Skip to content

Dataset · LEXam-Benchmark

LEXam legal reasoning benchmark

LEXam-Benchmark/LEXam

340 legal exams from Switzerland, the EU, and international law for evaluating reasoning models.

What's inside

LEXam is a legal reasoning evaluation set featuring 340 exam tasks from Switzerland, the EU, and international law. It includes multiple-choice questions broken down by the number of options: 1,655 with 4, 1,463 with 8, 1,028 with 16, and 550 with 32. Data is in English and German, in tabular and text formats; a leaderboard and code are included.

What you can do with it

  • Evaluating LLMs on legal exams
  • Testing reasoning models on MCQs
  • Benchmark for legal AI and RAG systems
Tasks
Text classification, Text generation
Languages
English, German
Size
1K–10K rows
Downloads this month
2.1K
Likes
81

How to load it

from datasets import load_dataset

ds = load_dataset("LEXam-Benchmark/LEXam")