Langhui Tech has deeply cultivated legal AI data services, building 18 high-quality annotated datasets across five major domains: judgment documents, laws and regulations, contract review, legal Q&A, and judicial reasoning. Covering 5M+ judgment documents, 2M+ legal question bank entries, and 1.5M+ legal QA pairs, we provide compliant, professional, and high-quality corpora for legal large model training.
From GPT-4 legal reasoning to AI judge assistants, from intelligent contract review to legal retrieval large models, AI is reshaping legal services and judicial systems.
A latest Harvard study shows that GPT-4 reaches lawyer-level performance on legal reasoning tasks, with accuracy above 85% on multiple benchmarks. It performs close to top law school graduates on LSAT, Bar Exam, and other legal exams, triggering a paradigm shift in legal education.
China's Supreme People's Court is advancing the "Digital Case Handling" project, with AI assisting in judgment document generation, similar case recommendations, and sentencing references, covering 3,500+ courts nationwide. 2025 data shows AI assistance improves judge efficiency by 40%, with similar case recommendation accuracy reaching 92%.
Trained on millions of legal corpora, LawGPT supports legal Q&A, document generation, case retrieval, and sentencing prediction. Nominated for Best Paper at ACL 2025, it comprehensively outperforms GPT-4 on the LegalBench legal reasoning benchmark.
A Stanford-released contract review AI achieves 96% accuracy in risk clause identification, surpassing the average level of senior lawyers. It has been applied to large enterprise contract review workflows, reducing per-contract review time from 4 hours to 5 minutes.
A legal retrieval system based on dense retrieval + LLM achieves over 90% retrieval accuracy across hundreds of millions of judgment documents. A SIGIR 2025 paper proposes LegalRetriever, a legal retrieval-specific large model that outperforms traditional keyword retrieval on complex legal questions.
The EU is advancing multilingual legal AI systems that support legal text understanding and translation across 24 official languages. The EU AI Act 2026 establishes strict compliance requirements for high-risk AI systems in legal domains, driving the development of transparent and explainable legal AI.
Deeply cultivating five major domains — judgment documents, laws and regulations, contract review, legal Q&A, and judicial reasoning — to build a full-link legal data infrastructure from professional annotation to compliant delivery.
Multi-level judgment documents from Supreme Court, High Courts, and Intermediate Courts, with structured annotations including case type, verdict, and legal basis.
Full-hierarchy regulation library covering Constitution, laws, administrative regulations, and local regulations, with annotations for efficacy hierarchy, revision history, and citation relationships.
Multiple contract types including sales, lease, labor, and equity transfer, with annotations for risk clauses, compliance, and abnormal clauses.
Legal consultations, exam question banks, and judicial interpretation Q&A, with annotations for legal article citations, case correlations, and answer logic chains.
Structured annotations including entity recognition, relation extraction, event extraction, legal article citations, and verdict results, supporting full-scenario legal NLP.
An expert review team composed of practicing lawyers, law professors, and senior judges, ensuring the professionalism and authority of legal annotations.
Strictly complying with the Data Security Law and Personal Information Protection Law. All data is desensitized, with compliant authorization delivery solutions provided.
Self-developed legal AI-assisted annotation platform supporting automatic extraction, intelligent classification, and legal article association, boosting annotation efficiency by 10x.
Focusing on five major domains — judgment documents, laws and regulations, contract review, legal Q&A, and judicial reasoning — to build the data foundation for legal AI training.
Covering civil, criminal, administrative, and enforcement cases, with metadata annotations including court level, verdict, and legal basis.
Summaries and key point extractions of judgment documents, with structured summary annotations including case facts, disputed issues, and ruling essentials.
Criminal case sentencing data, with annotations for charges, circumstances, sentencing ranges, and sentencing consistency, supporting sentencing assistant AI.
Court enforcement case data, with annotations for parties subject to enforcement, enforcement targets, enforcement results, and default information.
Full-hierarchy regulation library covering Constitution, laws, administrative regulations, local regulations, and departmental rules, with efficacy hierarchy and revision history annotations.
A relationship graph of citation, modification, and repeal among legal provisions, supporting legal traceability and impact analysis.
Parallel legal text corpora in Chinese, English, Japanese, Korean, and other languages, supporting cross-lingual legal retrieval and translation.
Multiple contract types including sales, lease, labor, and equity transfer, with annotations for risk clauses, compliance, and abnormal clauses.
Litigation documents such as legal complaints and answers, with annotations for litigation claims, factual reasons, and legal basis.
Various standard contract templates, with annotations for clause structure, risk points, and mandatory clauses, supporting contract generation AI training.
Contract data with risk clause annotations, supporting automatic identification of breach of contract liability, unfair clauses, and compliance risks.
Legal consultation Q&A pairs, with annotations for question classification, legal article citations, case correlations, and answer logic chains.
Common law legal exam question bank, with questions, answers, analyses, and legal article citations, covering core Anglo-American law domains.
Lawyer-client multi-turn dialogue data, with annotations for intent recognition, sentiment analysis, and legal recommendations.
Entity annotations for person names, organizations, locations, legal articles, charges, and other entities in legal text, in BIO format, supporting legal NLP.
Judicial cases with reasoning chain annotations, covering the complete reasoning process from fact-finding to law application, supporting legal reasoning large models.
A relationship graph of legal entities such as legal articles, cases, charges, and parties, supporting legal knowledge Q&A and reasoning.
Case outcome prediction data, with annotations for case features, verdicts, and appeal results, supporting judicial outcome prediction AI.
Showing the JSON annotation structure for four data types: judgment documents, laws and regulations, contract review, and legal Q&A.
Six typical AI application scenarios based on Langhui legal datasets.
A legal large model trained on legal QA data, providing 24/7 intelligent legal consultation services with accuracy above 90%.
AI automatically analyzes judgment documents, extracting case elements, verdicts, and legal basis, supporting similar case retrieval and sentencing references.
AI automatically reviews contract risk clauses, identifying breach of contract liability and unfair clauses, boosting review efficiency by 48x.
A legal retrieval system based on dense retrieval, achieving over 90% retrieval accuracy across hundreds of millions of judgment documents.
Predicting verdicts and appeal success rates based on case features, providing data support for lawyers' litigation strategies.
AI detects corporate compliance risks, automatically identifying legal compliance issues and reducing enterprise legal risks.
Frequently asked questions about legal datasets
Langhui legal datasets cover five major domains — judgment documents, laws and regulations, contract review, legal Q&A, and judicial reasoning — with 18 core datasets. These include Chinese judgment documents (5M), legal case summaries, sentencing references, enforcement cases, laws and regulations library (800K entries), legal article citation relationships, multilingual legal corpus, contract review (300K), complaints and answers, contract templates, contract risk identification, legal Q&A (1.5M pairs), legal question bank (2M), legal consultation dialogues, legal named entity recognition, judicial reasoning, legal knowledge graph, and judicial prediction — totaling 5M+ judgment documents, 2M+ legal question bank entries, and 1.5M+ legal QA pairs.
We strictly comply with the Data Security Law, Personal Information Protection Law, Cybersecurity Law, and other relevant laws and regulations. All data involving personal privacy is desensitized (names, ID numbers, addresses, etc.). Judgment document data is sourced from public channels with authorized use. We provide a complete compliance authorization chain and support enterprise-grade compliant delivery solutions, ensuring the legality and compliance of data use.
Suitable for six typical scenarios: intelligent legal Q&A, judgment document analysis, intelligent contract review, legal retrieval, judicial prediction, and compliance risk detection. It supports the training and fine-tuning of legal large models (LawGPT), providing professional AI training data for courts, law firms, corporate legal departments, and legal tech companies, and has served multiple legal institutions.
We adopt a "legal expert + AI assistance" dual quality control system. Practicing lawyers, law professors, and senior judges jointly conduct professional reviews of annotated content; our self-developed AI-assisted annotation platform performs format validation, consistency checks, and other automated quality inspections. Legal entity recognition accuracy exceeds 92%, and legal article citation accuracy exceeds 95%, ensuring the professionalism and authority of the data.
We support multiple delivery formats including text formats (TXT/DOCX/PDF), structured formats (JSON/XML/CSV), annotation formats (BIO/JSONL/COCO), and knowledge graph formats (RDF/OWL/N-Triples). We also provide supporting resources such as data dictionaries, annotation specification documents, and legal entity reference tables, supporting API integration and customized delivery.
Contact Langhui data experts to get the complete legal dataset catalog, trial access, and customized solutions.