Definition
XQuAD (Cross-lingual Question Answering Dataset) is a benchmark for evaluating the performance of multilingual LLMs and RAG pipelines by providing professionally translated question-answer pairs across 10 languages based on SQuAD v1.1. It highlights the trade-off between using specialized monolingual models for accuracy versus the zero-shot transfer capabilities of large multilingual models.
A multilingual evaluation benchmark, not a model architecture or a training algorithm.
"A Rosetta Stone used as a standardized test to ensure a student understands a story equally well in Spanish, German, or Chinese."
- SQuAD(Prerequisite)
- Zero-shot Cross-lingual Transfer(Component)
- Cross-lingual Information Retrieval (CLIR)(Component)
- Multilingual Embeddings(Component)
Conceptual Overview
XQuAD (Cross-lingual Question Answering Dataset) is a benchmark for evaluating the performance of multilingual LLMs and RAG pipelines by providing professionally translated question-answer pairs across 10 languages based on SQuAD v1.1. It highlights the trade-off between using specialized monolingual models for accuracy versus the zero-shot transfer capabilities of large multilingual models.
Disambiguation
A multilingual evaluation benchmark, not a model architecture or a training algorithm.
Visual Analog
A Rosetta Stone used as a standardized test to ensure a student understands a story equally well in Spanish, German, or Chinese.