Pith. sign in

REVIEW 2 major objections 5 minor 73 references

From Judgments to Issues: Structured Extraction of Legal Reasoning with Citation-Hallucination Control

T0 review · 2 major / 5 minor · reviewed 2026-07-12 · grok-4.5

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

Pith's one-line read An automated pipeline turns Italian tax-court judgments into issue-level XML records with citation-hallucination filtering, validated by tax-law experts.

desk verdict Solid methods paper: first expert-validated issue-level IRAC/XML pipeline with citation-hallucination control for Italian tax courts, run at scale and honestly evaluated. read the letter →

arxiv 2607.03325 v1 pith:XMQMERJH submitted 2026-07-03 cs.CL cs.AIcs.IR

Giovanni Piccioli , Alessia Fidelangeli , Piera Santin , Pierpaolo Vivo This is my paper
classification cs.CLcs.AIcs.IR
keywords legalNLPissueextractionIRACsyllogismcitationhallucinationItaliantaxcourtsstructuredXMLschema
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 builds a scalable pipeline that breaks each Italian tax-court judgment into autonomous legal issues and fills a structured XML schema inspired by IRAC and the legal syllogism. The unit of analysis is the single legal question the judge actually decides, not the whole document, so each record is self-contained and usable for search, citation graphs, and reasoning datasets. Extraction runs on a cost-efficient general-purpose model over roughly 329,000 decisions; a post-hoc filter then checks every extracted citation against a dedicated legal-reference parser and drops anything that cannot be matched in the source text. On a 50-judgment sample annotated by two tax-law PhDs, issue extraction is high-precision and moderate-recall, residual citation hallucinations fall to under one percent, and free-text fields receive strong expert scores. The result is a concrete, expert-checked starting point for issue-level retrieval, cleaner citation networks, and large structured corpora of judicial reasoning.

What carries the argument

Issue-based XML schema plus citation-hallucination filter: each judgment is split into autonomous issues whose fields (Whether-clause, factual premises, typed legal references with citation reasons, judge reasoning, outcome, summary) are populated by the model, after which Linkoln-normalized identifiers are used to drop any reference not found in the source text.

What would settle it

Re-annotate a larger stratified sample of judgments (or re-run the same pipeline with a different mid-range model) and show that issue-level precision falls well below the reported ~94 percent, residual hallucination rate rises substantially above 0.9 percent, or free-text expert scores drop below the high 4-range.

Watch

Extended reading notes

Core claim

A cost-efficient LLM pipeline can decompose Italian tax-court judgments into issue-level XML records grounded in IRAC and the legal syllogism, and a post-extraction filter that matches citations against a normalized parser can drive residual citation hallucinations down to 0.9 percent while preserving nearly all valid references, as confirmed by expert annotation on a 50-judgment subsample.

Load-bearing premise

That the experts’ working definition of a legal issue, the fixed list of allowed principles, and the 50-judgment sample are stable enough for the reported precision, recall, and residual-hallucination rates to generalize to the full heterogeneous corpus and to other models or jurisdictions.

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

2 major / 5 minor

Summary. The paper presents a cost-efficient pipeline that decomposes Italian tax-court judgments into autonomous legal issues and extracts, for each issue, an IRAC/syllogism-inspired XML schema (Whether-clause, factual premises, typed legal references with citation reasons, judge reasoning, outcome, summary). Extraction uses DeepSeek V3 (with a two-step prompt for longer texts and a title-correction step), followed by XSD validation and a Linkoln-based hallucination filter that normalizes citations to URN-NIR/ECLI/CELEX and removes references not found in the source (with a separate whitelist of 182 principles). The pipeline is run on ~329k of 401k retrieved first- and second-instance decisions. Validation on 50 judgments (20 double-blind by two tax-law PhDs) reports high-precision issue extraction (P≈94%, R 70–82%), citation F1 ~72–77% after filtering, residual hallucination rate 0.9%, free-text Likert scores >4.6/5 with strong IAA (Gwet AC2), plus run-to-run non-determinism analysis and a Linkoln baseline comparison.

Significance. If the reported numbers hold, this is a concrete, reproducible methods contribution for legal NLP: an issue-level structured representation with explicit citation-hallucination control, expert-validated IAA ceilings, bootstrap CIs, filter confusion matrix, and released code/annotations. The cost figure (~$0.0035/judgment) and scale (~329k documents) make the resource usable for issue-level retrieval, cleaner citation networks, and reasoning datasets in a jurisdiction and court tier that have received little NLP attention. Strengths include the transparent validation protocol, the stand-alone filter evaluation (Tables 8–9), Appendix H on non-determinism, and the honest limitations section. The work is engineering/methods rather than a new theoretical result, but the combination of schema design, hallucination control, and expert validation is a solid template for similar corpora.

major comments (2)
  1. The central claim is sound for the stated scope (Italian tax courts, DeepSeek V3, N=50). The main load-bearing soft spot is already acknowledged in §7: generalization of the working definition of “legal issue” (§5.1), the 182-principle whitelist, and residual non-determinism (Appendix H, run-to-run F1 78.3% issues / 73.2% citations). These do not invalidate the reported metrics, but the abstract and contributions should keep the “first for this corpus” framing tightly scoped and avoid implying model- or jurisdiction-independence without further evidence.
  2. §6.2 / Appendix I: citation precision (~72–74%) is moderate because many non-hallucinated citations are judged irrelevant to the specific issue. The paper correctly notes that Linkoln on the Motivi section is competitive on recall. For the claimed use of issue-level citation networks, a short quantitative discussion of how much noise remains after issue attribution (vs. Motivi-only) would strengthen the “signal-to-noise” argument in §3.3; Appendix I is indicative but limited to 31 judgments and automatic matching.
minor comments (5)
  1. Abstract and §1.1: the “first …” claim is carefully hedged; ensure it remains so in any camera-ready version and that related Italian tax work (Dal Pont et al. PRODIGIT; Grundler et al.) is contrasted only on the issue-level + hallucination-control axes.
  2. §4.3: non-determinism at temperature 0 is documented in Appendix H; a one-sentence pointer in the main extraction section would help readers who skip appendices.
  3. Table 10 / Appendix F: the kappa paradox discussion is useful; keeping Gwet AC2 as primary and kappa in the appendix is the right choice—consider a brief note in the main text that high prevalence makes κ uninformative here.
  4. Listing 1 vs. Italian prompts (Appendix C): field names differ slightly between English schema and Italian tags (e.g. legal_references vs lista_riferimenti_diritto). A short mapping note would avoid confusion for reimplementers.
  5. Minor presentation: arXiv date line “July 7, 2026” and a few DOI/page notes in the bibliography look like placeholders; clean for final version. Ensure GitHub link remains stable.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: metrics are independent expert comparisons and pure set-difference filtering, not fitted or self-defined predictions.

full rationale

This is an engineering/methods paper whose load-bearing claims are (i) a designed IRAC/syllogism XML schema, (ii) a DeepSeek V3 extraction pipeline plus Linkoln-normalized citation filter, and (iii) empirical P/R/F1, residual-hallucination, and Likert numbers on 50 expert-annotated judgments. None of these reduce by construction to their inputs. The schema is a design choice refined by iterative prompting, not a quantity fitted to the validation set. The hallucination filter is a pure set-difference after URI/number-year/principle-list matching against the source text (Figure 2, Tables 8–9); it does not fit parameters to the residual rate it reports. Issue and citation metrics treat independent PhD annotations as ground truth (Sections 5–6), with IAA ceilings reported separately; no parameter is fitted on a subset and then ‘predicted’ on a related quantity. Self-citations (e.g. Linkoln, prior IRAC/syllogism prompting work) supply tools or background, not uniqueness theorems that force the present results. Residual non-determinism (Appendix H) and the 182-principle list are openly stated limitations, not circular steps. The derivation chain is therefore self-contained against external expert benchmarks; score 0 is the correct outcome.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The work is an empirical engineering pipeline; its load-bearing commitments are design choices and domain conventions rather than free parameters fitted to the evaluation set. The main free choices are length thresholds used for filtering and the expert-curated principle whitelist; the axioms are standard legal-NLP and IRAC assumptions plus the operational definition of a legal issue.

free parameters (3)
  • Motivi-section length filter (680 chars) and total-length filter (2300 chars)
    Hard thresholds chosen to discard judgments that contain no substantive reasoning; affect which documents enter the 328k processed corpus.
  • Two-step extraction length threshold (4100 chars)
    Heuristic that switches from single-prompt to two-step extraction for longer judgments; influences completeness of issue lists.
  • Whitelist of 182 allowed legal principles
    Expert-curated closed list used by the hallucination filter for type=princ references; any principle outside the list is discarded even if present in the text.
assumptions (4)
  • domain assumption A legal issue is a question explicitly decided by the judge in the reasoning section, requiring interpretation or application of a legal source, and supported by enough argumentative content to fill the XML fields.
    Operational definition jointly fixed by the two annotators (Section 5.1); residual subjectivity is acknowledged via IAA.
  • domain assumption IRAC and the legal syllogism supply a sufficiently coarse yet machine-extractable skeleton for judicial reasoning.
    Design premise of the XML schema (Section 3); single-step rather than multi-syllogism decomposition chosen for extractability.
  • domain assumption Linkoln plus URN-NIR/ECLI/CELEX normalization plus number-year window matching correctly identifies the set of citations present in a judgment.
    Foundation of the hallucination filter (Section 4.4); residual false negatives on authority mismatches are noted.
  • ad hoc to paper Zero-shot prompting of DeepSeek V3 at temperature 0 yields outputs whose quality is acceptable to tax-law experts at the chosen price point.
    Model-selection decision based on informal expert review of 10 development judgments (Appendix J).
invented entities (1)
  • Issue-level IRAC/XML schema with citation_reason and in_favor_of attributes
    purpose: Machine-readable, human-consultable unit of legal reasoning that supports retrieval, citation networks, and dataset construction.
    New postulated representation; independent evidence is the expert Likert scores and downstream-use arguments, not external physical measurement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Judgments to Issues: Structured Extraction of Legal Reasoning with Citation-Hallucination Control." pith.science (2026). https://pith.science/paper/XMQMERJH

@misc{pith2026260703325,
  author       = {Pith},
  title        = {Pith review of: From Judgments to Issues: Structured Extraction of Legal Reasoning with Citation-Hallucination Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XMQMERJH}},
  note         = {Machine review of arXiv:2607.03325}
}
abstract

We present an automated pipeline that decomposes Italian tax-court judgments into individual legal issues and extracts, for each issue, a structured XML representation grounded in the IRAC framework and the legal syllogism. The pipeline targets a corpus of approximately $330{,}000$ first- and second-instance decisions of the Italian tax courts and is built around a capable yet cost-efficient general-purpose model (DeepSeek V3), a choice driven by the need to process several hundred thousand documents at a sustainable cost. To address the well-documented unreliability of large language models on legal citations, we couple the extraction step with an automatic hallucination-detection filter that compares the references produced by the model with those identified in the judgment text by a dedicated parser (Linkoln), normalised to standard identifiers (URN-NIR, ECLI, CELEX). We validate the pipeline on $50$ judgments annotated by two PhDs in tax law, computing inter-annotator agreement and LLM-vs-expert agreement on both issue extraction and legal citations, together with a stand-alone evaluation of the hallucination filter. To the best of our knowledge, this is the first issue-level, expert-validated structured extraction pipeline with hallucination control for Italian tax-court decisions, and it provides a concrete starting point for downstream applications such as issue-level retrieval, citation-network analysis, and the construction of large-scale datasets of legal reasoning.

Figures

Figures reproduced from arXiv: 2607.03325 by the authors.

Figure 1
Figure 1. Stylized visualization of the extraction and verification pipeline. [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Flowchart of the hallucination check for items in <legal_references>. [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

73 extracted references · 3 canonical work pages

  1. [1]

    Knowledge of the Law in the Big Data Age , volume=

    Improving public access to legislation through legal citations detection: the linkoln project at the Italian senate , author=. Knowledge of the Law in the Big Data Age , volume=. 2019 , publisher=

    2019

  2. [2]

    2024 , url =

    Banca Dati della Giurisprudenza Tributaria , author =. 2024 , url =

    2024

  3. [3]

    2026 , url =

    PyMuPDF Documentation , author =. 2026 , url =

    2026

  4. [4]

    2023 , howpublished =

    Gerwin Klein and Steve Rowe and Régis Décamps , title =. 2023 , howpublished =

    2023

  5. [5]

    Informatica e diritto , volume=

    URN-based Identification of Legal Acts: The Case of the Italian Senate , author=. Informatica e diritto , volume=. 2010 , note=

    2010

  6. [6]

    Artificial Intelligence and law , volume=

    A Bayesian model of legal syllogistic reasoning , author=. Artificial Intelligence and law , volume=. 2024 , publisher=

    2024

  7. [7]

    The Review of Socionetwork Strategies , volume=

    Legal information retrieval and entailment based on bm25, transformer and semantic thesaurus methods , author=. The Review of Socionetwork Strategies , volume=. 2022 , publisher=

    2022

  8. [8]

    European Conference on Information Retrieval , pages=

    PARM: A paragraph aggregation retrieval model for dense document-to-document retrieval , author=. European Conference on Information Retrieval , pages=. 2022 , organization=

    2022

Show all 73 references
  1. [9]

    ACM Computing Surveys , volume=

    A comprehensive survey on legal summarization: Challenges and future directions , author=. ACM Computing Surveys , volume=. 2025 , publisher=

    2025

  2. [10]

    Artificial Intelligence and Law , pages=

    A case study for automated attribute extraction from legal documents using large language models , author=. Artificial Intelligence and Law , pages=. 2024 , publisher=. doi:10.1007/s10506-024-09425-7 , note=

    2024 doi

  3. [11]

    Artificial Intelligence and Law , pages=

    Automating IRAC analysis in Malaysian contract law using a semi-structured knowledge base , author=. Artificial Intelligence and Law , pages=. 2025 , publisher=

    2025

  4. [12]

    Proceedings of the 31st international conference on computational linguistics , pages=

    Can large language models perform relation-based argument mining? , author=. Proceedings of the 31st international conference on computational linguistics , pages=. 2025 , doi=

    2025

  5. [13]

    Artificial Intelligence and Law , pages=

    Challenges and Opportunities for Visual Analytics in Jurisprudence , author=. Artificial Intelligence and Law , pages=. 2025 , publisher=

    2025

  6. [14]

    arXiv preprint arXiv:2508.03366 , year=

    A comparative study of neurosymbolic ai approaches to interpretable logical reasoning , author=. arXiv preprint arXiv:2508.03366 , year=. doi:10.48550/arXiv.2508.03366 , note=

    doi arXiv

  7. [15]

    Artificial Intelligence and Law , pages=

    Effectiveness in retrieving legal precedents: exploring text summarization and cutting-edge language models toward a cost-efficient approach , author=. Artificial Intelligence and Law , pages=. 2025 , publisher=

    2025

  8. [16]

    Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , pages=

    EUR-lex-sum: A multi-and cross-lingual dataset for long-form summarization in the legal domain , author=. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , pages=. 2022 , doi=

    2022

  9. [17]

    Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics , pages=

    Finding the law: Enhancing statutory article retrieval via graph neural networks , author=. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics , pages=. 2023 , doi=

    2023

  10. [18]

    arXiv preprint arXiv:2308.04416 , year=

    Legal summarisation through llms: The prodigit project , author=. arXiv preprint arXiv:2308.04416 , year=

    arXiv

  11. [19]

    Artificial Intelligence and Law , pages=

    Legal citation prediction with LLMs: a comparative evaluation of instruction tuning, retrieval, and jurisdiction-specific pre-training on the AusLaw citation benchmark , author=. Artificial Intelligence and Law , pages=. 2026 , publisher=

    2026

  12. [20]

    arXiv preprint arXiv:2212.01326 , year=

    Legal prompting: Teaching a language model to think like a lawyer , author=. arXiv preprint arXiv:2212.01326 , year=

    arXiv

  13. [21]

    Artificial Intelligence and Law , pages=

    LegItBART: a summarization model for Italian legal documents , author=. Artificial Intelligence and Law , pages=. 2025 , publisher=

    2025

  14. [22]

    Artificial Intelligence and Law , volume=

    Mining legal arguments in court decisions , author=. Artificial Intelligence and Law , volume=. 2024 , publisher=

    2024

  15. [23]

    Artificial Intelligence and Law , volume=

    A RDF-based graph to representing and searching parts of legal documents , author=. Artificial Intelligence and Law , volume=. 2024 , publisher=

    2024

  16. [24]

    Artificial Intelligence and Law , pages=

    Specialized or general AI? a comparative evaluation of LLMs' performance in legal tasks , author=. Artificial Intelligence and Law , pages=. 2025 , publisher=

    2025

  17. [25]

    Frontiers in Artificial Intelligence and Applications , pages=

    Automated Extraction of Judicial Interpretative Formulas in EU Case Law on VAT , author=. Frontiers in Artificial Intelligence and Applications , pages=. 2025 , note=

    2025

  18. [26]

    Proceedings of the Nineteenth International Conference on Artificial Intelligence and Law , pages=

    Legal holding extraction from italian case documents using italian-legal-bert text summarization , author=. Proceedings of the Nineteenth International Conference on Artificial Intelligence and Law , pages=. 2023 , doi=

    2023

  19. [27]

    Proceedings of the Nineteenth International Conference on Artificial Intelligence and Law , pages=

    Argumentation structure prediction in CJEU decisions on fiscal state aid , author=. Proceedings of the Nineteenth International Conference on Artificial Intelligence and Law , pages=. 2023 , doi=

    2023

  20. [28]

    Proceedings of the Twentieth International Conference on Artificial Intelligence and Law , pages=

    Automated Mapping of Legal Criteria to the Texts of Adjudicatory Decisions Using LLMs , author=. Proceedings of the Twentieth International Conference on Artificial Intelligence and Law , pages=. 2025 , doi=

    2025

  21. [29]

    Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=

    From text to structure: Using large language models to support the development of legal expert systems , author=. Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=. 2023 , organization=

    2023

  22. [30]

    37th Annual Conference on Legal Knowledge and Information Systems, JURIX 2024 , pages=

    Automated semantic annotation pipeline for brazilian judicial decisions , author=. 37th Annual Conference on Legal Knowledge and Information Systems, JURIX 2024 , pages=. 2024 , organization=

    2024

  23. [31]

    Frontiers in Artificial Intelligence and Applications , volume=

    Unravelling the ECHR: components of legal case analysis , author=. Frontiers in Artificial Intelligence and Applications , volume=. 2024 , publisher=

    2024

  24. [32]

    arXiv preprint arXiv:2309.15016 , year=

    Question-Answering Approach to Evaluating Legal Summaries , author=. arXiv preprint arXiv:2309.15016 , year=

    arXiv

  25. [33]

    JURIX 2025-38th International Conference on Legal Knowledge and Information Systems , pages=

    Does chatgpt understand the law? a case study on road homicide in italy , author=. JURIX 2025-38th International Conference on Legal Knowledge and Information Systems , pages=. 2025 , doi=

    2025

  26. [34]

    Legal Knowledge and Information Systems: JURIX 2024: The Thirty-seventh Annual Conference, Brno, Czech Republic, 11-13 December 2024 , pages=

    Leveraging llm for identification and extraction of normative statements , author=. Legal Knowledge and Information Systems: JURIX 2024: The Thirty-seventh Annual Conference, Brno, Czech Republic, 11-13 December 2024 , pages=. 2024 , organization=

    2024

  27. [35]

    arXiv preprint arXiv:2310.17413 , year=

    Harnessing GPT-3.5-turbo for rhetorical role prediction in legal cases , author=. arXiv preprint arXiv:2310.17413 , year=. doi:10.48550/arXiv.2310.17413 , note=

    doi arXiv

  28. [36]

    Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=

    Legal Text Segmentation Through Breakpoint Detection , author=. Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=. 2023 , organization=

    2023

  29. [37]

    Proceedings of the Nineteenth International Conference on Artificial Intelligence and Law , pages=

    Unlocking practical applications in legal domain: Evaluation of gpt for zero-shot semantic annotation of legal texts , author=. Proceedings of the Nineteenth International Conference on Artificial Intelligence and Law , pages=. 2023 , doi=

    2023

  30. [38]

    Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=

    Can GPT alleviate the burden of annotation? , author=. Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=. 2023 , organization=

    2023

  31. [39]

    Frontiers in Artificial Intelligence and Applications , volume=

    An Expert-Validated LLM Framework for Transforming Legal Procurement Texts into Actionable Data , author=. Frontiers in Artificial Intelligence and Applications , volume=. 2025 , publisher=

    2025

  32. [40]

    Frontiers in Artificial Intelligence and Applications , year=

    Context-Aware Citation Networks: A Human--AI Dataset, Analysis, and Tool , author=. Frontiers in Artificial Intelligence and Applications , year=

  33. [41]

    Legal Knowledge and Information Systems: JURIX 2025 , pages=

    A Self-Supervised Method for Legal Reference Resolution , author=. Legal Knowledge and Information Systems: JURIX 2025 , pages=. 2025 , publisher=

    2025

  34. [42]

    Legal Knowledge and Information Systems , pages=

    From Court Decisions to Guiding Principles: Advancing Complex Legal Summarization with LLMs , author=. Legal Knowledge and Information Systems , pages=. 2025 , publisher=

    2025

  35. [43]

    Legal Knowledge and Information Systems: JURIX 2025 , pages=

    LR2: A Legal RAG for Legal Reasoning over Cases , author=. Legal Knowledge and Information Systems: JURIX 2025 , pages=. 2025 , publisher=

    2025

  36. [44]

    JURIX , pages=

    ConsRAG: Minimize LLM Hallucinations in the Legal Domain , author=. JURIX , pages=. 2024 , note=

    2024

  37. [45]

    Proceedings of the Twentieth International Conference on Artificial Intelligence and Law , pages=

    Identifying Legal Holdings with LLMs: A Systematic Study of Performance, Scale, and Memorization , author=. Proceedings of the Twentieth International Conference on Artificial Intelligence and Law , pages=. 2025 , doi=

    2025

  38. [46]

    Proceedings of the Twentieth International Conference on Artificial Intelligence and Law , pages=

    Llms provide unstable answers to legal questions , author=. Proceedings of the Twentieth International Conference on Artificial Intelligence and Law , pages=. 2025 , doi=

    2025

  39. [47]

    Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=

    Benchmarking abstractive models for italian legal news summarization , author=. Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=. 2023 , organization=

    2023

  40. [48]

    arXiv preprint arXiv:2601.05609 , year=

    Data Augmented Pipeline for Legal Information Extraction and Reasoning , author=. arXiv preprint arXiv:2601.05609 , year=. doi:10.48550/arXiv.2601.05609 , note=

    doi

  41. [49]

    Proceedings of the nineteenth international conference on artificial intelligence and law , pages=

    Legal syllogism prompting: Teaching large language models for legal judgment prediction , author=. Proceedings of the nineteenth international conference on artificial intelligence and law , pages=. 2023 , doi=

    2023

  42. [50]

    Proceedings of AI for Access to Justice Workshop at Jurix 2024 , pages=

    Using chatgpt to extract principles of law for the sake of prediction: an exploration conducted on italian judgments concerning lgbt (qia+) rights , author=. Proceedings of AI for Access to Justice Workshop at Jurix 2024 , pages=. 2024 , note=

    2024

  43. [51]

    arXiv preprint arXiv:2005.05257 , year=

    A dataset for statutory reasoning in tax law entailment and question answering , author=. arXiv preprint arXiv:2005.05257 , year=

    2005 arXiv

  44. [52]

    The Review of Socionetwork Strategies , volume=

    Overview and discussion of the competition on legal information extraction/entailment (COLIEE) 2021 , author=. The Review of Socionetwork Strategies , volume=. 2022 , publisher=

    2021

  45. [53]

    , title =

    Gwet, Kilem L. , title =. 2014 , publisher =

    2014

  46. [54]

    and Cicchetti, Domenic V

    Feinstein, Alvan R. and Cicchetti, Domenic V. , title =. Journal of Clinical Epidemiology , volume =. 1990 , doi =

    1990

  47. [55]

    , title =

    Wongpakaran, Nahathai and Wongpakaran, Tinakon and Wedding, Danny and Gwet, Kilem L. , title =. BMC Medical Research Methodology , number =. 2013 , publisher =

    2013

  48. [56]

    Proceedings of the 9th international conference on Artificial intelligence and law , pages=

    Automated extraction of normative references in legal texts , author=. Proceedings of the 9th international conference on Artificial intelligence and law , pages=. 2003 , doi=

    2003

  49. [57]

    Frontiers in Physics , volume=

    Measuring law over time: a network analytical framework with an application to statutes and regulations in the United States and Germany , author=. Frontiers in Physics , volume=. 2021 , publisher=

    2021

  50. [58]

    Frontiers in Physics , volume=

    Analysing eu treaty-making and litigation with network analysis and natural language processing , author=. Frontiers in Physics , volume=. 2021 , publisher=

    2021

  51. [59]

    2012 , publisher=

    Linee guida per la marcatura dei documenti normativi secondo gli standard Normeinrete , author=. 2012 , publisher=

    2012

  52. [60]

    Proceedings of the Natural Legal Language Processing Workshop 2023 , pages=

    Connecting symbolic statutory reasoning with legal information extraction , author=. Proceedings of the Natural Legal Language Processing Workshop 2023 , pages=. 2023 , doi=

    2023

  53. [61]

    ai: A global open-source legal language model , author=

    Openjustice. ai: A global open-source legal language model , author=. Legal Knowledge and Information Systems: JURIX 2023: The Thirty-sixth Annual Conference, Maastricht, the Netherlands, 18--20 December 2023 , pages=. 2023 , organization=

    2023

  54. [62]

    Handbook of Formal Argumentation , pages=

    Computational models of legal argument , author=. Handbook of Formal Argumentation , pages=. 2024 , publisher=

    2024

  55. [63]

    2017 , publisher=

    Artificial intelligence and legal analytics: new tools for law practice in the digital age , author=. 2017 , publisher=

    2017

  56. [64]

    1994 , publisher=

    Legal reasoning and legal theory , author=. 1994 , publisher=

    1994

  57. [65]

    Journal of Legal Analysis , volume=

    Large legal fictions: Profiling legal hallucinations in large language models , author=. Journal of Legal Analysis , volume=. 2024 , publisher=

    2024

  58. [66]

    Computer networks and ISDN systems , volume=

    The anatomy of a large-scale hypertextual web search engine , author=. Computer networks and ISDN systems , volume=. 1998 , publisher=

    1998

  59. [67]

    Compliance for Artificial Intelligence Systems , series=

    Evaluating AI for Law: Bridging the Gap with Open-Source Solutions , author=. Compliance for Artificial Intelligence Systems , series=. 2026 , publisher=. doi:10.1007/978-3-032-12795-2_5 , note=

    2026 doi

  60. [68]

    arXiv preprint arXiv:2412.19437 , year=

    Deepseek-v3 technical report , author=. arXiv preprint arXiv:2412.19437 , year=

    arXiv

  61. [69]

    La graduazione delle questioni di merito

    Villa, Alberto , volume=. La graduazione delle questioni di merito. Ammissibilit. 2023 , publisher=

    2023

  62. [70]

    Principi , author=

    Manuale di diritto processuale civile. Principi , author=. 2007 , publisher=

    2007

  63. [71]

    digesto discipline privatistiche , year=

    Accertamento giudiziale , author=. digesto discipline privatistiche , year=

  64. [72]

    CEUR Workshop Proceedings , volume=

    Explaining legal decisions using IRAC , author=. CEUR Workshop Proceedings , volume=. 2020 , organization=

    2020

  65. [73]

    Comparative Law: Unraveling Global Legal Systems , pages=

    Aristotle is Long Dead But His Wisdom Rules Us from the Grave: A Comparison of the Application of Logic in Legal Reasoning in Common Law and Civil Law Systems , author=. Comparative Law: Unraveling Global Legal Systems , pages=. 2024 , publisher=

    2024

Pith tools

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