CAIL2018: A Large-Scale Legal Dataset for Judgment Prediction

Chaojun Xiao, Haoxi Zhong, Zhipeng Guo, Cunchao Tu et al.

All authors

Chaojun Xiao, Haoxi Zhong, Zhipeng Guo, Cunchao Tu, Zhiyuan Liu, Maosong Sun, Yansong Feng, Xianpei Han, Zhen Hu, Heng Wang, Jianfeng Xu

2018arXiv preprint

Read the paper Publication source

What it adds

Introduces a large Chinese criminal-case dataset connecting court-written facts to charges, criminal-law articles and prison terms. Its baselines expose the difference between high overall accuracy and weak performance on less frequent labels.

Read with this limitation

The corpus retains single-defendant cases and removes low-frequency labels. Splits are random, and charges are heavily imbalanced. Predicting labels extracted from historical judgments does not establish sound sentencing, fair decisions or prospective courtroom performance.

Jurisdiction tags: China
Language: Chinese

Follow the source

Selected statements located in the linked primary sources. These results have not been independently reproduced here.

After filtering, the paper reports 2,676,075 criminal cases, 183 law articles and 202 charges.
Section 2, Dataset Construction
The ten most frequent charges account for 79.0% of cases; the ten least frequent account for 0.12%.
Section 2, Dataset Construction

LawBench: Benchmarking Legal Knowledge of Large Language Models

Both examine Chinese legal-model evaluation, with different tasks and generations of models. This is a thematic connection, not a citation claim.

Cited by

FairLex: A Multilingual Benchmark for Evaluating Fairness in Legal Text Processing

Identifies CAIL2018 as the original Chinese judgment-prediction resource; its fairness experiments use a derived setting.

Inspect the source

Cited by

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

Contrasts its Indian prediction-and-explanation task with the Chinese judgment-prediction corpus.

Inspect the source

LeCaRDv2: A Large-Scale Chinese Legal Case Retrieval Dataset

Both use Chinese criminal judgments; LeCaRDv2 studies case retrieval while CAIL2018 studies judgment prediction.

Entry status, source edition & rights

Index entry: Source-linked summary, prepared with agent assistance. Independent scholarly review is pending.
Archive admission: This bibliographic entry does not confer admission to the versioned archive.
Edition inspected: 1807.02478v1
Reading coverage: Exact arXiv v1 PDF inspected in full (four pages), including preprocessing, random split, metrics and references. No dataset access, licence verification, or experimental rerun performed; this entry describes the 2018 paper.
Rights: External source links only. No paper full text or underlying dataset is mirrored by the site; downstream reuse requires checking the source licence.
Entry updated: 2026-10-15

Source-linked summary · independent assessment pending.