Pith. sign in

REVIEW 1 major objections 6 minor 59 references

The Judge Variable: Challenging Judge-Agnostic Legal Judgment Prediction

T0 review · 1 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Judge identity measurably shapes French child custody outcomes: specialist models trained on a single judge's rulings outperform judge-agnostic generalists.

desk verdict Good new dataset, but the headline claim overreaches: the specialist-versus-generalist comparison can't separate judge identity from court and case mix, and the abstract's consistency claim doesn't hold up in the paper's own tables. read the letter →

arxiv 2507.13732 v1 pith:UANU46ES submitted 2025-07-18 cs.CL cs.LG

Guillaume Zambrano This is my paper
classification cs.CLcs.LG
keywords legaljudgmentpredictionjudgeidentityjudicialdiscretionrealismchildcustodyFrenchappellatecourtsmachinelearningLLMfeatureextraction
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 asks whether the identity of the judge hearing a case changes its outcome, a core question in the legal realism–formalism debate. Using 18,937 child-custody rulings from French appellate courts, it trains specialist machine-learning models on the past decisions of individual judges and a generalist model on a pooled, judge-agnostic dataset. The specialist models reach F1 scores as high as 92.85%, compared with 82.63% for the generalist, which was trained on 20 to 100 times more data, and specialist models lose accuracy when applied to other judges' cases. The paper interprets these results as evidence for legal realism: judges show stable, individually distinguishable decision patterns, so modeling the judge as a variable improves prediction.

What carries the argument

The 'judge variable' is operationalized through a two-phase pipeline. First, a large language model (Llama 3.3 70B) extracts structured categorical features — parental petitions, child preferences, behavioral and practical factors — and the case outcome from ruling texts, with judge identities pseudonymized to comply with French law. Second, three classic classifiers (Random Forest, XGBoost, Support Vector Classification) are trained either on judge-specific buckets ('specialist models,' for judges with more than 300 rulings) or on the pooled residual group of judges with fewer than 300 rulings ('generalist model'). The experimental design then compares in-domain F1 (specialist tested on its own judge) against cross-domain F1 (same specialist tested on other judges), with the generalist as the baseline.

What would settle it

If cases are randomly assigned to judges within the same court, and judge-specific models still outperform the court-level generalist on held-out cases, the judge effect is real. If the specialist advantage disappears once court and case-mix variables are controlled — for example, when the generalist is trained only on the same court's cases — the apparent judge effect is actually a venue or case-allocation effect.

Watch

Extended reading notes

Core claim

The central claim is that judicial identity is a measurable, predictive variable in legal outcomes, not just noise to be averaged away. In the paper's experiments, models trained only on a single judge's past rulings consistently beat the judge-agnostic generalist on that judge's held-out cases, and their performance drops on other judges' cases. The paper reads the drop as evidence that decision-making patterns are judge-specific: the same case features carry different weight for different judges, and the specialist models capture stable individual 'signatures' that do not transfer. This is presented as a direct challenge to the judge-agnostic assumption underlying most legal judgment prediction, where judges are treated as interchangeable.

Load-bearing premise

The specialist model's advantage is interpreted as reflecting each judge's individual decision style, but it could instead reflect systematic differences in the kinds of cases that flow to high-volume judges, who sit in larger courts, versus the residual pool of lower-volume judges.

Editorial extensions

If this is right

  • Legal judgment prediction systems should treat judge identity as a feature, not a nuisance, when past decisions are available.
  • Judicial discretion becomes quantifiable: the gap between specialist and generalist F1 gives a per-judge measure of how far that judge's decisions deviate from the pooled norm.
  • Because specialist models capture patterns that do not transfer to other judges, a single 'one-size-fits-all' prediction model is structurally limited in high-discretion legal domains.
  • Feature-importance variation across judges indicates that judges weight petitions, behavioral evidence, and practical considerations differently, with direct implications for how courts are audited.

Reading between the lines

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

  • The comparison conflates judge with venue: high-volume judges cluster in large courts, so the specialist advantage could partly reflect local court culture or case demographics rather than an individual judge's style; a same-court, same-judge design would separate these.
  • If the judge effect survives venue controls, the results imply that court transparency tools could use specialist models to flag outlier judges or give litigants realistic forecasts that include the assigned judge's tendencies.
  • A testable extension is a within-judge stability check: specialist models trained on a judge's early rulings should predict that judge's later rulings as well as models trained on temporally adjacent rulings, which would distinguish a stable personal style from evolving court practice.
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

1 major / 6 minor

Summary. The paper tests whether individual judges' decision-making patterns measurably influence child-custody outcomes in French appellate courts, using a legal-realism versus legal-formalism framing. The authors build a dataset of 18,937 rulings from 10,306 cases, extract structured features with Llama 3.3, and train specialist models on each of 13 high-volume judges and a generalist model on all remaining judges (the 'Generic Bucket'). They report in-domain and cross-domain macro-F1 scores for Random Forest, XGBoost, and SVC. The headline result is that specialist models often outperform the generalist on their own judges' test sets and degrade on other judges' cases, interpreted as evidence that judicial identity is a stable, predictive variable.

Significance. If the central claim were supported, the paper would be a valuable contribution to legal judgment prediction and to the empirical legal-realism literature. The dataset is substantial for the domain, the research question is clearly posed, and the paper engages seriously with prior work on judge effects (Medvedeva, Wang, Barale). The pipeline is transparent and the authors commit to releasing data and code. However, the current experimental design does not identify a judge effect as opposed to a venue or case-mix effect, and the statistical evidence is fragile given the small specialist test sets.

major comments (1)
  1. [§5 and Table 2] The cross-domain generalization results (Tables 20-22) are interpreted as evidence of 'idiosyncratic rules' and 'judicial distinctiveness,' but the observed degradation of specialist models on other judges' data is equally consistent with overfitting to small samples and to court-specific textual regularities. The generalist model itself is highly unstable across courts (e.g., the Gen row in Table 22 ranges from 65.83 to 88.59), which indicates strong venue effects. The paper does not test whether the specialist's in-domain advantage persists after controlling for court or case mix, so the cross-domain evidence does not disambiguate the judge-effect hypothesis from the venue-effect hypothesis.
minor comments (6)
  1. [Abstract] The statement that specialist models 'consistently achieve higher predictive accuracy' is an overstatement: for XGBoost and SVC, only 9 of 13 specialists exceed the generalist's F1 on the same judge's test set, and several specialists fall well below. Suggest reporting the full distribution and the median specialist-generalist difference.
  2. [§3.b] The paper does not provide the prompt used for fact extraction, nor any evaluation of the accuracy of the extracted features against a gold standard. Given that these features are the core input to all models, a validation study is necessary.
  3. [§5] The resampling ratios for run1 and run2 (34%/40% undersampling, 33%/30% oversampling) appear arbitrary, and the paper does not explain why these specific values were chosen or whether results are sensitive to them.
  4. [Tables 11-18] There are inconsistencies in test-set sizes across tables for the same bucket and run (e.g., Dacrons has n_test=157 in XGB and SVC run1 but n_test=127 in RF run1). The paper should explain whether these are different random splits or data exclusions.
  5. [§2] The claim that 'the large scale of our dataset provides a robust statistical buffer against the impact of missing entries' is not convincing for the specialist buckets, which contain only a few hundred rulings and are the basis of the main comparison.
  6. [Throughout] The manuscript contains frequent typographical and formatting errors (e.g., 'splitted', 'sampls', 'afer_appeal_review' in Figure 1, 'V olume' in references). A careful proofreading pass is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the specialist/generalist comparison is an empirical evaluation, and the leakage/confound concerns are validity threats rather than reductions.

full rationale

The paper's derivation chain is an empirical comparison rather than a formal derivation: LLM-based feature extraction, per-judge specialist training buckets, a residual judge-agnostic generalist bucket, and then evaluation on held-out test sets and cross-domain matrices. No step equates a predicted quantity to an input by construction. The specialist advantage is measured on each judge's own held-out rulings, which is a standard supervised evaluation, and the cross-domain matrices compare models on other judges' held-out data; the interpretation that the gap reflects a judicial signature is an empirical claim, not a tautology. The concern that Llama 3.3 extracts case-description features from full judgment texts that also contain the outcome is a label-leakage threat, but the paper does not state that the features are defined from the outcome and provides no equation showing the features reduce to the label, so under the hard rules this is not demonstrated circularity. The court/case-mix confound (high-volume judges concentrated in large courts) threatens the causal reading of the specialist advantage but does not make the prediction itself equivalent to its inputs. The references are external, with no load-bearing self-citations or imported uniqueness claims. The derivation is therefore self-contained and not circular.

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

The central claim rests on two unverified assumptions: the outcome-independence of LLM-extracted features, and the interpretation of the specialist advantage as a judge effect rather than a case-mix effect. The manual resampling ratios and the 300-ruling threshold are hand-chosen parameters that affect all reported scores.

free parameters (3)
  • judge_volume_threshold = 300 rulings
    Judges with more than 300 rulings are classified as specialists; this hand-chosen cutoff defines the specialist set and therefore the entire comparison.
  • resampling_ratios_run1 = majority undersampled to 34%; minorities oversampled to 33% each
    Class-balancing ratios are set manually for run 1 and affect all reported F1 scores.
  • resampling_ratios_run2 = majority undersampled to 40%; minorities oversampled to 30% each
    Second manual balancing configuration; results are slightly different from run 1.
assumptions (3)
  • domain assumption LLM-extracted features are outcome-independent
    Features are extracted from the same judgment text as the outcome; only outcome extraction is validated against a 500-ruling gold standard, not feature extraction (Section 3c).
  • domain assumption The specialist/generalist gap reflects judge-specific decision policies, not case mix or court effects
    The 13 specialists are high-volume judges from large courts; the generalist pool is the residual. No adjustment is made for court, city, or case characteristics (Sections 2 and 5).
  • domain assumption Case allocation to judges is not correlated with judge-specific decision styles in a way that biases the comparison
    The design assumes differences in performance are due to the judge, but if clerks assign cases based on judge reputation or workload, the comparison is biased.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Judge Variable: Challenging Judge-Agnostic Legal Judgment Prediction." pith.science (2026). https://pith.science/paper/UANU46ES

@misc{pith2026250713732,
  author       = {Pith},
  title        = {Pith review of: The Judge Variable: Challenging Judge-Agnostic Legal Judgment Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UANU46ES}},
  note         = {Machine review of arXiv:2507.13732}
}
read the original abstract

This study examines the role of human judges in legal decision-making by using machine learning to predict child physical custody outcomes in French appellate courts. Building on the legal realism-formalism debate, we test whether individual judges' decision-making patterns significantly influence case outcomes, challenging the assumption that judges are neutral variables that apply the law uniformly. To ensure compliance with French privacy laws, we implement a strict pseudonymization process. Our analysis uses 18,937 living arrangements rulings extracted from 10,306 cases. We compare models trained on individual judges' past rulings (specialist models) with a judge-agnostic model trained on aggregated data (generalist models). The prediction pipeline is a hybrid approach combining large language models (LLMs) for structured feature extraction and ML models for outcome prediction (RF, XGB and SVC). Our results show that specialist models consistently achieve higher predictive accuracy than the general model, with top-performing models reaching F1 scores as high as 92.85%, compared to the generalist model's 82.63% trained on 20x to 100x more samples. Specialist models capture stable individual patterns that are not transferable to other judges. In-Domain and Cross-Domain validity tests provide empirical support for legal realism, demonstrating that judicial identity plays a measurable role in legal outcomes. All data and code used will be made available.

Figures

Figures reproduced from arXiv: 2507.13732 by the authors.

Figure 1
Figure 1. Case Outcome Extraction Prompt. Prompt#1 [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

59 extracted references · 53 canonical work pages

  1. [2]

    Un tiers de divorces en moins en 15 ans

    PAILLET, Guillaume, "Un tiers de divorces en moins en 15 ans", Infos Rapides Justice, Numéro 19, Service de la statistique, des études et de la recherche (SSER), 28 novembre 2024: <https://www.justice.gouv.fr/sites/default/files/2024-11/Infos_rapides_justice_n19_1.pdf>

    work page 2024

  2. [3]

    Fairness perceptions of the postdivorce division of childcare and child‐related expenses

    KOSTER, Tara and POORTMAN, Anne‐Rigt. Fairness perceptions of the postdivorce division of childcare and child‐related expenses. Journal of Marriage and Family, 2025, vol. 87, no 2, p. 751-771

    work page 2025

  3. [4]

    MACKAAY , E., & ROBILLARD, P. (1974). Predicting judicial decisions: The nearest neighbour rule and visual representation of case patterns. University of Montreal, Law and Economics Research Paper

    work page 1974

  4. [5]

    ZELEZNIKOW, J., & HUNTER, D. (1994). Building intelligent legal information systems: Representation and reasoning in law. Kluwer Law and Taxation Publishers

    work page 1994

  5. [6]

    BRÜNINGHAUS, S., & ASHLEY , K. D. (2003). Combining case-based and model-based reasoning for predicting the outcome of legal cases. International Conference on Case-Based Reasoning

    work page 2003

  6. [7]

    RINGGER, Eric K., et al. (2008). Assessing the costs of machine-assisted corpus annotation through a user study. In Proceedings of the 6th International Conference on Language Resources and Evaluation (LREC 2008)

    work page 2008

  7. [8]

    LexGLUE: A benchmark dataset for legal language understanding in English

    CHALKIDIS, Ilias, JANA, Abhik, HARTUNG, Dirk, et al. LexGLUE: A benchmark dataset for legal language understanding in English. arXiv preprint arXiv:2110.00976, 2021

    arXiv 2021

  8. [9]

    How Does NLP Benefit Legal System: A Summary of Legal Artificial Intelligence

    ZHONG, Haoxi, XIAO, Chaojun, TU, Cunchao, et al. How does NLP benefit legal system: A summary of legal artificial intelligence. arXiv preprint arXiv:2004.12158, 2020

    work page Pith review arXiv 2004

Show all 59 references
  1. [10]

    Surprisingly Simple: Large Language Models are Zero-Shot Feature Extractors for Tabular and Text Data

    SHI, Zhiyi, KIM, Junsik, JEONG, Davin, et al. Surprisingly Simple: Large Language Models are Zero-Shot Feature Extractors for Tabular and Text Data. Under review as a conference paper at ICLR 2025

    2025

  2. [11]

    Leveraging LLM for Identification and Extraction of Normative Statements

    ZIN, May Myo, SA TOH, Ken, et BORGES, Georg. Leveraging LLM for Identification and Extraction of Normative Statements. In : Legal Knowledge and Information Systems. IOS Press, 2024. p. 215-225

    2024

  3. [13]

    Information extraction from employment tribunal judgments using a large language model

    RIBEIRO DE FARIA, Joana, XIE, Huiyuan, et STEFFEK, Felix. Information extraction from employment tribunal judgments using a large language model. Artificial Intelligence and Law, 2025, p. 1-22

    2025

  4. [14]

    Training language models to follow instructions with human feedback

    OUYANG, Long, WU, Jeffrey, JIANG, Xu, et al. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 2022, vol. 35, p. 27730-27744

    2022

  5. [15]

    A minimalist prompt for zero-shot policy learning

    SONG, Meng, WANG, Xuezhi, BIRADAR, Tanay, et al. A minimalist prompt for zero-shot policy learning. arXiv preprint arXiv:2405.06063, 2024

    2024 arXiv

  6. [16]

    Think inside the json: Reinforcement strategy for strict llm schema adherence

    AGARWAL, Bhavik, JOSHI, Ishan, et ROJKOV A, Viktoria. Think inside the json: Reinforcement strategy for strict llm schema adherence. arXiv preprint arXiv:2502.14905, 2025

    2025 arXiv

  7. [17]

    A brief history of the changing roles of case prediction in AI and law

    ASHLEY , Kevin D. A brief history of the changing roles of case prediction in AI and law. Law Context: A Socio-Legal J., 2019, vol. 36, p. 93

    2019

  8. [18]

    A systematic literature review of machine learning methods in predicting court decisions

    ROSILI, Nur Aqilah Khadijah, ZAKARIA, Noor Hidayah, HASSAN, Rohayanti, et al. A systematic literature review of machine learning methods in predicting court decisions. IAES International Journal of Artificial Intelligence, 2021, vol. 10, n4, p. 1091

    2021

  9. [20]

    Automatic judgement forecasting for pending applications of the European Court of Human Rights

    MEDVEDEV A, Masha, ÜSTÜN, Ahmet, XU, Xiao, et al. Automatic judgement forecasting for pending applications of the European Court of Human Rights. In : Proceedings of the Fifth Workshop on Automatec Semantic Analysis of Information in Legal Text (ASAIL 2021). CEUR Workshop Proc...

    2021

  10. [21]

    Inferring the Winning Party in the Supreme Court from the Pattern of Questioning at Oral Argument

    EPSTEIN, L., W. M. LANDES, and R. A. POSNER. “Inferring the Winning Party in the Supreme Court from the Pattern of Questioning at Oral Argument.” The Journal of Legal Studies. 2010. 39(2):433–467

    2010

  11. [24]

    The supreme court forecasting project: legal and political science approaches to predicting supreme court decisionmaking

    RUGER, Theodore W., KIM, Pauline T., MARTIN, Andrew D., et al. The supreme court forecasting project: legal and political science approaches to predicting supreme court decisionmaking. Colum. L. Rev., 2004, vol. 104, p. 1150. 21

    2004

  12. [25]

    Predicting the behavior of the supreme court of the united states: A general approach

    KATZ, Daniel Martin, BOMMARITO II, Michael J., et BLACKMAN, Josh. Predicting the behavior of the supreme court of the united states: A general approach. arXiv preprint arXiv:1407.6333, 2014

    2014 arXiv

  13. [26]

    Using Modern Neural Networks to Predict the Decisions of Supreme Court of the United States with State-of-the-Art Accuracy

    SHARMA, R.D., MITTAL, S., TRIPA THI, S., ACHARYA, S. Using Modern Neural Networks to Predict the Decisions of Supreme Court of the United States with State-of-the-Art Accuracy. In: Arik, S., Huang, T., Lai, W., Liu, Q. (eds) Neural Information Processing. ICONIP 2015. Lecture ...

    2015

  14. [27]

    A general approach for predicting the behavior of the Supreme Court of the United States

    KATZ, Daniel Martin, BOMMARITO II, Michael J., et BLACKMAN, Josh. A general approach for predicting the behavior of the Supreme Court of the United States. PloS one, 2017, vol. 12, no 4, p. e0174698

    2017

  15. [28]

    Crowdsourcing accurately and robustly predicts Supreme Court decisions

    KATZ, Daniel Martin, BOMMARITO II, Michael James, et BLACKMAN, Josh. Crowdsourcing accurately and robustly predicts Supreme Court decisions. arXiv preprint arXiv:1712.03846, 2017

    2017 arXiv

  16. [29]

    Improving supreme court forecasting using boosted decision trees

    KAUFMAN, Aaron Russell, KRAFT, Peter, et SEN, Maya. Improving supreme court forecasting using boosted decision trees. Political Analysis, 2019, vol. 27, no 3, p. 381-387

    2019

  17. [30]

    Efficient prediction of court judgments using an LSTM+ CNN neural network model with an optimal feature set

    ALGHAZZAWI, Daniyal, BAMASAG, Omaimah, ALBESHRI, Aiiad, et al. Efficient prediction of court judgments using an LSTM+ CNN neural network model with an optimal feature set. Mathematics, 2022, vol. 10, no 5, p. 683

    2022

  18. [31]

    DINU, and Josef V AN GENABITH

    SULEA, Octavia-Maria, Marcos ZAMPIERI, Shervin MALMASI, Mihaela VELA, Liviu P. DINU, and Josef V AN GENABITH. Exploring the use of text classification in the legal domain. In Proceedings of ASAIL@ICAIL, volume 2143, 2017

    2017

  19. [32]

    DINU, and Josef V AN GENABITH

    SULEA, Octavia-Maria, Marcos ZAMPIERI, Shervin MALMASI, Mihaela VELA, Liviu P. DINU, and Josef V AN GENABITH. Predicting the law area and decisions of French supreme court cases. In Proceedings of RANLP, pages 716–722, 2017

    2017

  20. [33]

    Legal judgment prediction for Canadian appeal cases

    ALMUSLIM, Intisar et INKPEN, Diana. Legal judgment prediction for Canadian appeal cases. In : 2022 7th International Conference on Data Science and Machine Learning Applications (CDMA). IEEE, 2022. p. 163- 168

    2022

  21. [34]

    Legal judgement prediction for UK courts

    STRICKSON, Benjamin et DE LA IGLESIA, Beatriz. Legal judgement prediction for UK courts. In : Proceedings of the 3rd International Conference on Information Science and Systems. 2020. p. 204-209

    2020

  22. [35]

    Swiss-judgment-prediction: A multilingual legal judgment prediction benchmark

    NIKLAUS, Joel; CHALKIDIS, Ilias; STÜRMER, Matthias. Swiss-judgment-prediction: A multilingual legal judgment prediction benchmark. arXiv preprint arXiv:2110.00806, 2021

    2021 arXiv

  23. [36]

    ILDC for CJPE: Indian Legal Documents Corpus for Court Judgment Prediction and Explanation

    MALIK, Vijit, SANJAY , Rishabh, NIGAM, Shubham Kumar, et al. ILDC for CJPE: Indian Legal Documents Corpus for Court Judgment Prediction and Explanation. In : Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Join...

    2021

  24. [37]

    Fact-based court judgment prediction

    NIGAM, Shubham Kumar; DEROY , Aniket. Fact-based court judgment prediction. En Proceedings of the 15th Annual Meeting of the Forum for Information Retrieval Evaluation. 2023. p. 78-82

    2023

  25. [38]

    Predicting Brazilian court decisions

    LAGE-FREITAS, André, ALLENDE-CID, Héctor, SANTANA, Orivaldo, et al. Predicting Brazilian court decisions. PeerJ Computer Science, 2022, vol. 8, p. e904

    2022

  26. [39]

    Predicting judicial decisions of the European Court of Human Rights: A natural language processing perspective

    ALETRAS, Nikolaos, TSARAPATSANIS, Dimitrios, PREOŢIUC-PIETRO, Daniel, et al. Predicting judicial decisions of the European Court of Human Rights: A natural language processing perspective. PeerJ computer science, 2016, vol. 2, p. e93

    2016

  27. [40]

    Predicting judicial decisions: A statistically rigorous approach and a new ensemble classifier

    VISENTIN, Andrea, NARDOTTO, Alessia, et O’SULLIV AN, Barry. Predicting judicial decisions: A statistically rigorous approach and a new ensemble classifier. In : 2019 IEEE 31st International Conference on Tools with Artificial Intelligence (ICTAI). IEEE, 2019. p. 1820-1824

    2019

  28. [41]

    Judicial decisions of the European Court of Human Rights: Looking into the crystal ball

    MEDVEDEV A, Masha, VOLS, Michel, et WIELING, Martijn. Judicial decisions of the European Court of Human Rights: Looking into the crystal ball. In : Proceedings of the conference on empirical legal studies

  29. [43]

    Neural Legal Judgment Prediction in English

    CHALKIDIS, Ilias, ANDROUTSOPOULOS, Ion, et ALETRAS, Nikolaos. Neural Legal Judgment Prediction in English. In : Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics, 2019

    2019

  30. [44]

    XU, Shanshan, SANTOSH, T. Y . S., ICHIM, Oana, et al. Through the lens of split vote: Exploring disagreement, difficulty and calibration in legal case outcome classification. arXiv preprint arXiv:2402.07214, 2024

    2024 arXiv

  31. [45]

    PILOT: Legal Case Outcome Prediction with Case Law

    CAO, Lang, WANG, Zifeng, XIAO, Cao, et al. PILOT: Legal Case Outcome Prediction with Case Law. In : Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (V olume 1: Long Papers). 2024. p. 609-621

    2024

  32. [46]

    Cail2018: A large-scale legal dataset for judgment prediction

    XIAO, Chaojun, ZHONG, Haoxi, GUO, Zhipeng, et al. Cail2018: A large-scale legal dataset for judgment prediction. arXiv preprint arXiv:1807.02478, 2018

    2018 arXiv

  33. [48]

    Learning to Predict Charges for Criminal Cases with Legal Basis

    LUO, Bingfeng, FENG, Yansong, XU, Jianbo, et al. Learning to Predict Charges for Criminal Cases with Legal Basis. In : Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 2017

    2017

  34. [49]

    Legal judgment prediction via topological learning

    ZHONG, Haoxi, GUO, Zhipeng, TU, Cunchao, et al. Legal judgment prediction via topological learning. In : Proceedings of the 2018 conference on empirical methods in natural language processing. 2018. p. 3540-3549

    2018

  35. [50]

    Few-shot charge prediction with discriminative legal attributes

    HU, Zikun, LI, Xiang, TU, Cunchao, et al. Few-shot charge prediction with discriminative legal attributes. In : Proceedings of the 27th international conference on computational linguistics. 2018. p. 487-498. 22

    2018

  36. [51]

    An external knowledge enhanced multi-label charge prediction approach with label number learning

    WEI, Duan et LIN, Li. An external knowledge enhanced multi-label charge prediction approach with label number learning. arXiv preprint arXiv:1907.02205, 2019

    1907 arXiv

  37. [52]

    Legal judgment prediction via multi-perspective bi- feedback network

    YANG, Wenmian, JIA, Weijia, ZHOU, Xiaojie, et al. Legal judgment prediction via multi-perspective bi- feedback network. arXiv preprint arXiv:1905.03969, 2019

    1905 arXiv

  38. [53]

    Distinguish Confusing Law Articles for Legal Judgment Prediction

    XU, Nuo, WANG, Pinghui, CHEN, Long, et al. Distinguish Confusing Law Articles for Legal Judgment Prediction. In : Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics

  39. [54]

    Neurjudge: A circumstance-aware neural framework for legal judgment prediction

    YUE, Linan, LIU, Qi, JIN, Binbin, et al. Neurjudge: A circumstance-aware neural framework for legal judgment prediction. In : Proceedings of the 44th international ACM SIGIR conference on research and development in information retrieval. 2021. p. 973-982

    2021

  40. [55]

    DONG Qian and Shuzi NIU. 2021. Legal Judgment Prediction via Relational Learning. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR '21). Association for Computing Machinery, New York, NY , USA, 983–992. h...

    2021

  41. [56]

    Ml-ljp: Multi-law aware legal judgment prediction

    LIU, Yifei, WU, Yiquan, ZHANG, Yating, et al. Ml-ljp: Multi-law aware legal judgment prediction. In : Proceedings of the 46th international ACM SIGIR conference on research and development in information retrieval. 2023. p. 1023-1034

    2023

  42. [57]

    RLJP: Legal Judgment Prediction via First-Order Logic Rule-enhanced with Large Language Models

    ZHANG, Yue, TIAN, Zhiliang, ZHOU, Shicheng, et al. RLJP: Legal Judgment Prediction via First-Order Logic Rule-enhanced with Large Language Models. arXiv preprint arXiv:2505.21281, 2025

    2025

  43. [58]

    Using Bidirectional Encoder Representations from Transformers (BERT) to predict criminal charges and sentences from Taiwanese court judgments

    PENG YT, Lei CL. Using Bidirectional Encoder Representations from Transformers (BERT) to predict criminal charges and sentences from Taiwanese court judgments. PeerJ Comput Sci. 2024 Jan 31;10:e1841. doi: 10.7717/peerj-cs.1841. PMID: 38435559; PMCID: PMC10909178

    2024 doi

  44. [59]

    Analysis and prediction of legal judgements in the Slovak criminal

    V ARGA, Dávid, SZOPLÁK, Zoltán, KRAJCI, Stanislav, et al. Analysis and prediction of legal judgements in the Slovak criminal. 2021. <https://ceur-ws.org/V ol-2962/paper13.pdf>

    2021

  45. [60]

    Equality before the law: legal judgment consistency analysis for fairness

    WANG, Yuzhong, XIAO, Chaojun, MA, Shirong, et al. Equality before the law: legal judgment consistency analysis for fairness. arXiv preprint arXiv:2103.13868, 2021

    2021 arXiv

  46. [61]

    WU Yifan, Reducing Judicial Inconsistency through AI: A Review of Legal Judgement Prediction Models, ITM Web of Conferences 70, 02009 (2025), DAI 2024: doi.org/10.1051/itmconf/20257002009

    2025

  47. [62]

    Using machine learning to predict decisions of the European Court of Human Rights

    MEDVEDEV A, Masha, VOLS, Michel, et WIELING, Martijn. Using machine learning to predict decisions of the European Court of Human Rights. Artificial Intelligence and Law, 2020, vol. 28, no 2, p. 237-266

    2020

  48. [63]

    BARALE Claire , Michael ROV ATSOS, Nehal BHUTA, When Fairness Isn’t Statistical: The Limits of Machine Learning in Evaluating Legal Reasoning, arXiv:2506.03913v1 [cs.CL] 04 Jun 2025

    2025 arXiv

  49. [64]

    Legal judgment prediction: If you are going to do it, do it right

    MEDVEDEV A, Masha et MCBRIDE, Pauline. Legal judgment prediction: If you are going to do it, do it right. In : Proceedings of the Natural Legal Language Processing Workshop 2023. 2023. p. 73-84

    2023

  50. [65]

    Legal Judgment Prediction: A Survey of the State of the Art

    FENG, Yi; LI, Chuanyi; NG, Vincent. Legal Judgment Prediction: A Survey of the State of the Art. En IJCAI

  51. [66]

    Predicting Brazilian court decisions

    LAGE-FREITAS, André, ALLENDE-CID, Héctor, SANTANA, Orivaldo, et al. Predicting Brazilian court decisions. PeerJ Computer Science, 2022, vol. 8, p. e904. 23

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.