Skip to content

Dataset · google

Multilingual question-answering benchmark

google/xquad

A parallel corpus of 1,190 question-answer pairs across 10+ languages for cross-lingual QA evaluation.

What's inside

It is based on 240 paragraphs and 1,190 question-answer pairs from the SQuAD v1.1 validation set. The texts are professionally translated into ten languages, including Russian, Spanish, German, Greek, Turkish, Arabic, Vietnamese, Thai, and Chinese.

What you can do with it

  • Evaluating the quality of multilingual LLMs in text-based question-answering (QA) tasks
  • Testing models' ability to transfer knowledge across languages without fine-tuning
  • Testing RAG systems on multilingual knowledge bases
Tasks
Question answering
Languages
Arabic, German, Greek, English, Spanish, Hindi, Romanian, Russian and 4 more
Size
10K–100K rows
Downloads this month
12.9K
Likes
43

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("google/xquad")