LEXam: Legal Reasoning Benchmark on 340 Law Exams

A benchmark suite containing 4,696 multiple-choice questions from 340 law examinations across Swiss, EU, and international jurisdictions. The dataset was created by LEXam-Benchmark and updated in January 2026. It is designed to evaluate AI performance on legal reasoning tasks.

Use Cases

• Benchmarking AI legal reasoning capabilities based on multiple-choice questions from law exams
• Evaluating model performance on questions with varying numbers of answer choices (4, 8, 16, 32)
• Training or fine-tuning language models for legal question-answering tasks
• Analyzing the difficulty and diversity of legal examination questions across different jurisdictions

Strengths

• Contains 4,696 total multiple-choice questions aggregated from 340 distinct law exams
• Questions are organized into four files by the number of answer choices: 1,655 (4 choices), 1,463 (8 choices), 1,028 (16 choices), and 550 (32 choices)
• Sourced from a diverse set of Swiss, EU, and international law examinations

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown for the overall dataset, which may limit suitability assessment
• Description metadata is limited; actual data quality requires manual inspection after download

Provenance

Source

LEXam-Benchmark

Collection Method

Likely compiled from publicly available law examination materials

Freshness

Last updated 2026-01-26 17:11:03; freshness should be verified

Geography

Swiss, EU, and international jurisdictions