Skip to content

Dataset · AdaptLLM

Legal tasks benchmark for evaluating LLMs

AdaptLLM/law-tasks

A collection of English legal test tasks for evaluating model domain adaptation.

What's inside

The dataset includes 1 to 10 thousand examples in English in text and tabular format. It covers text classification, zero-shot classification, and legal question-answering tasks. The dataset was compiled by the authors of an ICLR 2024 paper to evaluate how question-answering skills are preserved after domain pre-training of models.

What you can do with it

  • Evaluating LLM quality on legal tasks and documents
  • Checking the preservation of question-answering capability after domain pre-training
  • Testing models in zero-shot legal text classification tasks
Tasks
Text classification, Question answering, Zero shot classification
Languages
English
Size
1K–10K rows
Downloads this month
337
Likes
37

How to load it

from datasets import load_dataset

ds = load_dataset("AdaptLLM/law-tasks")