Benchmarks & Datasets
Evaluation suites and datasets for legal-domain systems.
Italian Senate bulk Akoma Ntoso data
The code used to evaluate embedding models on the Massive Legal Embedding Benchmark (MLEB).
[ICLR] This Repo provides code for evaluating LLMs on LEXam - a comprehensive benchmark evaluating AI system's legal reasoning ability with law exam questions, using both open and multiple-choice questions.
German Dataset for Legal Information Retrieval
CaseHOLD (Case Holdings On Legal Decisions) is a law dataset comprised of over 53,000+ multiple choice questions to identify the relevant holding of a cited case.
The world’s most comprehensive open-source patents AI assistant — Search and analyze the entire USPTO corpus, track innovation trends, link patents to federal spending and grants data, and uncover emerging technologies — all through natural conversation.
Dataset Card for "IL-TUR" Dataset Description Summary "IL-TUR": Benchmark for Indian Legal Text Understanding and Reasoning is a collaborative effort to establish a modern benchmark for training and evaluating AI/NLP models on Indian Law. IL-TUR consists of 8 foundational tasks, requiring different types of understanding and skills. Apart from English, some tasks involve Indic languages. This dataset repository has been created to unify the data… See the full description on the dataset page: https://huggingface.co/datasets/Exploration-Lab/IL-TUR.
Dataset Card for "LexFiles" Dataset Summary The LeXFiles is a new diverse English multinational legal corpus that we created including 11 distinct sub-corpora that cover legislation and case law from 6 primarily English-speaking legal systems (EU, CoE, Canada, US, UK, India). The corpus contains approx. 19 billion tokens. In comparison, the "Pile of Law" corpus released by Hendersons et al. (2022) comprises 32 billion in total, where the majority (26/30) of… See the full description on the dataset page: https://huggingface.co/datasets/lexlms/lex_files.
Dataset Summary This dataset was introduced in CaseSumm: A Large-Scale Dataset for Long-Context Summarization from U.S. Supreme Court Opinions. The CaseSumm dataset consists of U.S. Supreme Court cases and their official summaries, called syllabuses, from the period 1815-2019. Syllabuses are written by an attorney employed by the Court and approved by the Justices. The syllabus is therefore the gold standard for summarizing majority opinions, and ideal for evaluating other… See the full description on the dataset page: https://huggingface.co/datasets/ChicagoHAI/CaseSumm.
Legal data in machine-readable form
⚖️ IndianBailJudgments-1200: Annotated Indian Bail Order Dataset (1975–2025) IndianBailJudgments-1200 is a high-quality, structured dataset comprising 1,200 annotated Indian bail-related court orders spanning five decades (1975–2025). It captures granular legal information across 78 courts and 28 regions, including crime types, IPC sections invoked, judge names, legal issues, bail outcomes, and bias indicators. Designed for use in legal NLP, fairness analysis, and judicial… See the full description on the dataset page: https://huggingface.co/datasets/SnehaDeshmukh/IndianBailJudgments-1200.
Dataset Card for EurlexResources: A Corpus Covering the Largest EURLEX Resources Dataset Summary This dataset contains large text resources (~179GB in total) from EURLEX that can be used for pretraining language models. Use the dataset like this: from datasets import load_dataset config = "de_caselaw" # {lang}_{resource} dataset = load_dataset("joelito/eurlex_resources", config, split='train', streaming=True) Supported Tasks and Leaderboards The… See the full description on the dataset page: https://huggingface.co/datasets/joelniklaus/eurlex_resources.
Swiss Case Law Dataset 1,050,000+ published decision records (~909,000 unique decisions) from Swiss federal, cantonal, and regulatory bodies. Figures as of 2026-07-24 — refreshed daily; live counts at opencaselaw.ch. Full text, structured metadata, extracted case-citation references, and daily updates. The dataset contains German, French, and Italian decisions; the export schema also reserves rm for Romansh. Dataset Summary The largest open collection of… See the full description on the dataset page: https://huggingface.co/datasets/voilaj/swiss-caselaw.
Curated dataset and tools for tracking global AI legislation — US federal, state, and international frameworks.
German Legal Sentences (GLS) is an automatically generated training dataset for semantic sentence matching in the domain in german legal documents. It follows the concept of weak supervision, where imperfect labels are generated using multiple heuristics. For this purpose we use a combination of legal citation matching and BM25 similarity. The contained sentences and their citations are parsed from real judicial decisions provided by [Open Legal Data](http://openlegaldata.io/)
LegalSummarization An MTEB dataset Massive Text Embedding Benchmark The dataset consistes of 439 pairs of contracts and their summarizations from https://tldrlegal.com and https://tosdr.org/. Task category t2t Domains Legal, Written Reference https://github.com/lauramanor/legal_summarization How to evaluate on this task You can evaluate an embedding model on this dataset using the following code: import mteb task =… See the full description on the dataset page: https://huggingface.co/datasets/mteb/legal_summarization.
Dataset Card for Free-Law-Project/opinions-synthetic-query-512 This dataset is created from the opinions-metadata, and used for training the Free Law Project Semantic Search models, including Free-Law-Project/modernbert-embed-base_finetune_512. Dataset Details The dataset is curated by Free Law Project by selecting the train split from the opinions-metadata dataset. The dataset is created for finetuning encoder models for semantic search, with 512 context window.… See the full description on the dataset page: https://huggingface.co/datasets/freelawproject/opinions-synthetic-query-512.
OpenRouter for legal work: vendor-neutral, benchmark-grounded skills that route each legal task (drafting, extraction, research, review, translation) to the right LLM. AGPL-3.0.
Build one of these? Verify ownership through GitHub and take over your project's page: tagline, categories, maintainer's note.