The Belebele Benchmark: a Parallel Reading Comprehension Dataset in 122 Language Variants

TL;DR

Belebele is a comprehensive multilingual reading comprehension dataset in 122 languages, enabling evaluation of NLP models across diverse language resources and revealing insights into multilingual model capabilities.

Contribution

This paper introduces Belebele, a large-scale, parallel, multilingual MRC dataset covering 122 languages, expanding evaluation scope for language understanding models.

Findings

Smaller MLMs outperform larger LLMs on many languages.

Vocabulary size and construction impact low-resource language performance.

English-centric LLMs show significant cross-lingual transfer.

Abstract

We present Belebele, a multiple-choice machine reading comprehension (MRC) dataset spanning 122 language variants. Significantly expanding the language coverage of natural language understanding (NLU) benchmarks, this dataset enables the evaluation of text models in high-, medium-, and low-resource languages. Each question is based on a short passage from the Flores-200 dataset and has four multiple-choice answers. The questions were carefully curated to discriminate between models with different levels of general language comprehension. The English dataset on its own proves difficult enough to challenge state-of-the-art language models. Being fully parallel, this dataset enables direct comparison of model performance across all languages. We use this dataset to evaluate the capabilities of multilingual masked language models (MLMs) and large language models (LLMs). We present extensive results and find that despite significant cross-lingual transfer in English-centric LLMs, much smaller MLMs pretrained on balanced multilingual data still understand far more languages. We also observe that larger vocabulary size and conscious vocabulary construction correlate with better performance on low-resource languages. Overall, Belebele opens up new avenues for evaluating and analyzing the multilingual capabilities of NLP systems.