EXAMS — Multilingual school exam questions
EXAMS is a multilingual dataset designed for training and evaluating question-answer (QA) models in an educational context. It includes more than 24,000 high school exam questions in 16 different languages, covering subjects such as mathematics, biology, history, or geography. Each entry includes a question, context, and multiple answer choices (in some cases). It is a unique benchmark to test the ability of models to generalize across languages and school domains.
Description
EXAMS is a multilingual QA benchmark based on real school exams. It covers a wide range of languages and subjects, making it an ideal tool for testing the performance of NLP models in a variety of linguistic and thematic contexts.
What is this dataset for?
- Form models of Multilingual question and answer
- Evaluate cross-language generalization skills
- Testing educational models in a realistic context
Can it be enriched or improved?
Yes: by adding new topics or grade levels, by standardizing answer formats, or by translating questions into other languages for cross-language alignment.
🔎 In summary
🧠 Recommended for
- Multilingual QA
- Reading comprehension assessment
- Educational models
🔧 Compatible tools
- Hugging Face Transformers
- Haystack
- Langchain
- OpenBookQA
💡 Tip
Can be used to create AI assessments that simulate a human school test in multiple languages.
Frequently Asked Questions
What types of school subjects are covered by EXAMS?
The dataset covers 24 subjects, including biology, physics, mathematics, literature, history, and geography.
Are the questions accompanied by context or reading material?
Yes, in many cases, each question comes with a supporting paragraph or context explaining the topic.
Can the dataset be used for educational or AI tutoring models?
Absolutely, this is one of its main use cases: training AIs to answer realistic educational questions in several languages.
