{"url":"/dataset/mlqa","name":"MLQA","full_name":"MultiLingual Question Answering","description_markdown":"MLQA (MultiLingual Question Answering) is a benchmark dataset for evaluating cross-lingual question answering performance. MLQA consists of over 5K extractive QA instances (12K in English) in SQuAD format in seven languages - English, Arabic, German, Spanish, Hindi, Vietnamese and Simplified Chinese. MLQA is highly parallel, with QA instances parallel between 4 different languages on average.\r\n\r\nSource: [Facebook Research](https://github.com/facebookresearch/mlqa)\r\nImage Source: [https://github.com/facebookresearch/mlqa](https://github.com/facebookresearch/mlqa)","description_withheld":null,"homepage":"https://github.com/facebookresearch/mlqa","introduced_date":"2019-10-16","introduced_date_note":null,"introduced_by":{"paper":"/paper/mlqa-evaluating-cross-lingual-extractive","title":"MLQA: Evaluating Cross-lingual Extractive Question Answering","first_author":"Patrick Lewis","url":null},"license":{"name":"CC-BY-SA 3.0","url":"https://creativecommons.org/licenses/by-sa/3.0/legalcode"},"modalities":[{"name":"Texts","url":"/datasets/modality/texts"}],"tasks":[{"name":"Question Answering","url":"/task/question-answering","datasets_with_task":"/datasets/task/question-answering"},{"name":"Machine Translation","url":"/task/machine-translation","datasets_with_task":"/datasets/task/machine-translation"},{"name":"Cross-Lingual Transfer","url":"/task/cross-lingual-transfer","datasets_with_task":"/datasets/task/cross-lingual-transfer"},{"name":"Cross-Lingual Question Answering","url":"/task/cross-lingual-question-answering","datasets_with_task":"/datasets/task/cross-lingual-question-answering"}],"languages":[{"name":"English","url":"/datasets/language/english"},{"name":"Spanish","url":"/datasets/language/spanish"},{"name":"German","url":"/datasets/language/german"},{"name":"Chinese","url":"/datasets/language/chinese"},{"name":"Arabic","url":"/datasets/language/arabic"},{"name":"Czech","url":"/datasets/language/czech"},{"name":"Hindi","url":"/datasets/language/hindi"},{"name":"Korean","url":"/datasets/language/korean"},{"name":"Vietnamese","url":"/datasets/language/vietnamese"}],"variants":["MLQA"],"data_loaders":[{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/mlqa","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/TheTung/mlqa","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/facebook/mlqa","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/tensorflow/datasets","url":"https://www.tensorflow.org/datasets/catalog/mlqa","frameworks":["tf","jax"]},{"repo":"https://github.com/facebookresearch/MLQA","url":"https://github.com/facebookresearch/mlqa","frameworks":[]}],"num_papers_in_archive":167,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[{"leaderboard":"/sota/cross-lingual-question-answering-on-mlqa","task":"Cross-Lingual Question Answering","dataset_variant":"MLQA","rows":3,"metrics":["F1","EM"],"first_row_in_archive_order":{"model":"ByT5 XXL","paper":"/paper/byt5-towards-a-token-free-future-with-pre","metrics":{"EM":"54.9","F1":"71.6"},"code_links":[{"title":"huggingface/transformers","url":"https://github.com/huggingface/transformers/tree/master/src/transformers/models/byt5"},{"title":"google-research/byt5","url":"https://github.com/google-research/byt5"},{"title":"ufal/multilexnorm2021","url":"https://github.com/ufal/multilexnorm2021"},{"title":"2024-MindSpore-1/Code2","url":"https://github.com/2024-MindSpore-1/Code2/tree/main/model-1/byt5"},{"title":"yoreG123/Paddle-ByT5","url":"https://github.com/yoreG123/Paddle-ByT5"}]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"}],"papers_with_a_benchmark_row":[{"paper":"/paper/byt5-towards-a-token-free-future-with-pre","title":"ByT5: Towards a token-free future with pre-trained byte-to-byte models","date":"2021-05-28","rows_on_this_dataset":1,"code_links":5,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":6,"samples_ran":0,"samples_unverified":6,"pointer_only_for_licence":0,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/rethinking-embedding-coupling-in-pre-trained-1","title":"Rethinking embedding coupling in pre-trained language models","date":"2020-10-24","rows_on_this_dataset":2,"code_links":4,"syntology":null}],"syntology_totals":{"read_at":"2026-09-24T18:15:14+00:00","papers_with_samples":1,"samples_harvested":6,"samples_ran":0,"samples_unverified":6,"pointer_only_for_licence":0,"papers_with_no_sample_that_ran":1,"note":"the per-paper counts above, summed; not a rate"},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks."}