Dataset · textmachinelab
Multiple-choice questions for reading comprehension
textmachinelab/quail
15 thousand English-language questions on short texts from four domains for reading comprehension evaluation.
What's inside
The balanced English-language dataset contains 15,000 multiple-choice questions labeled by type. The questions are formulated for text passages of 300–350 tokens from four sources: news, user stories, fiction, and blogs.
What you can do with it
- Evaluating reading comprehension and logical reasoning skills of language models
- Fine-tuning (SFT) models on question-answering tasks given a context
- Analyzing LLM errors depending on the question type and text genre
- Tasks
- Multiple choice
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 20.6K
- Likes
- 8
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("textmachinelab/quail")