Pith. sign in

REVIEW 5 major objections 6 minor 39 references

TrustDataFilter:Leveraging Trusted Knowledge Base Data for More Effective Filtering of Unknown Information

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A framework that checks candidate statements against trusted knowledge reports consistent accuracy gains in domain data filtering.

desk verdict A modest but real engineering contribution whose gains are probably correct on its own benchmarks, but the label construction makes the stronger claim about true filtering accuracy unproven. read the letter →

arxiv 2502.15714 v1 pith:2U5LGFVB submitted 2025-01-25 cs.IR cs.AIcs.CL

Jinghong Zhang , Yidong Cui , Weiling Wang , Xianyou Cheng This is my paper · ORCID
classification cs.IRcs.AIcs.CL
keywords domain-specificknowledgebasedatafilteringnaturallanguageinferencelargemodelsconstructioncontradictiondetectioniterativedecisiontree
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

The paper is trying to establish that the reliability of a domain knowledge base can be improved by exploiting the fact that facts inside a domain are interconnected, so a new statement can be checked against statements already known to be true. It proposes the self-NLI-TDF framework, which combines a large language model's confidence judgment with a natural-language-inference check against a trusted knowledge base, and a decision tree that fuses the two signals. Across biology, science, and radiation datasets, the framework reports an average accuracy gain of about three percentage points over direct filtering, with larger gains where the base model is weakest. If true, this offers a cheap, model-agnostic way to improve domain-specific knowledge collection without retraining.

What carries the argument

The load-bearing object is the self-NLI-TDF framework, an iterative pipeline. Its core mechanism: a vector database of trusted statements is built, either from authoritative sources or from a small manually annotated seed; for each candidate statement the system retrieves the nearest trusted statement by cosine similarity; a natural-language-inference model trained on the multi-genre NLI task labels the relationship between the candidate and the retrieved trusted statement as entailment, contradiction, or neutral, with a probability-based confidence score; a large language model independently gives a reliability label and confidence score; and a decision tree trained on five percent of the data fuses the four numbers into a final accept or reject decision. Accepted statements are vectorized and added to the trusted base, so the matching improves over five iterations.

What would settle it

An external evaluation set with human-verified labels, in any domain, would settle it: run the same three models through direct filtering and through self-NLI-TDF filtering on identical seed knowledge, and compare accuracy. If the gain disappears or reverses, the earlier result measured agreement with model-generated labels, not improved truthfulness.

Watch

Extended reading notes

Core claim

The paper's central claim is that filtering accuracy for domain knowledge improves when each candidate statement is judged not only by a large language model's self-reported confidence but also by whether it can be inferred from, or contradicts, a similar statement that has already been trusted. On three constructed datasets the framework's accuracy averaged about three percentage points higher than direct filtering, with the largest gains in the radiation domain where the base model had the least prior knowledge. The paper also reports that replacing the trusted knowledge with random irrelevant statements did not produce the same gains, which it takes as evidence that the comparison against genuinely related trusted knowledge is what carries the improvement.

Load-bearing premise

The evaluation assumes that the labels marking statements as correct or incorrect are genuine ground truth, but in two of the three datasets those labels were produced by large language models rather than by independent expert verification.

Editorial extensions

If this is right

  • Domain knowledge bases can be cleaned iteratively: each accepted statement becomes part of the trusted base used to judge later statements, so filtering improves as the base grows.
  • In domains where a large language model has little prior knowledge, the framework's gain is largest, suggesting it can compensate for weak model familiarity with domain-specific trusted references.
  • Using irrelevant or random reference statements does not reproduce the gain, so the quality of the trusted-knowledge matching, not just the extra module, is what drives the improvement.
  • Because the decision tree fuses a reliability label, a contradiction label, and their confidence scores, the framework can keep novel non-contradictory statements even when the LLM is uncertain, while still rejecting contradictions.

Reading between the lines

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

  • An implicit extension is to use this as a continuous quality-control loop for live knowledge bases: the compounding effect of accepted statements on later judgments may grow beyond the five iterations the paper tested.
  • Because the biological and science labels were themselves produced by large language models, a natural follow-up is to re-measure the gain on human-validated labels; the reported three-point gain may partly reflect alignment with LLM-generated labels.
  • The framework's dependence on a seed set of trusted statements means any bias in that seed will propagate; a testable next step is varying the seed composition and measuring how filtering accuracy and topic clusters shift.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes Self-NLI-TDF, an iterative pipeline for filtering domain-specific knowledge-base entries. The pipeline first asks a large language model for a binary reliability judgment and a confidence score (Confidence Evaluation), retrieves the most similar trusted entry from a vector knowledge base, obtains an NLI relation (entailment/contradiction/neutral) and confidence via RoBERTa-MNLI (Contradiction Evaluation), and fuses the four signals with a decision tree (Decision Evaluation). Accepted items are added to the knowledge base over five iterations. The authors construct three labeled datasets in biology, science, and radiation and compare three methods (Basic prompting, Self-NLI, and a 'Fake' ablation with random out-of-domain premises) using RoBERTa, Qwen2-7B, and GPT-3.5. They report average accuracy, precision, and F1 gains of roughly 3, 2.6, and 1.7 percentage points, plus a BERTopic analysis intended to show more compact filtering.

Significance. If the empirical claims held, the contribution would be a useful modular plug-in for domain knowledge-base construction: it is simple, open-sourced, and the Fake-knowledge ablation is a sensible control. The construction of three domain testbeds is also a useful resource. However, the current evaluation does not establish the central claim because the labels used as ground truth are generated or assumed by the same kinds of models under test, and because all results are single runs without uncertainty quantification. The stress-test concern about label circularity lands: the measured gains may reflect self-consistency with the labeling pipeline rather than better filtering of true domain knowledge. The framework idea is plausible, but the evidence presented is not yet sufficient for the strength of the claims.

major comments (5)
  1. [Sec. IV.A-C, Table I] The central claim requires that the correct/incorrect labels in the three test sets are trustworthy ground truth, but this condition is not met. In the biological dataset, GPT-3.5 generates both the correct and the deliberately distorted incorrect statements under a self-instruct protocol, and GPT-3.5 is also one of the models used in the Confidence Evaluation; the labels are therefore the same model's prior, not an independent standard. In the science dataset, correct/incorrect labels come from ARC multiple-choice options, but the test statements are LLM extractions and paraphrases, with no verification that the extracted propositions are true or false outside the original question. In the radiation dataset, correctness is assumed because the crawled websites are authoritative, and negative items are machine-generated corruptions (Sec. IV.C). No human validation, inter-annotator agreement, or external gold standard is reported. Because every metric in Table I is computed against these labels, the reported gains may reflect agreement with the labeling process rather than improved filtering of true domain knowledge. This is the principal threat to the paper's conclusion.
  2. [Sec. V.B, Table I] All results are from single runs and no error bars, confidence intervals, or significance tests are reported. Several headline differences are below one percentage point (Biological-RoBERTa 0.8799 to 0.8831; Science-Qwen2 0.8953 to 0.8986; Science-GPT 0.9315 to 0.9353), so the claimed improvement is not distinguishable from run-to-run noise. The abstract and introduction claim of a 1%-5% accuracy improvement is also inconsistent with Table I, which includes gains of 0.32% and 0.38% in the biological and science domains and gains of 5.09-7.02% in the radiation domain. The '+3% average' claim needs a defined aggregation rule and a measure of dispersion before it can be evaluated.
  3. [Table I, Biological and Radiation rows] The introduction states the framework improves accuracy by reducing false negatives, but the reported results show recall decreasing in several conditions (Biological-Qwen2 0.9914 to 0.9797; Biological-GPT 0.9900 to 0.9358; Biological-RoBERTa 0.9045 to 0.8833; Radiation-RoBERTa 0.9999 to 0.9774). In these rows the accuracy gain comes from a large precision increase, i.e., a shift in operating point, not from a reduction of false negatives. Since filtering decisions have asymmetric costs, accuracy alone is insufficient; the paper should report precision-recall trade-offs, threshold-matched comparisons, or a cost-sensitive evaluation.
  4. [Sec. V.A Step 6, Table I] The Fake ablation uses random, out-of-domain ASSET sentences as matched knowledge. This shows that injecting unrelated text as an NLI premise degrades or fails to help filtering, but it does not isolate the effect of the 'trusted' property of the knowledge base. A matched but untrusted in-domain premise, or a randomly selected in-domain premise, would be needed to demonstrate that trustworthiness rather than topical relevance drives the improvement. The current control therefore supports a weaker conclusion than the title and contribution statements.
  5. [Sec. II-C, III-A, V.A] The framework is described as convenient and as requiring no extensive model training, yet the implemented pipeline needs a 5% annotated seed to train the decision tree, to fine-tune RoBERTa, and to initialize the trusted knowledge base. Meanwhile the GPT/Qwen baselines are used zero-shot. This makes the comparison uneven and the 'plug-and-play' claim overstated; the annotation budget should be treated as a variable and its effect reported in a sensitivity analysis.
minor comments (6)
  1. [Sec. II-C] The expression 'c2 ∈ [0,1}' contains a typo and should read '[0,1]'. Also, the mapping between y2 values and the terms 'entailment/contradiction/neutral' is given earlier but not repeated in the decision-tree feature description, which is easy to misread.
  2. [Sec. V.C.3, Table II] Fewer BERTopic clusters is not by itself evidence of better filtering; a filter that discards legitimate but diverse statements would also reduce the cluster count. The paper should motivate why topic compactness is a quality measure or support it with a downstream evaluation.
  3. [Sec. V.B] The statement 'on average, our method achieved (3%) improvement' does not specify whether the average is over models, datasets, or both; the paper should report the full per-cell results with variance and the aggregation rule.
  4. [References] The reference list contains a duplicate: [16] and [30] are the same Williams et al. paper. Several in-text citation markers (e.g., 'Figure.1' and 'Figure.2' in Sec. I) are also malformed.
  5. [Sec. IV.A] The claim that the biological dataset is 'reliable' and 'comprehensive' is not supported by any annotation-quality experiments; a brief analysis of label agreement or a sample audit would help.
  6. [Sec. I] The code and dataset links should be accompanied by versioned DOIs or archival references, since the current URLs may not be stable.

Circularity Check

1 steps flagged · score 6.0 of 10

Biological benchmark labels are GPT-3.5 self-generations, so the GPT-3.5 accuracy gain is partly a self-consistency measure rather than independent filtering quality.

  1. self definitional [Section IV.A (Biological Dataset) and Section V.A Step 3 (Basic Filtering), evaluated in Table I]
    "The Biological dataset was created using GPT-3.5 based on its prior knowledge. We implemented and modified the self-instruction framework to enable the large language model to generate both correct and incorrect biological knowledge points based on given topics and subtopics. ... Using prompt-based techniques, we directly instructed the three large language models (RoBERTa, Qwen2, and GPT-3.5) to determine whether the data was trustworthy, and the results were recorded for further analysis and comparison."

    The biological test labels are not externally verified; they are GPT-3.5's own self-instruct outputs. GPT-3.5 is also the model whose filtering output is scored against those labels. Thus the GPT-3.5 accuracy values in Table I measure agreement between one GPT-3.5 generation and another GPT-3.5 judgment, i.e., self-consistency, not filtering against independently established biological truth. Because the Self-NLI pipeline is initialized from the same GPT-3.5-generated training labels, part of the reported 1%-5% improvement is a self-agreement artifact rather than a validation of general filtering quality. The Science and Radiation datasets have external origins, so the circularity is partial rather than total.

full rationale

The framework's architecture is internally coherent: Confidence Evaluation, Contradiction Evaluation, and Decision Evaluation are described explicitly, and the Fake Filtering ablation provides a meaningful control for the value of trusted knowledge matching. There is no load-bearing self-citation chain; the cited works are background support, not the authors' own prior results. However, the Biological dataset is generated by GPT-3.5 under self-instruct prompts, and GPT-3.5 is one of the filters evaluated. The correctness labels for that dataset are therefore defined by the same model family whose filtering accuracy is being measured. This makes the biological GPT-3.5 results at least partly a measure of agreement with the labeling process, reducing one load-bearing part of the empirical claim to a self-consistency check by construction. The Science labels derive from the externally sourced ARC multiple-choice data, and the Radiation labels rest on an explicit authority assumption; these are weaker evidence than independent gold standards but are not circular in the same way. Overall, the paper has partial circularity in one of its three evaluation tracks, warranting a score of 6.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on one fitted threshold, a chosen iteration count, and a set of domain assumptions about knowledge interconnectedness, NLI reliability, and label validity. The benchmark labels are the most fragile part because they are generated by the same kinds of models being evaluated.

free parameters (3)
  • cosine similarity threshold = 0.85 to 0.90
    Used to decide which trusted knowledge entries are matched; paper states it is 'heuristically assigned based on empirical observations' in Section III-C.
  • number of filtering iterations = 5
    Self-NLI filtering was run for five iterations; no sensitivity analysis is provided (Section V-A, Step 5).
  • annotated seed ratio = 5% (plus 5% validation)
    The initial trusted knowledge base and training sets are built from 5% of each dataset; chosen by convention, not studied.
assumptions (5)
  • domain assumption Initial trusted knowledge base entries are correct
    The framework seeds its vector database with training-set 'correct' statements or authoritative-source extractions; if these are wrong, errors propagate. The authors acknowledge the reliance in Section VI-A.
  • domain assumption NLI entailment and contradiction labels from RoBERTa-MNLI indicate factual correctness against trusted knowledge
    The Contradiction Evaluation module treats entailment as credible and contradiction as erroneous; this mapping is assumed, not validated on domain data.
  • domain assumption LLM self-reported confidence scores track factual reliability
    Confidence Evaluation uses the model's own probability or self-assessment as a trustworthiness signal; this is contested in the cited literature and not calibrated here.
  • domain assumption Domain knowledge is interconnected enough that a nearest trusted statement can support or refute a candidate
    The whole iterative design depends on the premise stated in the introduction that 'internal knowledge is interconnected.'
  • ad hoc to paper The generated dataset labels are valid ground truth
    Biological labels are GPT-3.5 outputs, science labels are LLM extractions from ARC options, and radiation labels are assumed from authoritative sources; no human verification is reported (Section IV).

how reviews work

0 comments
Cite this review

Pith. "Pith review of TrustDataFilter:Leveraging Trusted Knowledge Base Data for More Effective Filtering of Unknown Information." pith.science (2026). https://pith.science/paper/2U5LGFVB

@misc{pith2026250215714,
  author       = {Pith},
  title        = {Pith review of: TrustDataFilter:Leveraging Trusted Knowledge Base Data for More Effective Filtering of Unknown Information},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2U5LGFVB}},
  note         = {Machine review of arXiv:2502.15714}
}
read the original abstract

With the advancement of technology and changes in the market, the demand for the construction of domain-specific knowledge bases has been increasing, either to improve model performance or to promote enterprise innovation and competitiveness. The construction of domain-specific knowledge bases typically relies on web crawlers or existing industry databases, leading to problems with accuracy and consistency of the data. To address these challenges, we considered the characteristics of domain data, where internal knowledge is interconnected, and proposed the Self-Natural Language Inference Data Filtering (self-nli-TDF) framework. This framework compares trusted filtered knowledge with the data to be filtered, deducing the reasoning relationship between them, thus improving filtering performance. The framework uses plug-and-play large language models for trustworthiness assessment and employs the RoBERTa-MNLI model from the NLI domain for reasoning. We constructed three datasets in the domains of biology, radiation, and science, and conducted experiments using RoBERTa, GPT3.5, and the local Qwen2 model. The experimental results show that this framework improves filter quality, producing more consistent and reliable filtering results.

Figures

Figures reproduced from arXiv: 2502.15714 by the authors.

Figure 1
Figure 1. Comparison of traditional filtering and the proposed [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. self-nli-datafilter architecture Overall Process: The process begins with the Data to Process stage, where each knowledge item undergoes initial filtering. Subsequently, the knowledge items are passed to the Trusted Knowledge Matching stage, where they are matched with trusted items in the vector knowledge base. If a match is successful, the knowledge item proceeds for further evaluation. If the match fails, the ite… view at source ↗
Figure 3
Figure 3. self-nli-datafilter architecture C. Trusted Knowledge Matching The process of aligning data with a trusted knowledge base is fundamental to effective contradiction evaluation. This process is organized into three critical components: knowledge base construction, vector matching techniques, and dynamic knowledge base expansion. Knowledge Base Construction: To begin, we construct a domain-specific knowledge base conta… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Distribution of Topics [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qwen Accuracy Improvement improvements over the Basic group, while being significantly outperformed by the Self-NLI group. For instance: In the Biological dataset, for the Qwen2 model, the accuracy of the Fake group slightly improves from the Basic group’s 0.8863 to 0.…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 18 canonical work pages

  1. [30]

    Bow- man

    Williams, Adina, Nikita Nangia, and Samuel R. Bow- man. ”A broad-coverage challenge corpus for sen- tence understanding through inference.” arXiv preprint arXiv:1704.05426 (2017)

    arXiv 2017

  2. [1]

    Wang, Huan, and Yan-Fu Li. ”Large Language Model Empowered by Domain-Specific Knowledge Base for In- dustrial Equipment Operation and Maintenance.” 2023 5th International Conference on System Reliability and Safety Engineering (SRSE). IEEE, 2023

    work page 2023

  3. [2]

    ”Yago 4: A reason-able knowledge base.” The Semantic Web: 17th International Conference, ESWC 2020, Heraklion, Crete, Greece, May 31–June 4, 2020, Proceedings 17

    Pellissier Tanon, Thomas, Gerhard Weikum, and Fabian Suchanek. ”Yago 4: A reason-able knowledge base.” The Semantic Web: 17th International Conference, ESWC 2020, Heraklion, Crete, Greece, May 31–June 4, 2020, Proceedings 17. Springer International Publishing, 2020

    work page 2020

  4. [3]

    Kumar, Sunil, and Ilyoung Chong. ”Correlation analysis to identify the effective data in machine learning: Prediction of depressive disorder and emotion states.” International journal of environmental research and public health 15.12 (2018): 2907

    work page 2018

  5. [4]

    ”RoBERTa: A robustly optimized BERT pre- training approach

    Liu, Y ., et al. ”RoBERTa: A robustly optimized BERT pre- training approach. arXiv [Preprint](2019).” arXiv preprint arXiv:1907.11692 (1907)

    arXiv 2019

  6. [5]

    ”Qwen2 technical report.” arXiv preprint arXiv:2407.10671 (2024)

    Yang, An, et al. ”Qwen2 technical report.” arXiv preprint arXiv:2407.10671 (2024)

    arXiv 2024

  7. [6]

    ”Language models are few-shot learners.” arXiv preprint arXiv:2005.14165 (2020)

    Brown, Tom B. ”Language models are few-shot learners.” arXiv preprint arXiv:2005.14165 (2020)

    arXiv 2020

  8. [7]

    Shalaby W, Arantes A, GonzalezDiaz T, et al. Build- ing chatbots from large scale domain-specific knowledge bases: Challenges and opportunities[C]//2020 IEEE Inter- national Conference on Prognostics and Health Manage- ment (ICPHM). IEEE, 2020: 1-8

    work page 2020

Show all 39 references
  1. [8]

    ”Domain-specific knowledge graphs: A survey.” Journal of Network and Computer Applications 185 (2021): 103076

    Abu-Salih, Bilal. ”Domain-specific knowledge graphs: A survey.” Journal of Network and Computer Applications 185 (2021): 103076

    2021

  2. [9]

    ”Precision at Scale: Domain-Specific Datasets On-Demand.” arXiv preprint arXiv:2407.03463 (2024)

    Rodr ´ıguez-de-Vera, Jes ´us M., et al. ”Precision at Scale: Domain-Specific Datasets On-Demand.” arXiv preprint arXiv:2407.03463 (2024)

    2024 arXiv

  3. [10]

    Malviya-Thakur, Addi, and Audris Mockus. ”The Role of Data Filtering in Open Source Software Ranking and Selection.” Proceedings of the 1st IEEE/ACM Interna- tional Workshop on Methodological Issues with Empirical Studies in Software Engineering. 2024

    2024

  4. [11]

    ”Data filtering networks.” arXiv preprint arXiv:2309.17425 (2023)

    Fang, Alex, et al. ”Data filtering networks.” arXiv preprint arXiv:2309.17425 (2023)

    2023 arXiv

  5. [12]

    Zheng, Danna, Mirella Lapata, and Jeff Z. Pan. ”Large language models as reliable knowledge bases?.” arXiv preprint arXiv:2407.13578 (2024)

    2024 arXiv

  6. [13]

    ”Calibrating Reasoning in Lan- guage Models with Internal Consistency.” arXiv preprint arXiv:2405.18711 (2024)

    Xie, Zhihui, et al. ”Calibrating Reasoning in Lan- guage Models with Internal Consistency.” arXiv preprint arXiv:2405.18711 (2024)

    2024 arXiv

  7. [14]

    ”Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks.” arXiv preprint arXiv:1908.10084 (2019)

    Reimers, N. ”Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks.” arXiv preprint arXiv:1908.10084 (2019)

    2019 arXiv

  8. [15]

    ”Milvus: A purpose-built vector data management system.” Proceedings of the 2021 Inter- national Conference on Management of Data

    Wang, Jianguo, et al. ”Milvus: A purpose-built vector data management system.” Proceedings of the 2021 Inter- national Conference on Management of Data. 2021

    2021

  9. [17]

    ”Deep learning.” nature 521.7553 (2015): 436-444

    LeCun, Yann, Yoshua Bengio, and Geoffrey Hinton. ”Deep learning.” nature 521.7553 (2015): 436-444

    2015

  10. [18]

    ”Retrieval-augmented generation for large language models: A survey.” arXiv preprint arXiv:2312.10997 (2023)

    Gao, Yunfan, et al. ”Retrieval-augmented generation for large language models: A survey.” arXiv preprint arXiv:2312.10997 (2023)

    2023 arXiv

  11. [19]

    ”Retrieval-augmented generation for knowledge-intensive nlp tasks.” Advances in Neural Information Processing Systems 33 (2020): 9459-9474

    Lewis, Patrick, et al. ”Retrieval-augmented generation for knowledge-intensive nlp tasks.” Advances in Neural Information Processing Systems 33 (2020): 9459-9474

    2020

  12. [20]

    ”Natural Language Processing with Commonsense Knowledge: A Survey.” IEEE Transactions on Knowledge and Data Engineering (2024)

    Xie, Yubo, et al. ”Natural Language Processing with Commonsense Knowledge: A Survey.” IEEE Transactions on Knowledge and Data Engineering (2024)

    2024

  13. [21]

    & Forte, G

    Morelli, M., Casagrande, M. & Forte, G. Decision Making: a Theoretical Review. Integr. psych. behav. 56, 609–629 (2022). https://doi.org/10.1007/s12124-021- 09669-x

    2022 doi

  14. [22]

    ”Auto-RAG: Autonomous Retrieval-Augmented Generation for Large Language Models.” arXiv preprint arXiv:2411.19443 (2024)

    Yu, Tian, Shaolei Zhang, and Yang Feng. ”Auto-RAG: Autonomous Retrieval-Augmented Generation for Large Language Models.” arXiv preprint arXiv:2411.19443 (2024)

    2024 arXiv

  15. [23]

    ”A large annotated corpus for learning natural language inference.” arXiv preprint arXiv:1508.05326 (2015)

    Bowman, Samuel R., et al. ”A large annotated corpus for learning natural language inference.” arXiv preprint arXiv:1508.05326 (2015)

    2015 arXiv

  16. [24]

    Quinlan, J. Ross. ”Induction of decision trees.” Machine learning 1 (1986): 81-106

    1986

  17. [25]

    ”A survey of language model confidence estimation and calibration.” arXiv preprint arXiv:2311.08298 (2023)

    Geng, Jiahui, et al. ”A survey of language model confidence estimation and calibration.” arXiv preprint arXiv:2311.08298 (2023)

    2023 arXiv

  18. [26]

    A survey on semi- supervised learning[J]

    Van Engelen J E, Hoos H H. A survey on semi- supervised learning[J]. Machine learning, 2020, 109(2): 373-440

    2020

  19. [27]

    ”Retrieval augmented language model pre-training.” International conference on machine learn- ing

    Guu, Kelvin, et al. ”Retrieval augmented language model pre-training.” International conference on machine learn- ing. PMLR, 2020

    2020

  20. [28]

    ”Dense passage retrieval for open-domain question answering.” arXiv preprint arXiv:2004.04906 (2020)

    Karpukhin, Vladimir, et al. ”Dense passage retrieval for open-domain question answering.” arXiv preprint arXiv:2004.04906 (2020)

    2020 arXiv

  21. [29]

    JOURNAL OF LATEX CLASS FILES, VOL

    Steck, Harald, Chaitanya Ekanadham, and Nathan Kallus. JOURNAL OF LATEX CLASS FILES, VOL. 18, NO. 9, SEPTEMBER 2020 13 ”Is cosine-similarity of embeddings really about simi- larity?.” Companion Proceedings of the ACM on Web Conference 2024. 2024

    2020

  22. [31]

    ”Factual Confidence of LLMs: on Reliability and Robustness of Current Estimators.” arXiv preprint arXiv:2406.13415 (2024)

    Mahaut, Mat ´eo, et al. ”Factual Confidence of LLMs: on Reliability and Robustness of Current Estimators.” arXiv preprint arXiv:2406.13415 (2024)

    2024 arXiv

  23. [32]

    ”Assessing the reliability of large language model knowledge.” arXiv preprint arXiv:2310.09820 (2023)

    Wang, Weixuan, et al. ”Assessing the reliability of large language model knowledge.” arXiv preprint arXiv:2310.09820 (2023)

    2023 arXiv

  24. [33]

    Rasoul, and David Landgrebe

    Safavian, S. Rasoul, and David Landgrebe. ”A survey of decision tree classifier methodology.” IEEE transactions on systems, man, and cybernetics 21.3 (1991): 660-674

    1991

  25. [34]

    ”Self-instruct: Aligning language models with self-generated instructions.” arXiv preprint arXiv:2212.10560 (2022)

    Wang, Yizhong, et al. ”Self-instruct: Aligning language models with self-generated instructions.” arXiv preprint arXiv:2212.10560 (2022)

    2022 arXiv

  26. [35]

    ”Think you have solved question answering? try arc, the ai2 reasoning challenge.” arXiv preprint arXiv:1803.05457 (2018)

    Clark, Peter, et al. ”Think you have solved question answering? try arc, the ai2 reasoning challenge.” arXiv preprint arXiv:1803.05457 (2018)

    2018 arXiv

  27. [36]

    ”Large language models for generative information extraction: A survey.” Frontiers of Computer Science 18.6 (2024): 186357

    Xu, Derong, et al. ”Large language models for generative information extraction: A survey.” Frontiers of Computer Science 18.6 (2024): 186357

    2024

  28. [37]

    ”Large Language Models Help Humans Verify Truthfulness–Except When They Are Convincingly Wrong.” arXiv preprint arXiv:2310.12558 (2023)

    Si, Chenglei, et al. ”Large Language Models Help Humans Verify Truthfulness–Except When They Are Convincingly Wrong.” arXiv preprint arXiv:2310.12558 (2023)

    2023 arXiv

  29. [38]

    ”ASSET: A dataset for tuning and evaluation of sentence simplification models with multiple rewriting transformations.” arXiv preprint arXiv:2005.00481 (2020)

    Alva-Manchego, Fernando, et al. ”ASSET: A dataset for tuning and evaluation of sentence simplification models with multiple rewriting transformations.” arXiv preprint arXiv:2005.00481 (2020)

    2020 arXiv

  30. [39]

    ”BERTopic: Neural topic model- ing with a class-based TF-IDF procedure.” arXiv preprint arXiv:2203.05794 (2022)

    Grootendorst, Maarten. ”BERTopic: Neural topic model- ing with a class-based TF-IDF procedure.” arXiv preprint arXiv:2203.05794 (2022)

    2022 arXiv

  31. [40]

    ”Logic-consistency text generation from semantic parses.” arXiv preprint arXiv:2108.00577 (2021)

    Shu, Chang, et al. ”Logic-consistency text generation from semantic parses.” arXiv preprint arXiv:2108.00577 (2021)

    2021 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.