims. Its ability to effectively conduct fact-checking across various platforms marks a critical step forward in the fight against health-related misinformation, enhancing trust and accuracy in the digital information age.

Figures

Figures reproduced from arXiv: 2505.07891 by the authors.

Figure 1
Figure 1. The architecture of TrumorGPT, showcasing the workflow from user input to fact verification. TrumorGPT leverages advanced algorithms and an [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. An illustrative semantic health knowledge graph, highlighting the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the impact of the damping factor and [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: A sequence diagram illustrating the workflow of GraphRAG. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Comparison of average number of sentences generated by LLMs for [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Correlation between prediction accuracy and explanation length in [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 9
Figure 9. Figure 9: Performance of TrumorGPT in constructing semantic health knowl [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 8
Figure 8. Figure 8: A normalized confusion matrix for visualizing the performance of [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 48 canonical work pages

  1. [1]

    A model of online misin- formation,

    D. Acemoglu, A. Ozdaglar, and J. Siderius, “A model of online misin- formation,” Review of Economic Studies , p. rdad111, 2023

  2. [2]

    MEGA: Machine learning-enhanced graph analytics for infodemic risk manage- ment,

    C. N. Hang, P.-D. Yu, S. Chen, C. W. Tan, and G. Chen, “MEGA: Machine learning-enhanced graph analytics for infodemic risk manage- ment,” IEEE Journal of Biomedical and Health Informatics , vol. 27, no. 12, pp. 6100–6111, 2023

  3. [3]

    Misinformation: Strategic sharing, homophily, and endoge- nous echo chambers,

    J. Siderius, “Misinformation: Strategic sharing, homophily, and endoge- nous echo chambers,” Technical Report, National Bureau of Economic Research, Tech. Rep., 2021

  4. [4]

    Network security and contagion,

    D. Acemoglu, A. Malekian, and A. Ozdaglar, “Network security and contagion,” Journal of Economic Theory , vol. 166, pp. 536–585, 2016

  5. [5]

    Opinion dynamics and learning in social networks,

    D. Acemoglu and A. Ozdaglar, “Opinion dynamics and learning in social networks,” Dynamic Games and Applications , vol. 1, pp. 3–49, 2011

  6. [6]

    The spreading of misinformation online,

    M. Del Vicario, A. Bessi, F. Zollo, F. Petroni, A. Scala, G. Caldarelli, H. E. Stanley, and W. Quattrociocchi, “The spreading of misinformation online,” Proceedings of the National Academy of Sciences , vol. 113, no. 3, pp. 554–559, 2016

  7. [7]

    The spread of true and false news online,

    S. V osoughi, D. Roy, and S. Aral, “The spread of true and false news online,” Science, vol. 359, no. 6380, pp. 1146–1151, 2018

  8. [8]

    Contagion source detection in epidemic and infodemic outbreaks: Mathematical analysis and network algorithms,

    C. W. Tan and P.-D. Yu, “Contagion source detection in epidemic and infodemic outbreaks: Mathematical analysis and network algorithms,” Found. Trends® Netw., vol. 13, no. 2-3, pp. 107–251, 2023

Show all 56 references
  1. [9]

    Spread of misinformation in social net- works: Analysis based on Weibo tweets,

    H. Luo, M. Cai, and Y . Cui, “Spread of misinformation in social net- works: Analysis based on Weibo tweets,” Security and Communication Networks, vol. 2021, no. 1, p. 7999760, 2021

  2. [10]

    The simple macroeconomics of AI,

    D. Acemoglu, “The simple macroeconomics of AI,” National Bureau of Economic Research, Tech. Rep., 2024

  3. [11]

    Fact-checking: A meta-analysis of what works and for whom,

    N. Walter, J. Cohen, R. L. Holbert, and Y . Morag, “Fact-checking: A meta-analysis of what works and for whom,” Political Communication, vol. 37, no. 3, pp. 350–375, 2020

  4. [12]

    Where is your evidence: Improving fact-checking by justification modeling,

    T. Alhindi, S. Petridis, and S. Muresan, “Where is your evidence: Improving fact-checking by justification modeling,” in Proceedings of the First Workshop on Fact Extraction and Verification (FEVER) , 2018, pp. 85–90

  5. [13]

    Making better use of the crowd: How crowdsourcing can advance machine learning research,

    J. W. Vaughan, “Making better use of the crowd: How crowdsourcing can advance machine learning research,” Journal of Machine Learning Research, vol. 18, no. 193, pp. 1–46, 2018

  6. [14]

    Cooperative AI: Machines must learn to find common ground,

    A. Dafoe, Y . Bachrach, G. Hadfield, E. Horvitz, K. Larson, and T. Grae- pel, “Cooperative AI: Machines must learn to find common ground,” 2021

  7. [15]

    LLaMA: Open and efficient foundation language models,

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozi `ere, N. Goyal, E. Hambro, F. Azhar, A. Rodriguez, A. Joulin, E. Grave, and G. Lample, “LLaMA: Open and efficient foundation language models,” arXiv:2302.13971, 2023

  8. [16]

    Training language models to follow instructions with human feedback,

    L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. Wainwright, P. Mishkin, C. Zhang, S. Agarwal, K. Slama, A. Ray et al. , “Training language models to follow instructions with human feedback,” in Advances in Neural Information Processing Systems , vol. 35, 2022, pp. 27 730– 27 744

  9. [17]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askellet al., “Language models are few-shot learners,” in Advances in Neural Information Processing Systems, vol. 33, 2020, pp. 1877–1901

  10. [18]

    PaLM 2 technical report,

    R. Anil, A. M. Dai, O. Firat, M. Johnson, D. Lepikhin, A. Passos, S. Shakeri, E. Taropa, P. Bailey, Z. Chen et al. , “PaLM 2 technical report,” arXiv:2305.10403, 2023

  11. [19]

    GPT-4 technical report,

    OpenAI, “GPT-4 technical report,” arXiv:2303.08774, 2023

  12. [20]

    Sentence centrality revisited for unsupervised summarization,

    H. Zheng and M. Lapata, “Sentence centrality revisited for unsupervised summarization,” arXiv:1906.03508, 2019

  13. [21]

    TextRank: Bringing order into text,

    R. Mihalcea and P. Tarau, “TextRank: Bringing order into text,” in Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing, 2004, pp. 404–411

  14. [22]

    An introduction to information retrieval,

    S. Ceri, A. Bozzon, M. Brambilla, E. Della Valle, P. Fraternali, S. Quar- teroni, S. Ceri, A. Bozzon, M. Brambilla, E. Della Valle et al. , “An introduction to information retrieval,” Web Information Retrieval , pp. 3–11, 2013

  15. [23]

    Biased TextRank: Unsu- pervised graph-based content extraction,

    A. Kazemi, V . P ´erez-Rosas, and R. Mihalcea, “Biased TextRank: Unsu- pervised graph-based content extraction,” arXiv:2011.01026, 2020

  16. [24]

    PositionRank: An unsupervised approach to keyphrase extraction from scholarly documents,

    C. Florescu and C. Caragea, “PositionRank: An unsupervised approach to keyphrase extraction from scholarly documents,” inProceedings of the 55th Annual Meeting of the Association for Computational Linguistics (volume 1: long papers) , 2017, pp. 1105–1115

  17. [25]

    Retrieval- augmented generation for knowledge-intensive NLP tasks,

    P. Lewis, E. Perez, A. Piktus, F. Petroni, V . Karpukhin, N. Goyal, H. K ¨uttler, M. Lewis, W.-t. Yih, T. Rockt ¨aschel et al. , “Retrieval- augmented generation for knowledge-intensive NLP tasks,” in Advances in Neural Information Processing Systems , vol. 33, 2020, pp. 9459– 9474

  18. [26]

    From local to global: A graph RAG approach to query- focused summarization,

    D. Edge, H. Trinh, N. Cheng, J. Bradley, A. Chao, A. Mody, S. Truitt, and J. Larson, “From local to global: A graph RAG approach to query- focused summarization,” arXiv:2404.16130, 2024

  19. [27]

    Med-BERT: Pretrained contextualized embeddings on large-scale structured electronic health records for disease prediction,

    L. Rasmy, Y . Xiang, Z. Xie, C. Tao, and D. Zhi, “Med-BERT: Pretrained contextualized embeddings on large-scale structured electronic health records for disease prediction,” NPJ Digital Medicine , vol. 4, no. 1, p. 86, 2021

  20. [28]

    Medical semantic similarity with a neural language model,

    L. De Vine, G. Zuccon, B. Koopman, L. Sitbon, and P. Bruza, “Medical semantic similarity with a neural language model,” in Proceedings of the 23rd ACM International Conference on Conference on Information and Knowledge Management , 2014, pp. 1819–1822. 15

  21. [29]

    One LLM is not enough: Harnessing the power of ensemble learning for medical question answering,

    H. Yang, M. Li, Y . Xiao, H. Zhou, R. Zhang, and Q. Fang, “One LLM is not enough: Harnessing the power of ensemble learning for medical question answering,” medRxiv 2023.12.21.23300380, 2023

  22. [30]

    Qilin-Med: Multi- stage knowledge injection advanced medical large language model,

    Q. Ye, J. Liu, D. Chong, P. Zhou, Y . Hua, and A. Liu, “Qilin-Med: Multi- stage knowledge injection advanced medical large language model,” arXiv:2310.09089, 2023

  23. [31]

    Large language model AI chatbots require approval as medical devices,

    S. Gilbert, H. Harvey, T. Melvin, E. V ollebregt, and P. Wicks, “Large language model AI chatbots require approval as medical devices,”Nature Medicine, vol. 29, no. 10, pp. 2396–2398, 2023

  24. [32]

    Evaluating large language models on medical evidence summarization,

    L. Tang, Z. Sun, B. Idnay, J. G. Nestor, A. Soroush, P. A. Elias, Z. Xu, Y . Ding, G. Durrett, J. F. Rousseau et al. , “Evaluating large language models on medical evidence summarization,” NPJ Digital Medicine , vol. 6, no. 1, p. 158, 2023

  25. [33]

    The impact of multimodal large language models on health care’s future,

    B. Mesk ´o, “The impact of multimodal large language models on health care’s future,” Journal of Medical Internet Research, vol. 25, p. e52865, 2023

  26. [34]

    How to fight an infodemic,

    J. Zarocostas, “How to fight an infodemic,” The lancet , vol. 395, no. 10225, p. 676, 2020

  27. [35]

    A framework of AI-based approaches to improving eHealth literacy and combating infodemic,

    T. Liu and X. Xiao, “A framework of AI-based approaches to improving eHealth literacy and combating infodemic,” Frontiers in Public Health , vol. 9, p. 755808, 2021

  28. [36]

    As- sessing the risks of ‘infodemics’ in response to COVID-19 epidemics,

    R. Gallotti, F. Valle, N. Castaldo, P. Sacco, and M. De Domenico, “As- sessing the risks of ‘infodemics’ in response to COVID-19 epidemics,” Nat. Hum. Behav. , vol. 4, pp. 1285–1293, 2020

  29. [37]

    A survey on automated fact-checking,

    Z. Guo, M. Schlichtkrull, and A. Vlachos, “A survey on automated fact-checking,” Transactions of the Association for Computational Lin- guistics, vol. 10, pp. 178–206, 2022

  30. [38]

    Fully automated fact checking using external sources,

    G. Karadzhov, P. Nakov, L. M `arquez, A. Barr ´on-Cede˜no, and I. Koychev, “Fully automated fact checking using external sources,” arXiv:1710.00341, 2017

  31. [39]

    Physiological indicators for user trust in machine learning with influence enhanced fact-checking,

    J. Zhou, H. Hu, Z. Li, K. Yu, and F. Chen, “Physiological indicators for user trust in machine learning with influence enhanced fact-checking,” in Machine Learning and Knowledge Extraction , 2019, pp. 94–113

  32. [40]

    Language-aware truth assessment of fact candidates,

    N. Nakashole and T. Mitchell, “Language-aware truth assessment of fact candidates,” in Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , 2014, pp. 1009–1019

  33. [41]

    Arabic fake news detection: A fact checking based deep learning approach,

    F. Harrag and M. K. Djahli, “Arabic fake news detection: A fact checking based deep learning approach,” Transactions on Asian and Low-Resource Language Information Processing , vol. 21, no. 4, pp. 1– 34, 2022

  34. [42]

    Computational fact checking from knowledge networks,

    G. L. Ciampaglia, P. Shiralkar, L. M. Rocha, J. Bollen, F. Menczer, and A. Flammini, “Computational fact checking from knowledge networks,” PLoS ONE, vol. 10, no. 6, p. e0128193, 2015

  35. [43]

    Tracy: Tracing facts over knowledge graphs and text,

    M. H. Gad-Elrab, D. Stepanova, J. Urbani, and G. Weikum, “Tracy: Tracing facts over knowledge graphs and text,” in The World Wide Web Conference, 2019, pp. 3516–3520

  36. [44]

    Fact checking in heterogeneous information networks,

    B. Shi and T. Weninger, “Fact checking in heterogeneous information networks,” in Proceedings of the 25th International Conference Com- panion on World Wide Web , 2016, pp. 101–102

  37. [45]

    FACE-KEG: Fact checking explained using knowledge graphs,

    N. Vedula and S. Parthasarathy, “FACE-KEG: Fact checking explained using knowledge graphs,” in Proceedings of the 14th ACM International Conference on Web Search and Data Mining , 2021, pp. 526–534

  38. [46]

    Knowledge graph informed fake news classification via heterogeneous representation ensembles,

    B. Koloski, T. S. Perdih, M. Robnik- ˇSikonja, S. Pollak, and B. ˇSkrlj, “Knowledge graph informed fake news classification via heterogeneous representation ensembles,” Neurocomputing, vol. 496, pp. 208–226, 2022

  39. [47]

    Compare to the knowledge: Graph neural fake news detection with external knowledge,

    L. Hu, T. Yang, L. Zhang, W. Zhong, D. Tang, C. Shi, N. Duan, and M. Zhou, “Compare to the knowledge: Graph neural fake news detection with external knowledge,” in Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th Internationa...

  40. [48]

    DEAP-FAKED: Knowledge graph based approach for fake news detection,

    M. Mayank, S. Sharma, and R. Sharma, “DEAP-FAKED: Knowledge graph based approach for fake news detection,” in 2022 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), 2022, pp. 47–51

  41. [49]

    Eigenvalues of rank-one updated matrices with some applications,

    J. Ding and A. Zhou, “Eigenvalues of rank-one updated matrices with some applications,” Applied Mathematics Letters , vol. 20, no. 12, pp. 1223–1226, 2007

  42. [50]

    D. A. Levin, Y . Peres, and E. L. Wilmer, Markov Chains and Mixing Times. Providence, Rhode Island: American Mathematical Society, 2017

  43. [51]

    Can we soft prompt LLMs for graph learning tasks?

    Z. Liu, X. He, Y . Tian, and N. V . Chawla, “Can we soft prompt LLMs for graph learning tasks?” in Companion Proceedings of the ACM on Web Conference 2024, 2024, pp. 481–484

  44. [52]

    Exploring the potential of large language models (LLMs) in learning on graphs,

    Z. Chen, H. Mao, H. Li, W. Jin, H. Wen, X. Wei, S. Wang, D. Yin, W. Fan, H. Liu et al., “Exploring the potential of large language models (LLMs) in learning on graphs,” ACM SIGKDD Explorations Newsletter, vol. 25, no. 2, pp. 42–61, 2024

  45. [53]

    DBpedia: A nucleus for a web of open data,

    S. Auer, C. Bizer, G. Kobilarov, J. Lehmann, R. Cyganiak, and Z. Ives, “DBpedia: A nucleus for a web of open data,” in International Semantic Web Conference, 2007, pp. 722–735

  46. [54]

    Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context,

    M. Reid, N. Savinov, D. Teplyashin, D. Lepikhin, T. Lillicrap, J.-b. Alayrac, R. Soricut, A. Lazaridou, O. Firat, J. Schrittwieser et al. , “Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context,” arXiv:2403.05530, 2024

  47. [55]

    Hy- bridrag: Integrating knowledge graphs and vector retrieval augmented generation for efficient information extraction,

    B. Sarmah, D. Mehta, B. Hall, R. Rao, S. Patel, and S. Pasquali, “Hy- bridrag: Integrating knowledge graphs and vector retrieval augmented generation for efficient information extraction,” in Proceedings of the 5th ACM International Conference on AI in Finance , 2024, pp. 608– 616

  48. [56]

    Lightrag: Simple and fast retrieval-augmented generation,

    Z. Guo, L. Xia, Y . Yu, T. Ao, and C. Huang, “Lightrag: Simple and fast retrieval-augmented generation,” arXiv:2410.05779, 2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.