Pith. sign in

REVIEW 4 major objections 5 minor 85 references

Enhancing Health Information Retrieval with RAG by Prioritizing Topical Relevance and Factual Accuracy

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

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

Pith's one-line read A three-stage retrieval-augmented pipeline that generates a cited evidence summary and ranks documents by alignment with it improves health search on both topical relevance and factual accuracy.

desk verdict Plausible RAG-based health IR pipeline with a new GenText reference idea, but the reported gains over WISENLI need significance tests before they can be believed. read the letter →

arxiv 2502.04666 v1 pith:3CHWSJ5P submitted 2025-02-07 cs.IR

Rishabh Uapadhyay , Marco Viviani This is my paper · ORCID
classification cs.IR
keywords HealthInformationRetrievalConsumerSearchMisinformationRetrieval-AugmentedGenerationFactualAccuracyMultidimensionalRelevanceLargeLanguageModelsGenText
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 paper tries to establish that health-information search can be made safer by using retrieval-augmented generation: instead of scoring documents only by topical match, first retrieve evidence passages from a curated scientific literature database, have an LLM synthesize a short referenced answer text, then rank documents by how closely they align with that text in both stance and meaning. On two public health-search benchmark collections, the authors report that this three-stage pipeline beats existing retrieval baselines on combined relevance-and-credibility metrics, especially with a Llama-based configuration. The larger point is that a generated evidence summary can serve as a reusable yardstick for factual accuracy and as an explanation to users of why a document was judged credible.

What carries the argument

The load-bearing object is GenText, a short paragraph the LLM is prompted to write using only the top-k retrieved scientific passages, capped at 64 words, with a reference for every sentence. GenText plays two roles: it is the comparison standard for factual accuracy and a user-facing explanation. Factual accuracy is computed as $F(d,G)=\alpha\cdot\text{stance}(d,G)+(1-\alpha)\cdot\cos(d,G)$, where stance comes from a biomedical text-to-text model and cosine from biomedical language-model embeddings; the final ranking is $RSV(d,q,G)=\beta\cdot BM25(d,q)+(1-\beta)\cdot F(d,G)$.

What would settle it

Run the pipeline on queries whose top retrieved passages contain a known falsehood, then check whether a document repeating that falsehood earns a high factual-accuracy score; if GenText propagates the error, the ranking is not measuring truth.

Watch

Extended reading notes

Core claim

The paper's central claim is that a single LLM-generated paragraph, built only from topically selected scientific passages, can act as a proxy for the evidence base against which candidate documents are judged. Documents are ranked by a weighted blend of BM25 topicality and a factual-accuracy score that combines stance detection with semantic similarity, both comparing the document to the generated evidence summary. The authors show that this approach improves ranking quality relative to models that retrieve first and then verify, and relative to credibility-aware baselines, on two benchmark collections.

Load-bearing premise

The whole factual-accuracy score depends on GenText being a faithful, correct summary of the retrieved evidence; if the LLM adds, distorts, or omits a claim while generating that summary, the error flows into every document score.

Editorial extensions

If this is right

  • Health search systems can re-rank results without supervised misinformation labels, using only a curated literature API and an LLM prompt.
  • The same GenText can be shown alongside results, giving users a citation-linked rationale for why a page was judged factually unsupported.
  • The pipeline's factual-accuracy score is model-agnostic: any LLM that follows the context-only instruction can serve as the generator, so improvements in generator faithfulness should transfer directly to ranking quality.
  • Because factual accuracy is approximated by alignment with GenText, the method can absorb better stance models or similarity measures without changing the retrieval architecture.

Reading between the lines

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

  • I infer that the approach is only as comprehensive as the passage set: if the top-k retrieval misses the decisive evidence, GenText cannot repair the gap, so on queries where the knowledge base is thin the margin over baselines should shrink.
  • I infer that GenText could be reused as a query expansion for a second retrieval round, surfacing documents that match the evidence semantically but not the original query terms.
  • I infer that the method may transfer to other high-stakes domains with trusted document collections, such as legal or financial search, wherever a cited synthesis can serve as a verification anchor.
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. The paper proposes a three-stage RAG-driven health information retrieval model. In the first stage, top-k passages are retrieved from PubMed Central via BM25 and BioBERT-based similarity with a named-entity discount factor. In the second stage, an LLM generates a concise, citation-backed summary (GenText) from these passages. In the third stage, documents are scored for topicality by BM25 and for factual accuracy by combining stance detection (SCIFIVE) and cosine similarity against GenText, then linearly fused into a final ranking. The model is evaluated on the CLEF eHealth 2020 and TREC Health Misinformation 2020 datasets using CAM MAP and CAM NDCG, comparing three LLM-based configurations (GPT, Llama, Falcon) against BM25, DigiLab, CiTIUS, WISE, and WISENLI baselines. The authors report that the Llama RAG configuration achieves the best performance on both datasets, leading to the claim that the model 'outperforms existing baseline models' in retrieving both topically relevant and factually accurate health information.

Significance. If the empirical claims are robust, the paper makes a useful contribution by operationalizing factual accuracy through an LLM-generated evidence summary and integrating it into a ranking pipeline. The proposed three-stage architecture is a novel combination of existing components (BM25, BioBERT, NER discounting, stance detection, semantic similarity) and offers a potential explainability channel through generated citations. The use of two standard external benchmarks is a strength, as is the explicit description of the prompt and pipeline. However, the significance is currently limited by the lack of statistical validation of the reported improvements and by unresolved questions about the faithfulness of GenText as a truth proxy. The contribution is therefore incremental rather than transformative, and the paper would require strengthening of the empirical evidence to justify its claims.

major comments (4)
  1. [Section 4.4, Tables 1 and 2] The central claim that Llama RAG 'outperforms' all baselines is not supported by any uncertainty quantification. No confidence intervals, significance tests, or per-query variance are reported for any of the CAM MAP or CAM NDCG differences. With roughly 45 test queries per dataset (50 CLEF topics and 46 TREC topics minus the 5 tuning queries removed per dataset), the observed margins (e.g., CLEF top-5 CAM MAP: Llama 0.1079 vs GPT 0.1045 vs WISENLI 0.0883) may well be within noise. Please provide bootstrap confidence intervals or paired significance tests (e.g., paired bootstrap or randomization test over queries) and report the distribution of per-query differences. Without this, the headline 'outperforms' claim is not statistically established.
  2. [Section 4.4 and footnote 15] Hyperparameters k, alpha, beta, and dNE are optimized by grid search on only 5 queries per dataset, with those queries then removed from the test set. The paper acknowledges this procedure, but the selection process on a 5-query sample introduces substantial variability into the reported test scores. The reported point estimates are conditional on a tuning choice that could easily be affected by chance. Please provide a sensitivity analysis: e.g., evaluate the final ranking under different random 5-query tuning subsets, or report results across the grid to show that the improvement over baselines is not driven by a lucky hyperparameter selection. Also report the performance on the tuning set itself and the variance across random seeds.
  3. [Sections 3.2 and 3.3] The factual accuracy score F(d,G) is computed by comparing each document to GenText, an LLM-generated summary. The paper itself concedes in Section 5 that RAG does not guarantee the LLM confines itself to the provided context and that 'there remain risks if the system is not strictly limited to such reliable sources.' If GenText contains hallucinated or unfaithful content, the stance and similarity scores do not measure factual accuracy—they measure agreement with a potentially unreliable synthesis. This threatens the construct validity of the core 'factual accuracy' claim. Please provide an evaluation of GenText faithfulness, for example by comparing GenText statements against the source passages (automatically or via human annotation), and discuss how errors in GenText would propagate into the ranking. A qualitative inspection of a few examples would already help, but a systematic check is needed to support the claim that the model prioritizes factually accurate information.
  4. [Tables 1 and 2] The paper does not explain whether the baseline results (DigiLab, CiTIUS, WISE, WISENLI) were rerun with the same evaluation protocol, the same document collections, and the same implementations, or whether the numbers are taken from the original papers. If the latter, differences in preprocessing, query sets, and evaluation scripts could bias the comparison. Please clarify the provenance of each baseline score and, if possible, rerun all baselines on the same data with the same evaluation code. At minimum, state explicitly how each baseline was implemented or where its numbers came from, and whether the same document collection and relevance judgments were used.
minor comments (5)
  1. [Throughout] There are several typos and formatting issues: 'likelyhood' in Section 3.2, 'detais' in footnote 16, 'maily' in Section 4.2, and 'PCM' in Section 4.4 (should be PMC).
  2. [Section 3.1] The definition of the discount factor dNE would benefit from a specification of its allowed range and typical value; the grid-search range is not stated. Also, Figure 2 is referenced but not described in the text; please add a brief explanation of what the figure illustrates.
  3. [Section 4.2] The implementation section mentions 'training process' and 'training phase,' but the proposed pipeline is unsupervised and no training is performed. Please rephrase to avoid implying that any model weights are learned in this work.
  4. [Section 4.5] The explainability example in Figure 4 is suggestive but anecdotal. It would strengthen the paper to outline a concrete protocol for how GenText would be presented to users alongside search results and how its usefulness as an explanation could be evaluated.
  5. [Data availability] The data availability statement only mentions the external datasets. Since the paper introduces a specific prompting strategy and a pipeline, consider releasing the code and the exact prompts to support reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the RAG pipeline is scored against external benchmark labels, and the self-citations set design parameters rather than predetermine the outcome.

full rationale

The proposed pipeline is an unsupervised, three-stage scoring function: BM25 plus BioBERT passage selection from PMC, LLM summarization into GenText, and document scoring RSV = beta * BM25(d,q) + (1-beta) * [alpha * stance(d,GenText) + (1-alpha) * cos(d,GenText)]. Nothing in this chain is fitted to, or defined in terms of, the CLEF/TREC gold topicality and credibility labels used for evaluation. GenText is generated from the query and PMC passages, not from the documents being ranked or from the evaluation labels, so the factual-accuracy score is not the evaluation target by construction. The hyperparameters k, alpha, beta, and d_NE are tuned on five queries per dataset that are then removed from the test queries; this is an overfitting and small-sample risk, but it is not circularity, because the reported CAM MAP and CAM NDCG values are computed on unseen queries with held-out gold labels. The self-citations to references [42] (grid-search tuning and use of a single scientific article) and [78] (sentence granularity) guide design choices, but the claim that Llama RAG outperforms the baselines rests on the measured external benchmark numbers, not on those citations. There is no equation equating the prediction to an input, no parameter fitted to the predicted quantity and renamed as a result, and no uniqueness argument imported from the authors' prior work. Statistical-significance concerns about the small margins over WISENLI are a correctness and robustness risk, not a circularity concern.

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

The model introduces four tunable parameters (dNE, k, alpha, beta) and relies on a set of domain assumptions about the reliability of PMC, the validity of credibility labels, and the faithfulness of LLM-generated summaries. No invented entities are postulated, but GenText itself is a synthetic artifact whose accuracy is asserted rather than verified.

free parameters (5)
  • dNE = not reported (optimized via grid search)
    Discount factor applied to passage similarity when named entities for medicines/diseases do not match the query. Tuned on 5 randomly selected queries per dataset using F1, following the method in [42].
  • k = 10
    Number of top passages from PMC used to generate GenText. Searched from 5 to 20 on a 5-query subset per dataset.
  • alpha = 0.65
    Weight for stance detection versus cosine similarity in the factual accuracy score. Optimized on the tuning subset.
  • beta = 0.45
    Weight for topicality versus factual accuracy in the final RSV. Optimized on the tuning subset.
  • number_of_retrieved_articles = 1
    Set to 1 based on previous work [42] rather than tuned in this paper; still a modeling choice that affects GenText.
assumptions (5)
  • standard math BM25, cosine similarity, and linear score fusion are appropriate for topical relevance and factual accuracy approximation.
    Used without derivation in Section 3.
  • domain assumption Articles in PubMed Central are scientifically reliable and constitute valid evidence for factual accuracy.
    The system uses PMC exclusively as the knowledge base (Section 3.1).
  • domain assumption A document's factual accuracy can be approximated by its stance and semantic similarity to LLM-generated GenText derived from PMC passages.
    This is the core modeling assumption behind the F score (Section 3.3).
  • domain assumption The credibility labels in the CLEF eHealth and TREC Health Misinformation datasets are a valid proxy for factual accuracy.
    The evaluation uses credibility as the ground truth (Section 4.1).
  • domain assumption The retrieved top-k passages are sufficient and relevant for generating a factually accurate GenText.
    The pipeline relies on the retrieval stage (BM25 + BioBERT + NER discount) to supply accurate context (Section 3.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Health Information Retrieval with RAG by Prioritizing Topical Relevance and Factual Accuracy." pith.science (2026). https://pith.science/paper/3CHWSJ5P

@misc{pith2026250204666,
  author       = {Pith},
  title        = {Pith review of: Enhancing Health Information Retrieval with RAG by Prioritizing Topical Relevance and Factual Accuracy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3CHWSJ5P}},
  note         = {Machine review of arXiv:2502.04666}
}
read the original abstract

The exponential surge in online health information, coupled with its increasing use by non-experts, highlights the pressing need for advanced Health Information Retrieval models that consider not only topical relevance but also the factual accuracy of the retrieved information, given the potential risks associated with health misinformation. To this aim, this paper introduces a solution driven by Retrieval-Augmented Generation (RAG), which leverages the capabilities of generative Large Language Models (LLMs) to enhance the retrieval of health-related documents grounded in scientific evidence. In particular, we propose a three-stage model: in the first stage, the user's query is employed to retrieve topically relevant passages with associated references from a knowledge base constituted by scientific literature. In the second stage, these passages, alongside the initial query, are processed by LLMs to generate a contextually relevant rich text (GenText). In the last stage, the documents to be retrieved are evaluated and ranked both from the point of view of topical relevance and factual accuracy by means of their comparison with GenText, either through stance detection or semantic similarity. In addition to calculating factual accuracy, GenText can offer a layer of explainability for it, aiding users in understanding the reasoning behind the retrieval. Experimental evaluation of our model on benchmark datasets and against baseline models demonstrates its effectiveness in enhancing the retrieval of both topically relevant and factually accurate health information, thus presenting a significant step forward in the health misinformation mitigation problem.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

85 extracted references · 52 canonical work pages

  1. [1]

    Buchanan, J., Kock, N.: Information overload: A decision making perspective. In: Multiple Criteria Decision Making in the New Millennium: Proceedings of the Fif- teenth International Conference on Multiple Criteria Decision Making (MCDM) Ankara, Turkey, July 10–14, 2000, pp. 49–58 (2001). Springer

    2001

  2. [2]

    : CLEF eHealth Evaluation Lab 2021

    Goeuriot, L., Suominen, H., Kelly, L., Alemany, L.A., Brew-Sam, N., Cotik, V., Filippo, D., Gonzalez Saez, G., Luque, F., Mulhem, P., et al. : CLEF eHealth Evaluation Lab 2021. In: Advances in Information Retrieval: 43rd European Con- ference on IR Research, ECIR 2021, Virtual Event, March 28–April 1, 2021, Proceedings, Part II 43, pp. 593–600 (2021). Springer

    2021

  3. [3]

    In: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp

    Goodrich, B., Rao, V., Liu, P.J., Saleh, M.: Assessing the factual accuracy of gen- erated text. In: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 166–175 (2019)

    2019

  4. [4]

    International Journal of Environmental Research and Public Health 19(4), 2173 (2022)

    Di Sotto, S., Viviani, M.: Health misinformation detection in the social web: an overview and a data science approach. International Journal of Environmental Research and Public Health 19(4), 2173 (2022)

    2022

  5. [5]

    In: 2022 Ninth International Conference on Social Networks Analysis, Management and Security (SNAMS), pp

    Abdullah, M., Madain, A., Jararweh, Y.: Chatgpt: Fundamentals, applications and social impacts. In: 2022 Ninth International Conference on Social Networks Analysis, Management and Security (SNAMS), pp. 1–8 (2022). Ieee

    2022

  6. [6]

    arXiv preprint arXiv:2307.09288 (2023)

    Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bash- lykov, N., Batra, S., Bhargava, P., Bhosale, S., et al.: Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288 (2023)

    arXiv 2023

  7. [7]

    arXiv preprint 24https://kurami.disco.unimib.it/ 18 arXiv:2103.10360 (2021)

    Du, Z., Qian, Y., Liu, X., Ding, M., Qiu, J., Yang, Z., Tang, J.: Glm: General language model pretraining with autoregressive blank infilling. arXiv preprint 24https://kurami.disco.unimib.it/ 18 arXiv:2103.10360 (2021)

    arXiv 2021

  8. [8]

    In: International Conference on AI-generated Content, pp

    Ackerman, R., Balyan, R.: Automatic multilingual question generation for health data using llms. In: International Conference on AI-generated Content, pp. 1–11 (2023). Springer

    work page 2023

Show all 85 references
  1. [9]

    arXiv preprint arXiv:2403.01924 (2024)

    Frisoni, G., Cocchieri, A., Presepi, A., Moro, G., Meng, Z.: To generate or to retrieve? on the effectiveness of artificial contexts for medical open-domain question answering. arXiv preprint arXiv:2403.01924 (2024)

    2024 arXiv

  2. [10]

    arXiv preprint arXiv:2402.01700 (2024)

    Kell, G., Roberts, A., Umansky, S., Qian, L., Ferrari, D., Soboczenski, F., Wallace, B., Patel, N., Marshall, I.J.: Question answering systems for health professionals at the point of care–a systematic review. arXiv preprint arXiv:2402.01700 (2024)

    2024 arXiv

  3. [11]

    arXiv preprint arXiv:2302.04023 (2023)

    Bang, Y., Cahyawijaya, S., Lee, N., Dai, W., Su, D., Wilie, B., Lovenia, H., Ji, Z., Yu, T., Chung, W., et al.: A multitask, multilingual, multimodal eval- uation of chatgpt on reasoning, hallucination, and interactivity. arXiv preprint arXiv:2302.04023 (2023)

    2023 arXiv

  4. [12]

    arXiv preprint arXiv:2301.07597 (2023)

    Guo, B., Zhang, X., Wang, Z., Jiang, M., Nie, J., Ding, Y., Yue, J., Wu, Y.: How close is chatgpt to human experts? comparison corpus, evaluation, and detection. arXiv preprint arXiv:2301.07597 (2023)

    2023 arXiv

  5. [13]

    In: Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp

    Cao, M., Dong, Y., Wu, J., Cheung, J.C.K.: Factual error correction for abstrac- tive summarization models. In: Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 6251–6258. Association for Computational Linguistics, Online (2020)

    2020

  6. [14]

    In: Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp

    Raunak, V., Menezes, A., Junczys-Dowmunt, M.: The curious case of hallucina- tions in neural machine translation. In: Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp. 1172–1183. ...

    2021

  7. [15]

    ACM Comput

    Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y.J., Madotto, A., Fung, P.: Survey of hallucination in natural language generation. ACM Comput. Surv. 55(12) (2023)

    2023

  8. [16]

    arXiv preprint arXiv:2311.13878 (2023)

    Saxena, S., Prasad, S., Prakash, M., Shankar, A., Vaddina, V., Gopalakrishnan, S., et al.: Minimizing factual inconsistency and hallucination in large language models. arXiv preprint arXiv:2311.13878 (2023)

    2023 arXiv

  9. [17]

    arXiv preprint arXiv:2311.05232 (2023) 19

    Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., Chen, Q., Peng, W., Feng, X., Qin, B., et al.: A survey on hallucination in large language models: Prin- ciples, taxonomy, challenges, and open questions. arXiv preprint arXiv:2311.05232 (2023) 19

    2023 arXiv

  10. [18]

    arXiv preprint arXiv:2309.01219 (2023)

    Zhang, Y., Li, Y., Cui, L., Cai, D., Liu, L., Fu, T., Huang, X., Zhao, E., Zhang, Y., Chen, Y., et al.: Siren’s song in the ai ocean: a survey on hallucination in large language models. arXiv preprint arXiv:2309.01219 (2023)

    2023 arXiv

  11. [19]

    arXiv preprint arXiv:2301.00303 (2022)

    He, H., Zhang, H., Roth, D.: Rethinking with retrieval: Faithful large language model inference. arXiv preprint arXiv:2301.00303 (2022)

    2022 arXiv

  12. [20]

    arXiv preprint arXiv:2305.05862 (2023)

    Li, X., Zhu, X., Ma, Z., Liu, X., Shah, S.: Are chatgpt and gpt-4 general-purpose solvers for financial text analytics? an examination on several typical tasks. arXiv preprint arXiv:2305.05862 (2023)

    2023 arXiv

  13. [21]

    arXiv preprint arXiv:2304.08979 (2023)

    Shen, X., Chen, Z., Backes, M., Zhang, Y.: In chatgpt we trust? measuring and characterizing the reliability of chatgpt. arXiv preprint arXiv:2304.08979 (2023)

    2023 arXiv

  14. [22]

    Advances in Neural Information Processing Systems 33, 9459–9474 (2020)

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

    2020

  15. [23]

    In: 2024 47th MIPRO ICT and Electronics Convention (MIPRO), pp

    Perkovi´ c, G., Drobnjak, A., Botiˇ cki, I.: Hallucinations in llms: Understanding and addressing challenges. In: 2024 47th MIPRO ICT and Electronics Convention (MIPRO), pp. 2084–2088 (2024). IEEE

    2024

  16. [24]

    In: International Conference on Machine Learning, pp

    Borgeaud, S., Mensch, A., Hoffmann, J., Cai, T., Rutherford, E., Millican, K., Van Den Driessche, G.B., Lespiau, J.-B., Damoc, B., Clark, A., et al.: Improving lan- guage models by retrieving from trillions of tokens. In: International Conference on Machine Learning, pp. 2206–...

    2022

  17. [25]

    In: International Conference on Machine Learning, pp

    Guu, K., Lee, K., Tung, Z., Pasupat, P., Chang, M.: Retrieval augmented language model pre-training. In: International Conference on Machine Learning, pp. 3929– 3938 (2020). PMLR

    2020

  18. [26]

    Journal of Machine Learning Research24(251), 1–43 (2023)

    Izacard, G., Lewis, P., Lomeli, M., Hosseini, L., Petroni, F., Schick, T., Dwivedi- Yu, J., Joulin, A., Riedel, S., Grave, E.: Atlas: Few-shot learning with retrieval augmented language models. Journal of Machine Learning Research24(251), 1–43 (2023)

    2023

  19. [27]

    arXiv preprint arXiv:2404.07221 (2024)

    Setty, S., Jijo, K., Chung, E., Vidra, N.: Improving retrieval for rag based ques- tion answering models on financial documents. arXiv preprint arXiv:2404.07221 (2024)

    2024 arXiv

  20. [28]

    arXiv preprint arXiv:2404.16160 (2024)

    Kang, C., Novak, D., Urbanova, K., Cheng, Y., Hu, Y.: Domain-specific improve- ment on psychotherapy chatbot using assistant. arXiv preprint arXiv:2404.16160 (2024)

    2024 arXiv

  21. [29]

    20 arXiv preprint arXiv:2307.16877 (2023)

    Adlakha, V., BehnamGhader, P., Lu, X.H., Meade, N., Reddy, S.: Evaluating cor- rectness and faithfulness of instruction-following models for question answering. 20 arXiv preprint arXiv:2307.16877 (2023)

    2023 arXiv

  22. [30]

    In: TREC (2020)

    Clarke, C.L., Maistro, M., Smucker, M.D., Zuccon, G.: Overview of the trec 2020 health misinformation track. In: TREC (2020)

    2020

  23. [31]

    arXiv preprint arXiv:2303.01241 (2023)

    Zhao, R., Arana-Catania, M., Zhu, L., Kochkina, E., Gui, L., Zubiaga, A., Procter, R., Liakata, M., He, Y.: Panacea: An automated misinformation detection system on covid-19. arXiv preprint arXiv:2303.01241 (2023)

    2023 arXiv

  24. [32]

    arXiv preprint arXiv:2212.09683 (2022)

    Mendes, E., Chen, Y., Xu, W., Ritter, A.: Human-in-the-loop evaluation for early misinformation detection: A case study of covid-19 treatments. arXiv preprint arXiv:2212.09683 (2022)

    2022 arXiv

  25. [33]

    In: Proceed- ings of the 31st ACM International Conference on Information & Knowledge Management, pp

    Yue, Z., Zeng, H., Kou, Z., Shang, L., Wang, D.: Contrastive domain adaptation for early misinformation detection: A case study on covid-19. In: Proceed- ings of the 31st ACM International Conference on Information & Knowledge Management, pp. 2423–2433 (2022)

    2022

  26. [34]

    Information Processing & Management 59(5), 103029 (2022)

    Jiang, G., Liu, S., Zhao, Y., Sun, Y., Zhang, M.: Fake news detection via knowl- edgeable prompt learning. Information Processing & Management 59(5), 103029 (2022)

    2022

  27. [35]

    Chen, C., Shu, K.: Can llm-generated misinformation be detected? arXiv preprint arXiv:2309.13788 (2023)

    2023 arXiv

  28. [36]

    Multimedia Tools and Applications 82(4), 5271–5290 (2023)

    Upadhyay, R., Pasi, G., Viviani, M.: Vec4cred: a model for health misinformation detection in web pages. Multimedia Tools and Applications 82(4), 5271–5290 (2023)

    2023

  29. [37]

    In: Proceedings of the Conference on Information Technology for Social Good

    Upadhyay, R., Pasi, G., Viviani, M.: Health misinformation detection in web con- tent: A structural-, content-based, and context-aware approach based on web2vec. In: Proceedings of the Conference on Information Technology for Social Good. GoodIT ’21, pp. 19–24. Association for...

    2021

  30. [38]

    In: Proceedings of the 3rd International Workshop on Open Challenges in Online Social Networks

    Upadhyay, R., Pasi, G., Viviani, M.: Leveraging socio-contextual information in bert for fake health news detection in social media. In: Proceedings of the 3rd International Workshop on Open Challenges in Online Social Networks. OASIS ’23, pp. 38–46. Association for Computing ...

    2023

  31. [39]

    : E-bart: Jointly pre- dicting and explaining truthfulness

    Brand, E., Roitero, K., Soprano, M., Demartini, G., et al. : E-bart: Jointly pre- dicting and explaining truthfulness. In: Proceedings of the Conference for Truth and Trust Online (2021)

    2021

  32. [40]

    21 Proceedings of the ACM on Human-Computer Interaction 6(GROUP), 1–25 (2022)

    Kou, Z., Shang, L., Zhang, Y., Wang, D.: Hc-covid: A hierarchical crowdsource knowledge graph approach to explainable covid-19 misinformation detection. 21 Proceedings of the ACM on Human-Computer Interaction 6(GROUP), 1–25 (2022)

    2022

  33. [41]

    In: Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp

    Wu, J., Liu, Q., Xu, W., Wu, S.: Bias mitigation for evidence-aware fake news detection by causal intervention. In: Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 2308–2313 (2022)

    2022

  34. [42]

    Frontiers in Artificial Intelligence 6, 1184851 (2023)

    Upadhyay, R., Knoth, P., Pasi, G., Viviani, M.: Explainable online health infor- mation truthfulness in consumer health search. Frontiers in Artificial Intelligence 6, 1184851 (2023)

    2023

  35. [43]

    In: 2022 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), pp

    Shang, L., Zhang, Y., Yue, Z., Choi, Y., Zeng, H., Wang, D.: A knowledge-driven domain adaptive approach to early misinformation detection in an emergent health domain on social media. In: 2022 IEEE/ACM International Conference on Advances in Social Networks Analysis and Minin...

    2022

  36. [44]

    In: Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp

    Hassan, N., Arslan, F., Li, C., Tremayne, M.: Toward automated fact-checking: Detecting check-worthy factual claims by claimbuster. In: Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 1803–1812 (2017)

    2017

  37. [45]

    Transactions of the Association for Computational Linguistics10, 178–206 (2022)

    Guo, Z., Schlichtkrull, M., Vlachos, A.: A survey on automated fact-checking. Transactions of the Association for Computational Linguistics10, 178–206 (2022)

    2022

  38. [46]

    In: Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp

    Zeng, X., La Barbera, D., Roitero, K., Zubiaga, A., Mizzaro, S.: Combining large language models and crowdsourcing for hybrid human-ai misinformation detection. In: Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval,...

    2024

  39. [47]

    In: CEUR Workshop Proceedings, pp

    Goeuriot, L., Suominen, H., Pasi, G., Bassani, E., Brew-Sam, N., Gonz´ alez-S´ aez, G., Kelly, L., Mulhem, P., Seneviratne, S., Upadhyay, R.,et al.: Consumer health search at clef ehealth 2021. In: CEUR Workshop Proceedings, pp. 1–19 (2021). CEUR

    2021

  40. [48]

    IMS UniPD ad CLEF eHealth 2020 Task 2

    Di Nunzio, G.M., Marchesin, S., Vezzani, F.: A Study on Reciprocal Ranking Fusion in Consumer Health Search. IMS UniPD ad CLEF eHealth 2020 Task 2. In: CLEF (Working Notes) (2020)

    2020

  41. [49]

    In: Proceedings of the 32nd International ACM SIGIR Conference on Research and Development in Information Retrieval, pp

    Cormack, G.V., Clarke, C.L., Buettcher, S.: Reciprocal rank fusion outper- forms condorcet and individual rank learning methods. In: Proceedings of the 32nd International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 758–759 (2009)

    2009

  42. [50]

    22 In: CLEF 2020 (2020)

    Mulhem, P., Saez, G.G., Mannion, A., Schwab, D., Frej, J.: Lig-health at adhoc and spoken ir consumer health search: expanding queries using umls and fasttext. 22 In: CLEF 2020 (2020)

    2020

  43. [51]

    In: CLEF (Working Notes) (2020)

    Seneviratne, S., Daskalaki, E., Hossain, M.Z., Lenskiy, A.: Sandidoc at clef 2020- consumer health search: Adhoc ir task. In: CLEF (Working Notes) (2020)

    2020

  44. [52]

    In: TREC (2020)

    Fern´ andez-Pichel, M., Losada, D.E., Pichel, J.C., Elsweiler, D.: Citius at the trec 2020 health misinformation track. In: TREC (2020)

    2020

  45. [53]

    arXiv preprint arXiv:2202.06771 (2022)

    Zhang, B., Naderi, N., Jaume-Santero, F., Teodoro, D.: Ds4dh at trec health misinformation 2021: multi-dimensional ranking models with transfer learning and rank fusion. arXiv preprint arXiv:2202.06771 (2022)

    2022 arXiv

  46. [54]

    arXiv preprint arXiv:2112.06080 (2021)

    Schlicht, I.B., Paula, A.F.M., Rosso, P.: Upv at trec health misinformation track 2021 ranking with sbert and quality estimators. arXiv preprint arXiv:2112.06080 (2021)

    2021 arXiv

  47. [55]

    In: Proceedings of the Thirtieth REtrieval Conference Proceedings (TREC 2021)

    Abualsaud, M., Chen, I.X., Ghajar, K., Minh, L., Smucker, M., Tahami, A.V., Zhang, D.: Uwaterloomds at the trec 2021 health misinformation track. In: Proceedings of the Thirtieth REtrieval Conference Proceedings (TREC 2021). National Institute of Standards and Technology (NIST...

    2021

  48. [56]

    In: Proceedings of the Fifth Workshop on Technologies for Machine Translation of Low-Resource Languages (LoResMT 2022), pp

    Pankaj, S., Gautam, A.: Augmented bio-sbert: Improving performance for pair- wise sentence tasks in bio-medical domain. In: Proceedings of the Fifth Workshop on Technologies for Machine Translation of Low-Resource Languages (LoResMT 2022), pp. 43–47 (2022)

    2022

  49. [57]

    Journal of machine learning research 21(140), 1–67 (2020)

    Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., Liu, P.J.: Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of machine learning research 21(140), 1–67 (2020)

    2020

  50. [58]

    arXiv preprint arXiv:2402.10426 (2024)

    Wan, H., Feng, S., Tan, Z., Wang, H., Tsvetkov, Y., Luo, M.: Dell: Generating reactions and explanations for llm-based misinformation detection. arXiv preprint arXiv:2402.10426 (2024)

    2024 arXiv

  51. [59]

    In: Companion Proceedings of the ACM on Web Conference 2024, pp

    Choi, E.C., Ferrara, E.: Automated claim matching with large language mod- els: empowering fact-checkers in the fight against misinformation. In: Companion Proceedings of the ACM on Web Conference 2024, pp. 1441–1449 (2024)

    2024

  52. [60]

    Cao, Y., Nair, A.M., Eyimife, E., Soofi, N.J., Subbalakshmi, K., Wullert II, J.R., Basu, C., Shallcross, D.: Can large language models detect misinformation in scientific news reporting? arXiv preprint arXiv:2402.14268 (2024)

    2024

  53. [61]

    arXiv preprint arXiv:2402.17887 (2024) 23

    Wang, J., Yang, Z., Yao, Z., Yu, H.: Jmlr: Joint medical llm and retrieval training for enhancing reasoning and professional question answering capability. arXiv preprint arXiv:2402.17887 (2024) 23

    2024 arXiv

  54. [62]

    arXiv preprint arXiv:2405.14654 (2024)

    Khlaut, J., Dancette, C., Ferreres, E., Bennani, A., H´ erent, P., Manceron, P.: Effi- cient medical question answering with knowledge-augmented question generation. arXiv preprint arXiv:2405.14654 (2024)

    2024 arXiv

  55. [63]

    arXiv preprint arXiv:2301.12652 (2023)

    Shi, W., Min, S., Yasunaga, M., Seo, M., James, R., Lewis, M., Zettlemoyer, L., Yih, W.-t.: Replug: Retrieval-augmented black-box language models. arXiv preprint arXiv:2301.12652 (2023)

    2023 arXiv

  56. [64]

    Ren, R., Wang, Y., Qu, Y., Zhao, W.X., Liu, J., Tian, H., Wu, H., Wen, J.- R., Wang, H.: Investigating the Factual Knowledge Boundary of Large Language Models with Retrieval Augmentation (2023)

    2023

  57. [65]

    In: Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pp

    Izacard, G., Grave, E.: Leveraging passage retrieval with generative models for open domain question answering. In: Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pp. 874–880. Association for Computatio...

    2021

  58. [66]

    In: Proceedings of the 61st Annual Meeting of the Association for Computa- tional Linguistics (Volume 1: Long Papers), pp

    Trivedi, H., Balasubramanian, N., Khot, T., Sabharwal, A.: Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions. In: Proceedings of the 61st Annual Meeting of the Association for Computa- tional Linguistics (Volume 1: Long Papers)...

    2023

  59. [67]

    In: Findings of the Association for Computational Linguistics: ACL 2023, pp

    Li, D., Rawat, A.S., Zaheer, M., Wang, X., Lukasik, M., Veit, A., Yu, F., Kumar, S.: Large language models with controllable working memory. In: Findings of the Association for Computational Linguistics: ACL 2023, pp. 1774–1793. Association for Computational Linguistics, Toron...

    2023

  60. [68]

    Cai, D., Wang, Y., Bi, W., Tu, Z., Liu, X., Lam, W., Shi, S.: Skeleton-to-response: Dialogue generation guided by retrieval memory. In: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computa- tional Linguistics: Human Language Technolog...

    2019

  61. [69]

    Cai, D., Wang, Y., Bi, W., Tu, Z., Liu, X., Shi, S.: Retrieval-guided dialogue response generation via a matching-to-generation framework. In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on N...

    2019

  62. [70]

    arXiv preprint arXiv:2302.12813 (2023) 24

    Peng, B., Galley, M., He, P., Cheng, H., Xie, Y., Hu, Y., Huang, Q., Liden, L., Yu, Z., Chen, W., et al.: Check your facts and try again: Improving large language models with external knowledge and automated feedback. arXiv preprint arXiv:2302.12813 (2023) 24

    2023 arXiv

  63. [71]

    arXiv preprint arXiv:2306.16092 (2023)

    Cui, J., Li, Z., Yan, Y., Chen, B., Yuan, L.: Chatlaw: Open-source legal large language model with integrated external knowledge bases. arXiv preprint arXiv:2306.16092 (2023)

    2023 arXiv

  64. [72]

    In: The Eleventh International Conference on Learning Representations (2023)

    Zhou, S., Alon, U., Xu, F.F., Jiang, Z., Neubig, G.: Docprompting: Generat- ing code by retrieving the docs. In: The Eleventh International Conference on Learning Representations (2023)

    2023

  65. [73]

    arXiv preprint arXiv:2307.13528 (2023)

    Chern, I., Chern, S., Chen, S., Yuan, W., Feng, K., Zhou, C., He, J., Neubig, G., Liu, P., et al.: Factool: Factuality detection in generative ai–a tool aug- mented framework for multi-task and multi-domain scenarios. arXiv preprint arXiv:2307.13528 (2023)

    2023 arXiv

  66. [74]

    arXiv preprint arXiv:2305.12744 (2023)

    Pan, L., Wu, X., Lu, X., Luu, A.T., Wang, W.Y., Kan, M.-Y., Nakov, P.: Fact-checking complex claims with program-guided reasoning. arXiv preprint arXiv:2305.12744 (2023)

    2023 arXiv

  67. [75]

    arXiv preprint arXiv:2310.00305 (2023)

    Zhang, X., Gao, W.: Towards llm-based fact verification on news claims with a hierarchical step-by-step prompting method. arXiv preprint arXiv:2310.00305 (2023)

    2023 arXiv

  68. [76]

    Transactions of the Association for Computational Linguistics 12, 334–354 (2024)

    Zeng, F., Gao, W.: Justilm: Few-shot justification generation for explainable fact- checking of real-world claims. Transactions of the Association for Computational Linguistics 12, 334–354 (2024)

    2024

  69. [77]

    Foundations and Trends ® in Information Retrieval 3(4), 333–389 (2009)

    Robertson, S., Zaragoza, H., et al.: The probabilistic relevance framework: Bm25 and beyond. Foundations and Trends ® in Information Retrieval 3(4), 333–389 (2009)

    2009

  70. [78]

    In: Joint European Conference on Machine Learning and Knowledge Discovery in Databases, pp

    Upadhyay, R., Pasi, G., Viviani, M.: A passage retrieval transformer-based re- ranking model for truthful consumer health search. In: Joint European Conference on Machine Learning and Knowledge Discovery in Databases, pp. 355–371 (2023). Springer

    2023

  71. [79]

    Bioinformatics 36(4), 1234–1240 (2020)

    Lee, J., Yoon, W., Kim, S., Kim, D., Kim, S., So, C.H., Kang, J.: Biobert: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics 36(4), 1234–1240 (2020)

    2020

  72. [80]

    arXiv preprint arXiv:2007.01282 (2020)

    Izacard, G., Grave, E.: Leveraging passage retrieval with generative models for open domain question answering. arXiv preprint arXiv:2007.01282 (2020)

    2020 arXiv

  73. [81]

    arXiv preprint arXiv:2106.03598 (2021)

    Phan, L.N., Anibal, J.T., Tran, H., Chanana, S., Bahadroglu, E., Peltekian, A., Altan-Bonnet, G.: Scifive: a text-to-text transformer model for biomedical literature. arXiv preprint arXiv:2106.03598 (2021)

    2021 arXiv

  74. [82]

    arXiv preprint arXiv:1611.09268 (2016)

    Bajaj, P., Campos, D., Craswell, N., Deng, L., Gao, J., Liu, X., Majumder, R., McNamara, A., Mitra, B., Nguyen, T., et al.: MS MARCO: A human generated 25 machine reading comprehension dataset. arXiv preprint arXiv:1611.09268 (2016)

    2016 arXiv

  75. [83]

    In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp

    Schwarz, J., Morris, M.: Augmenting web pages and search results to support credibility assessment. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 1245–1254 (2011)

    2011

  76. [84]

    In: International Conference on Web Information Systems Engineering, pp

    Upadhyay, R., Pasi, G., Viviani, M.: An unsupervised approach to genuine health information retrieval based on scientific evidence. In: International Conference on Web Information Systems Engineering, pp. 119–135 (2022). Springer

    2022

  77. [85]

    In: Proceedings of the ACM SIGIR International Conference on Theory of Information Retrieval, pp

    Lioma, C., Simonsen, J.G., Larsen, B.: Evaluation measures for relevance and credibility in ranked lists. In: Proceedings of the ACM SIGIR International Conference on Theory of Information Retrieval, pp. 91–98 (2017) 26

    2017

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.