Pith. sign in

REVIEW 5 major objections 5 minor 31 references

ForPKG: A Framework for Constructing Forestry Policy Knowledge Graph and Application Analysis

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims a three-step pipeline using an open-source large language model can extract forestry policy triples with 76.2% precision and 62.6% recall, beating unsupervised baselines.

desk verdict The ontology and the promise of a released KG are the real contribution; the headline extraction numbers are plausible but unsupported because the paper never states that the 150 training samples and the 50 evaluation documents are disjoint. read the letter →

arxiv 2411.11090 v2 pith:F3NVYMUA submitted 2024-11-17 cs.IR

Jingyun Sun , Zhongze Luo This is my paper · ORCID
classification cs.IR
keywords knowledgegraphconstructionpolicyforestryontologydesigninformationextractionpromptlearninglargelanguagemodelsretrieval-augmentedgeneration
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 argues that a usable forestry policy knowledge graph can be built without large labeled datasets by combining a fine-grained ontology with a decomposed open-source large-language-model pipeline. The ontology defines 10 entity types and 15 relationship types, including normative relations such as duty, prohibition, and right drawn from deontic logic. The extraction pipeline first uses an LLM to find head entities, then a prompt-learned text classifier to pick the relation, then the LLM to find the tail entity; on 1,126 manually annotated triples from 50 Chinese forestry policy documents it reports 76.2% precision and 62.6% recall, above direct LLM calls and off-the-shelf extraction frameworks. The authors also show that when the resulting graph is used as a retrieval source for a large language model, answer quality improves, which is the practical payoff they care about.

What carries the argument

The load-bearing machinery is the three-step content-level extraction process: (1) Kimi, an open-source Chinese LLM, is prompted to recognize head entities from the full policy document; (2) a DeBERTa model fine-tuned with prompt learning, using relation names as the verbalizer, classifies which of the 15 ontology relations applies; (3) Kimi locates the relation word and extracts the tail entity. Step order matters because long tail entities tend to follow the relation word, so most of the difficult span-detection is deferred to a pattern-assisted step, and the abstract relations are converted into a bounded classification task. This pipeline is what carries the reported performance advantage and what makes the framework transferable to other policy domains.

What would settle it

Re-run the same three-step pipeline on 50 held-out forestry policy documents whose texts were never used to build the 150 relationship-classification examples, and compare precision and recall against Table III; if the numbers fall sharply, the published estimate depends on training/test overlap. A second independent annotator re-marking the same 50 documents would also test whether the 1,126 triples are stable enough to serve as a gold standard.

Watch

Extended reading notes

Core claim

The central claim is that the hard part of policy knowledge extraction is not language understanding but division of labour: asking one model to produce complete triples in one pass fails, while splitting the job into head-entity recognition, relation classification, and tail-entity recognition succeeds on long abstract normative entities like '退耕还林还草补助资金兑付' and on relations such as obligation and prohibition. The paper further claims that relation classification is best treated as a few-shot text classification problem, with 150 hand-built examples and a prompt-learning-fine-tuned DeBERTa model, rather than as an open-ended LLM generation task. On its test corpus, this design gives 76.2% precision and 62.6% recall, and the paper argues this confirms the ontology is expressive and extensible enough to support both direct policy applications and retrieval-augmented generation.

Load-bearing premise

The 150 relationship-classification examples used to train the prompt-learning classifier are never explicitly separated from the 50 documents used to measure precision and recall, and the authors' manual annotation is the only gold standard; if those overlap or the annotation is unreliable, the reported numbers do not measure real extraction performance.

Editorial extensions

If this is right

  • A policy knowledge graph that reaches inside the text, rather than just linking documents to metadata, can support compliance checking, policy analysis, and intelligent policy question answering.
  • The extraction recipe needs only about 150 labeled examples, so other low-resource policy fields can imitate it without building large annotated corpora.
  • Feeding the graph into a retrieval-augmented generation system lifts the generator's correctness, effectiveness, and fluency scores in the paper's evaluation.
  • The ontology's 10 entity types and 15 relationship types can be merged with general knowledge graphs, and the paper reports improved query response accuracy after fusion.
  • Because the ontology and pipeline are not forestry-specific, the same framework is intended to transfer to healthcare, insurance, and finance policy text.

Reading between the lines

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

  • The paper never explicitly separates the 150 prompt-learning training examples from the 50 evaluation documents, so the 76.2%/62.6% figures should be re-audited on fully held-out documents before the comparison is taken at face value.
  • Since tail-entity extraction relies on finding the relation word, sentences that express obligations or definitions without an explicit cue word are likely to be the pipeline's weak spot; a stress test with cue-poor sentences would reveal the true ceiling.
  • Calling the method unsupervised is softened by 150 hand-built examples; a zero-shot variant that omits the DeBERTa step would test which part of the pipeline is doing the work.
  • The application analysis depends on an LLM judge rather than human raters, so a small human evaluation would make the RAG improvement claim easier to rely on.
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

5 major / 5 minor

Summary. The paper proposes ForPKG, a framework for constructing a forestry policy knowledge graph. It defines a 10-type/15-relation ontology, a three-step extraction pipeline (Kimi-based head-entity recognition, DeBERTa prompt-learning relation classification, Kimi-based tail-entity recognition), and reports precision/recall on triple extraction from 50 Chinese forestry policy documents. It also presents a RAG application evaluation scored by GPT-4 and announces that the knowledge graph will be released on GitHub. The paper's main empirical claim is that the proposed extraction method outperforms unsupervised off-the-shelf tools and direct LLM baselines, with Table III reporting 76.2% precision and 62.6% recall.

Significance. If the reported extraction numbers were valid, the paper would be a useful contribution to domain-specific knowledge graph construction: the ontology is grounded in deontic logic, the pipeline is concrete and reproducible in principle, and the planned public release of the knowledge graph is a genuine community resource. The comparison against several off-the-shelf extraction tools and direct LLM baselines is a reasonable starting point. However, the central empirical claim is not currently supported because the paper does not separate the supervision source from the evaluation set, mischaracterizes the method as unsupervised, and provides no statistical or agreement evidence. As it stands, the contribution is a framework proposal whose headline quantitative results require substantial additional validation.

major comments (5)
  1. [Section IV-C2; Section V-B1; Table III] The paper never states that the 150 prompt-learning training samples (10 per relationship type) used to fine-tune DeBERTa are disjoint from the 50 documents whose 1,126 triples are used to compute Table III. If the training examples and evaluation documents overlap, the relation classifier can memorize entity/pattern associations, and the 76.2%/62.6% headline figures would overstate generalization. The authors must state the split explicitly and, if necessary, re-run the evaluation on held-out documents.
  2. [Abstract; Section IV-C2] The method is called 'unsupervised' in the abstract and in Section IV, but Section IV-C2 fine-tunes DeBERTa with 150 manually labeled examples. This makes the comparison against DeepKE, LTP, Jiagu, and DeepDive, described as 'unsupervised,' not apples-to-apples. The method should be described as few-shot or lightweight-supervised, and the baselines should include supervised or few-shot alternatives; otherwise the reported margin could be attributed to the use of labels rather than to the proposed architecture.
  3. [Section V-B1; Table III] No variance is reported for any row of Table III, and no inter-annotator agreement is given for the manually annotated gold triples. LLM API outputs are stochastic, so point estimates without repeated runs or confidence intervals do not support the claim that the improvement over the strongest direct-LLM baseline (50.7%/23.3%) is significant. The authors should report means and standard deviations over multiple runs, and at least one agreement metric for the gold annotation.
  4. [Section V-A; Figure 2] The ontology extensibility experiment is not described in sufficient detail: the text does not state how many queries were used, what the 'response accuracy' measure is, or how the fusion with OwnThink, OpenKG, CN-Dbpedia, DBpedia, and OpenConcept was performed. Without this information, Figure 2 cannot be interpreted and the claim that the ontology has 'good expressiveness and extensibility' is not supported.
  5. [Section V-C; Figure 5] The RAG evaluation uses GPT-4 as the scorer but reports no number of test questions, no sample prompts, no validation of the GPT-4 scoring, and no statistical test or confidence intervals. The claim that 'the practical application value of the knowledge graph ... is confirmed' goes beyond what this evidence can establish. The authors should provide the test set, scoring template, agreement between GPT-4 and human raters, and significance testing.
minor comments (5)
  1. [Algorithm 1, line 6] The formula uses a slash where a conditional-probability bar is intended; write log P([MASK] = phi(y_i) | T(s_i, h_i)).
  2. [Table II; Section IV-B] Table II contains the typos 'Iinverse Relationship' and 'Contrain'; also, the 'relevant' relation is described as reflexive, but the construction via pairwise cosine similarity yields an undirected/symmetric relation, not a reflexive one.
  3. [Section IV-B] The Chinese checkpoint of BigBird is mentioned without a citation; provide the source or model identifier so the encoding step is reproducible.
  4. [Section V-B2; Figure 3] The 'accuracy rates' shown in the radar chart should be defined as precision, recall, or F1, and the number of instances per entity/relationship type should be reported, because the figure is hard to interpret without denominators.
  5. [Abstract; Section VII] The GitHub repository is mentioned in the abstract, but the experimental section does not state which materials (documents, annotations, fine-tuned model, or evaluation scripts) are actually included; please clarify the release contents at the point where the experiments are described.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the main extraction result is compared against external, off-the-shelf baselines and direct LLM calls, so the central claim is not reduced to the paper's own inputs.

full rationale

The paper's central quantitative claim—that the proposed three-step extraction method reaches 76.2% precision and 62.6% recall on 1,126 manually annotated triples from 50 forestry policy documents—is evaluated against independent baselines: DeepKE, LTP, Jiagu, DeepDive, direct Kimi calls, and direct LLaMa2-Chinese calls (Table III). These baselines are not derived from the paper's ontology or extraction pipeline, so the comparison is externally grounded rather than circular. The prompt-learning relation classifier is trained on 150 manually constructed examples (Section IV-C2) and the method is evaluated on a separately described set of 50 manually annotated documents (Section V-B1). The paper does not state whether those 150 training examples are disjoint from the 50 evaluation documents; if they overlap, the reported numbers could overstate generalization. However, this is a potential experimental-design weakness and a validity risk, not a demonstrated circular reduction: the paper does not show that the evaluation set is a subset of the training set, nor does it fit a parameter to the evaluation labels and then rename that fit as a prediction. The ontology is used to define the annotation schema, but that is normal practice for domain knowledge-graph evaluation and does not by itself make the extraction result an input. The RAG evaluation uses GPT-4 as an external judge, and the paper's use of the term 'unsupervised' is inaccurate because 150 labeled examples are used; these are labeling and evaluation-quality concerns, not instances of the derivation being equivalent to its inputs. No self-citation chain, imported uniqueness theorem, or ansatz-smuggling-via-citation is present. Given the external benchmarks, the appropriate finding is that the central derivation is self-contained, with a minor unstated train/evaluation overlap risk that does not rise to demonstrated circularity.

Assumptions & free parameters 3 free parameters · 5 assumptions · 1 invented entities

The central claim rests on a newly designed ontology and a manual evaluation set created by the authors. There are no external benchmarks for the ontology, a potentially overlapping train/test design, and a proprietary LLM API; these choices are documented in Sections IV and V.

free parameters (3)
  • Cosine similarity threshold lambda = not reported
    Section IV-B: a 'relevant' edge is added only when BigBird document embedding similarity exceeds lambda; no value or sensitivity analysis is given, so the document-level graph depends on an unstated choice.
  • Prompt-learning training set size (10 examples per relation) = 150 samples
    Section IV-C2: this size was chosen by hand; Figure 4 shows performance degrades with fewer samples, making the reported relation classification quality dependent on this manual choice.
  • DeBERTa fine-tuning hyperparameters = not reported
    Section IV-C2 and Algorithm 1 give no learning rate, epochs, or batch size, so the trained relation classifier is not fully specified.
assumptions (5)
  • domain assumption The deontic-logic categories (ACT, STATE, duty, prohibit, hasRight) are the correct semantic representation for Chinese forestry policy norms.
    Section III introduces these categories with a single example sentence and cites Von Wright's deontic logic; no corpus-driven or user validation is provided.
  • domain assumption Kimi API outputs are stable and accurate enough for repeated head-entity and tail-entity recognition.
    Section IV-C relies on Kimi for both entity recognition steps but does not measure API variability or provide the exact prompts.
  • domain assumption Cosine similarity in BigBird embedding space is a valid measure of policy-document relevance.
    Section IV-B uses BigBird vector cosine similarity to create 'relevant' edges and set a threshold, without validating this similarity against human judgments.
  • domain assumption GPT-4 scores are a valid proxy for human judgment of answer correctness, effectiveness, and fluency.
    Section V-C uses GPT-4 as the only evaluator of RAG output quality with no human validation or correlation analysis.
  • domain assumption A single manual annotation pass on 50 documents yields reliable gold-standard triples.
    Section V-B1 describes manual annotation of 1,126 triples but reports no inter-annotator agreement or adjudication protocol.
invented entities (1)
  • ACT and STATE entity types plus duty, prohibit, and hasRight relationship types
    purpose: to model normative content such as obligations and prohibitions in forestry policy text
    Introduced in Section III based on deontic logic; they are not validated against an external ontology or user study, and their contribution is measured only through the authors' own extraction experiment.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ForPKG: A Framework for Constructing Forestry Policy Knowledge Graph and Application Analysis." pith.science (2026). https://pith.science/paper/F3NVYMUA

@misc{pith2026241111090,
  author       = {Pith},
  title        = {Pith review of: ForPKG: A Framework for Constructing Forestry Policy Knowledge Graph and Application Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F3NVYMUA}},
  note         = {Machine review of arXiv:2411.11090}
}
read the original abstract

A policy knowledge graph can provide decision support for tasks such as project compliance, policy analysis, and intelligent question answering, and can also serve as an external knowledge base to assist the reasoning process of related large language models. Although there have been many related works on knowledge graphs, there is currently a lack of research on the construction methods of policy knowledge graphs. This paper, focusing on the forestry field, designs a complete policy knowledge graph construction framework, including: firstly, proposing a fine-grained forestry policy domain ontology; then, proposing an unsupervised policy information extraction method, and finally, constructing a complete forestry policy knowledge graph. The experimental results show that the proposed ontology has good expressiveness and extensibility, and the policy information extraction method proposed in this paper achieves better results than other unsupervised methods. Furthermore, by analyzing the application of the knowledge graph in the retrieval-augmented-generation task of the large language models, the practical application value of the knowledge graph in the era of large language models is confirmed. The knowledge graph resource will be released on an open-source platform and can serve as the basic knowledge base for forestry policy-related intelligent systems. It can also be used for academic research. In addition, this study can provide reference and guidance for the construction of policy knowledge graphs in other fields. Our data is provided on Github https://github.com/luozhongze/ForPKG.

Figures

Figures reproduced from arXiv: 2411.11090 by the authors.

Figure 1
Figure 1. The workflow for building a forestry policy knowledge graph using [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The improvement in response accuracy after merging the forestry [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Accuracy on different entity types and relationship types. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Training sample dependency analysis. C. Application-based practical analysis To further analyze the practical application effects of the forestry policy knowledge graph we have built, we used it as the RAG knowledge source for a large model. We used the LangChain frame…
Figure 5
Figure 5. Figure 5: Evaluation results of answer generation quality before and after [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 26 canonical work pages

  1. [1]

    Knowledge graph embedding by translating on hyperplanes,

    Z. Wang, J. Zhang, J. Feng, and Z. Chen, “Knowledge graph embedding by translating on hyperplanes,” in Pro- ceedings of the AAAI conference on artificial intelligence, vol. 28, no. 1, 2014

    work page 2014

  2. [2]

    A review: Knowledge reasoning over knowledge graph,

    X. Chen, S. Jia, and Y . Xiang, “A review: Knowledge reasoning over knowledge graph,” Expert systems with applications, vol. 141, p. 112948, 2020

    work page 2020

  3. [3]

    Introduction: what is a knowledge graph?

    D. Fensel, U. S ¸ims ¸ek, K. Angele, E. Huaman, E. K¨arle, O. Panasiuk, I. Toma, J. Umbrich, A. Wahler, D. Fensel et al., “Introduction: what is a knowledge graph?”Knowl- edge graphs: Methodology, tools and selected use cases , pp. 1–10, 2020

    work page 2020

  4. [4]

    Kgscs—a smart care system for elderly with geriatric chronic diseases: a knowledge graph approach,

    A. Li, C. Han, X. Xing, Q. Wei, Y . Chi, and F. Pu, “Kgscs—a smart care system for elderly with geriatric chronic diseases: a knowledge graph approach,” BMC Medical Informatics and Decision Making, vol. 24, no. 1, p. 73, 2024

    work page 2024

  5. [5]

    Modal-adaptive knowledge-enhanced graph- based financial prediction from monetary policy confer- ence calls with llm,

    K. Ouyang, Y . Liu, S. Li, R. Bao, K. Harimoto, and X. Sun, “Modal-adaptive knowledge-enhanced graph- based financial prediction from monetary policy confer- ence calls with llm,” arXiv preprint arXiv:2403.16055 , 2024

    arXiv 2024

  6. [6]

    Dbpedia: A nucleus for a web of open data,

    S. Auer, C. Bizer, G. Kobilarov, J. Lehmann, R. Cyga- niak, and Z. Ives, “Dbpedia: A nucleus for a web of open data,” in international semantic web conference . Springer, 2007, pp. 722–735

    work page 2007

  7. [7]

    Cn-dbpedia: A never-ending chinese knowl- edge extraction system,

    B. Xu, Y . Xu, J. Liang, C. Xie, B. Liang, W. Cui, and Y . Xiao, “Cn-dbpedia: A never-ending chinese knowl- edge extraction system,” in International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems . Springer, 2017, pp. 428– 438

    work page 2017

  8. [8]

    Yago: a core of semantic knowledge,

    F. M. Suchanek, G. Kasneci, and G. Weikum, “Yago: a core of semantic knowledge,” in Proceedings of the 16th international conference on World Wide Web , 2007, pp. 697–706

    work page 2007

Show all 31 references
  1. [9]

    Knowledge graph construction based on ship collision accident reports to improve maritime traffic safety,

    L. Gan, B. Ye, Z. Huang, Y . Xu, Q. Chen, and Y . Shu, “Knowledge graph construction based on ship collision accident reports to improve maritime traffic safety,” Ocean & Coastal Management , vol. 240, p. 106660, 2023

    2023

  2. [10]

    Knowledge graphs: Opportunities and challenges,

    C. Peng, F. Xia, M. Naseriparsa, and F. Osborne, “Knowledge graphs: Opportunities and challenges,” Ar- tificial Intelligence Review , vol. 56, no. 11, pp. 13 071– 13 102, 2023

    2023

  3. [11]

    A comprehensive survey on automatic knowledge graph construction,

    L. Zhong, J. Wu, Q. Li, H. Peng, and X. Wu, “A comprehensive survey on automatic knowledge graph construction,” ACM Computing Surveys , vol. 56, no. 4, pp. 1–62, 2023

    2023

  4. [12]

    Survey of open-domain knowledge graph question an- swering

    C. Zirui, W. Xin, W. Lin, X. Dawei, and J. Yongzhe, “Survey of open-domain knowledge graph question an- swering.” Journal of Frontiers of Computer Science & Technology, vol. 15, no. 10, 2021

    2021

  5. [13]

    Domain-specific knowledge graphs: A survey,

    B. Abu-Salih, “Domain-specific knowledge graphs: A survey,” Journal of Network and Computer Applications , vol. 185, p. 103076, 2021

    2021

  6. [14]

    Wikidata: a free collab- orative knowledgebase,

    D. Vrande ˇci´c and M. Kr ¨otzsch, “Wikidata: a free collab- orative knowledgebase,” Communications of the ACM , vol. 57, no. 10, pp. 78–85, 2014

    2014

  7. [15]

    An automatic knowledge graph construction system for k- 12 education,

    P. Chen, Y . Lu, V . W. Zheng, X. Chen, and X. Li, “An automatic knowledge graph construction system for k- 12 education,” in Proceedings of the fifth annual ACM conference on learning at scale , 2018, pp. 1–4

    2018

  8. [16]

    Knowledge graph- enhanced molecular contrastive learning with functional prompt,

    Y . Fang, Q. Zhang, N. Zhang, Z. Chen, X. Zhuang, X. Shao, X. Fan, and H. Chen, “Knowledge graph- enhanced molecular contrastive learning with functional prompt,” Nature Machine Intelligence , vol. 5, no. 5, pp. 542–553, 2023

    2023

  9. [17]

    Construction and application of an ontology-based domain-specific knowl- edge graph for petroleum exploration and development,

    X. Tang, Z. Feng, Y . Xiao, M. Wang, T. Ye, Y . Zhou, J. Meng, B. Zhang, and D. Zhang, “Construction and application of an ontology-based domain-specific knowl- edge graph for petroleum exploration and development,” Geoscience Frontiers, vol. 14, no. 5, p. 101426, 2023

    2023

  10. [18]

    Real- world data medical knowledge graph: construction and applications,

    L. Li, P. Wang, J. Yan, Y . Wang, S. Li, J. Jiang, Z. Sun, B. Tang, T.-H. Chang, S. Wang et al. , “Real- world data medical knowledge graph: construction and applications,” Artificial intelligence in medicine, vol. 103, p. 101817, 2020

    2020

  11. [19]

    Chinese named entity recognition: The state of the art,

    P. Liu, Y . Guo, F. Wang, and G. Li, “Chinese named entity recognition: The state of the art,” Neurocomputing, vol. 473, pp. 37–53, 2022

    2022

  12. [20]

    Relation classification via convolutional deep neural network,

    D. Zeng, K. Liu, S. Lai, G. Zhou, and J. Zhao, “Relation classification via convolutional deep neural network,” in Proceedings of COLING 2014, the 25th international conference on computational linguistics: technical pa- pers, 2014, pp. 2335–2344

    2014

  13. [21]

    Joint entity and relation extraction for legal documents with legal feature enhancement,

    Y . Chen, Y . Sun, Z. Yang, and H. Lin, “Joint entity and relation extraction for legal documents with legal feature enhancement,” in Proceedings of the 28th Inter- national Conference on Computational Linguistics, 2020, pp. 1561–1571

    2020

  14. [22]

    Structured infor- mation extraction from scientific text with large language models,

    J. Dagdelen, A. Dunn, S. Lee, N. Walker, A. S. Rosen, G. Ceder, K. A. Persson, and A. Jain, “Structured infor- mation extraction from scientific text with large language models,” Nature Communications, vol. 15, no. 1, p. 1418, 2024

    2024

  15. [23]

    Finbert: A large language model for extracting information from financial text,

    A. H. Huang, H. Wang, and Y . Yang, “Finbert: A large language model for extracting information from financial text,” Contemporary Accounting Research, vol. 40, no. 2, pp. 806–841, 2023

    2023

  16. [24]

    Deontic logic,

    G. H. V on Wright, “Deontic logic,” Mind, vol. 60, no. 237, pp. 1–15, 1951

    1951

  17. [25]

    Glm: General language model pretrain- ing with autoregressive blank infilling,

    Z. Du, Y . Qian, X. Liu, M. Ding, J. Qiu, Z. Yang, and J. Tang, “Glm: General language model pretrain- ing with autoregressive blank infilling,” arXiv preprint arXiv:2103.10360, 2021

    2021 arXiv

  18. [26]

    Llama 2: Open foundation and fine-tuned chat models,

    H. Touvron, L. Martin, K. Stone, P. Albert, A. Almahairi, Y . Babaei, N. Bashlykov, S. Batra, P. Bhargava, S. Bhos- ale et al., “Llama 2: Open foundation and fine-tuned chat models,” arXiv preprint arXiv:2307.09288 , 2023

    2023 arXiv

  19. [27]

    Baichuan 2: Open large-scale language models,

    A. Yang, B. Xiao, B. Wang, B. Zhang, C. Bian, C. Yin, C. Lv, D. Pan, D. Wang, D. Yan et al. , “Baichuan 2: Open large-scale language models,” arXiv preprint arXiv:2309.10305, 2023

    2023 arXiv

  20. [28]

    Exploiting cloze questions for few shot text classification and natural language inference,

    T. Schick and H. Sch ¨utze, “Exploiting cloze questions for few shot text classification and natural language inference,” arXiv preprint arXiv:2001.07676 , 2020

    2001 arXiv

  21. [29]

    Deepke: A deep learning based knowledge extraction toolkit for knowledge base population,

    N. Zhang, X. Xu, L. Tao, H. Yu, H. Ye, S. Qiao, X. Xie, X. Chen, Z. Li, L. Li et al. , “Deepke: A deep learning based knowledge extraction toolkit for knowledge base population,” arXiv preprint arXiv:2201.03335 , 2022

    2022 arXiv

  22. [30]

    N-ltp: An open- source neural language technology platform for chinese,

    W. Che, Y . Feng, L. Qin, and T. Liu, “N-ltp: An open- source neural language technology platform for chinese,” arXiv preprint arXiv:2009.11616 , 2020

    2009 arXiv

  23. [31]

    Deepdive: Web-scale knowledge-base construction using statistical learning and inference

    F. Niu, C. Zhang, C. R ´e, and J. W. Shavlik, “Deepdive: Web-scale knowledge-base construction using statistical learning and inference.” VLDS, vol. 12, pp. 25–28, 2012

    2012

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.