Language Contamination Helps Explains the Cross-lingual Capabilities of English Pretrained Models
Open accessAbstract
English pretrained language models, which make up the backbone of many modern NLP systems, require huge amounts of unlabeled training data.These models are generally presented as being trained only on English text but have been found to transfer surprisingly well to other languages.We investigate this phenomenon and find that common English pretraining corpora actually contain significant amounts of non-English text: even when less than 1% of data is not English (well within the error rate of strong language classifiers), this leads to hundreds of millions of foreign language tokens in large-scale datasets.We then demonstrate that even these small percentages of non-English data facilitate cross-lingual transfer for models trained on them, with target language performance strongly correlated to the amount of in-language data seen during pretraining.In light of these findings, we argue that no model is truly monolingual when pretrained at scale, which should be considered when evaluating cross-lingual transfer.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Most LLMs are trained mainly on English corpora.
“Most LLMs are trained mainly on English corpora35, while many countries’ medical systems use minority languages that most LLMs are not familiar with.”
From Probabilistic medical predictions of large language models (2024) · cites this paper as [35]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
The citation lists haven’t been fetched yet
References 22 · Cited by 27
Research works printed but not linked: 2
- Evangelia Gogoulou, Ariel Ekgren, Tim Isbister, and Magnus Sahlgren. 2021. Cross-lingual trans- fer of monolingual models. arXiv preprint arXiv:2109.07348. · arXiv
- Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. 2019. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9.
- Authors
- Blevins, T., & Zettlemoyer, L.
- Affiliations
- University of Washington (US)
- ORCID
- Zettlemoyer, Luke
- Venue
- Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing
- Published
- 2022
- Pages
- 3563-3574
- OpenAlex
- W4385573174
- Cited by
- 27
- References
- 22
- Citations by year
- 2026: 5 · 2025: 6 · 2024: 5 · 2023: 11
- FWCI
- 3.04
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Conference on Empirical Methods in Natural Language Processing (EMNLP) (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- language contamination, cross-lingual transfer, pre-trained language models, multilingual corpora, cross-lingual evaluation, multilingual NLP
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
