{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/xnli-evaluating-cross-lingual-sentence","title":"XNLI: Evaluating Cross-lingual Sentence Representations","arxiv_id":"1809.05053","date":"2018-09-13","proceeding":"EMNLP 2018 10","authors":["Alexis Conneau","Guillaume Lample","Ruty Rinott","Adina Williams","Samuel R. Bowman","Holger Schwenk","Veselin Stoyanov"],"abstract":"State-of-the-art natural language processing systems rely on supervision in\nthe form of annotated data to learn competent models. These models are\ngenerally trained on data in a single language (usually English), and cannot be\ndirectly used beyond that language. Since collecting data in every language is\nnot realistic, there has been a growing interest in cross-lingual language\nunderstanding (XLU) and low-resource cross-language transfer. In this work, we\nconstruct an evaluation set for XLU by extending the development and test sets\nof the Multi-Genre Natural Language Inference Corpus (MultiNLI) to 15\nlanguages, including low-resource languages such as Swahili and Urdu. We hope\nthat our dataset, dubbed XNLI, will catalyze research in cross-lingual sentence\nunderstanding by providing an informative standard evaluation task. In\naddition, we provide several baselines for multilingual sentence understanding,\nincluding two based on machine translation systems, and two that use parallel\ndata to train aligned multilingual bag-of-words and LSTM encoders. We find that\nXNLI represents a practical and challenging evaluation suite, and that directly\ntranslating the test data yields the best performance among available\nbaselines.","url_abs":"http://arxiv.org/abs/1809.05053v1","url_pdf":"http://arxiv.org/pdf/1809.05053v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/1-punchMan/CLTS","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/Somefive/XNLI","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/blazejdolicki/multilingual-analysis","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/deterministic-algorithms-lab/Large-XLM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/facebookresearch/XLM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/facebookresearch/XNLI","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/feyzaakyurek/XLM-LwLL","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/fshdnc/enfi-XLM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"xnli-evaluating-cross-lingual-sentence","repo_url":"https://github.com/kheeong/XLM_OWN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"cross-lingual-natural-language-inference","task_name":"Cross-Lingual Natural Language Inference"},{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"translation","task_name":"Translation"}],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[{"slug":"xnli","name":"XNLI","full_name":"Cross-lingual Natural Language Inference"}],"methods_introduced":[],"results":[{"leaderboard":"/sota/natural-language-inference-on-xnli-french","task":"Natural Language Inference","dataset":"XNLI French","model":"BiLSTM-max","rank_in_archive_order":6,"of":6,"metrics":{"Accuracy":"68.3"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1809.05053","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}