Matches in SemOpenAlex for { <https://semopenalex.org/work/W3211449872> ?p ?o ?g. }
- W3211449872 abstract "Transformers that are pre-trained on multilingual corpora, such as, mBERT and XLM-RoBERTa, have achieved impressive cross-lingual transfer capabilities. In the zero-shot transfer setting, only English training data is used, and the fine-tuned model is evaluated on another target language. While this works surprisingly well, substantial variance has been observed in target language performance between different fine-tuning runs, and in the zero-shot setup, no target-language development data is available to select among multiple fine-tuned models. Prior work has relied on English dev data to select among models that are fine-tuned with different learning rates, number of steps and other hyperparameters, often resulting in suboptimal choices. In this paper, we show that it is possible to select consistently better models when small amounts of annotated data are available in auxiliary pivot languages. We propose a machine learning approach to model selection that uses the fine-tuned model's own internal representations to predict its cross-lingual capabilities. In extensive experiments we find that this method consistently selects better models than English validation data across twenty five languages (including eight low-resource languages), and often achieves results that are comparable to model selection using target language development data." @default.
- W3211449872 created "2021-11-22" @default.
- W3211449872 creator A5020819058 @default.
- W3211449872 creator A5039096905 @default.
- W3211449872 date "2020-10-12" @default.
- W3211449872 modified "2023-10-16" @default.
- W3211449872 title "Model Selection for Cross-Lingual Transfer" @default.
- W3211449872 cites W2126725946 @default.
- W3211449872 cites W2143331230 @default.
- W3211449872 cites W2156387975 @default.
- W3211449872 cites W2229177960 @default.
- W3211449872 cites W2493916176 @default.
- W3211449872 cites W2604763608 @default.
- W3211449872 cites W2739967986 @default.
- W3211449872 cites W2915429162 @default.
- W3211449872 cites W2950733326 @default.
- W3211449872 cites W2952303884 @default.
- W3211449872 cites W2962739339 @default.
- W3211449872 cites W2963341956 @default.
- W3211449872 cites W2963403868 @default.
- W3211449872 cites W2963826397 @default.
- W3211449872 cites W2964022985 @default.
- W3211449872 cites W2964114970 @default.
- W3211449872 cites W2964121744 @default.
- W3211449872 cites W2970697704 @default.
- W3211449872 cites W2970854433 @default.
- W3211449872 cites W2971145411 @default.
- W3211449872 cites W2971675764 @default.
- W3211449872 cites W2983040767 @default.
- W3211449872 cites W2990761674 @default.
- W3211449872 cites W3011411500 @default.
- W3211449872 cites W3034776473 @default.
- W3211449872 cites W3035032094 @default.
- W3211449872 cites W3035497479 @default.
- W3211449872 cites W3100198908 @default.
- W3211449872 cites W3101516616 @default.
- W3211449872 cites W3102483398 @default.
- W3211449872 cites W3104820280 @default.
- W3211449872 cites W3167030571 @default.
- W3211449872 cites W759515131 @default.
- W3211449872 doi "https://doi.org/10.48550/arxiv.2010.06127" @default.
- W3211449872 hasPublicationYear "2020" @default.
- W3211449872 type Work @default.
- W3211449872 sameAs 3211449872 @default.
- W3211449872 citedByCount "0" @default.
- W3211449872 crossrefType "posted-content" @default.
- W3211449872 hasAuthorship W3211449872A5020819058 @default.
- W3211449872 hasAuthorship W3211449872A5039096905 @default.
- W3211449872 hasBestOaLocation W32114498721 @default.
- W3211449872 hasConcept C119857082 @default.
- W3211449872 hasConcept C121332964 @default.
- W3211449872 hasConcept C121955636 @default.
- W3211449872 hasConcept C137293760 @default.
- W3211449872 hasConcept C144133560 @default.
- W3211449872 hasConcept C150899416 @default.
- W3211449872 hasConcept C154945302 @default.
- W3211449872 hasConcept C165801399 @default.
- W3211449872 hasConcept C173608175 @default.
- W3211449872 hasConcept C196083921 @default.
- W3211449872 hasConcept C204321447 @default.
- W3211449872 hasConcept C2776145971 @default.
- W3211449872 hasConcept C2776175482 @default.
- W3211449872 hasConcept C41008148 @default.
- W3211449872 hasConcept C62520636 @default.
- W3211449872 hasConcept C66322947 @default.
- W3211449872 hasConcept C81917197 @default.
- W3211449872 hasConcept C8642999 @default.
- W3211449872 hasConcept C93959086 @default.
- W3211449872 hasConceptScore W3211449872C119857082 @default.
- W3211449872 hasConceptScore W3211449872C121332964 @default.
- W3211449872 hasConceptScore W3211449872C121955636 @default.
- W3211449872 hasConceptScore W3211449872C137293760 @default.
- W3211449872 hasConceptScore W3211449872C144133560 @default.
- W3211449872 hasConceptScore W3211449872C150899416 @default.
- W3211449872 hasConceptScore W3211449872C154945302 @default.
- W3211449872 hasConceptScore W3211449872C165801399 @default.
- W3211449872 hasConceptScore W3211449872C173608175 @default.
- W3211449872 hasConceptScore W3211449872C196083921 @default.
- W3211449872 hasConceptScore W3211449872C204321447 @default.
- W3211449872 hasConceptScore W3211449872C2776145971 @default.
- W3211449872 hasConceptScore W3211449872C2776175482 @default.
- W3211449872 hasConceptScore W3211449872C41008148 @default.
- W3211449872 hasConceptScore W3211449872C62520636 @default.
- W3211449872 hasConceptScore W3211449872C66322947 @default.
- W3211449872 hasConceptScore W3211449872C81917197 @default.
- W3211449872 hasConceptScore W3211449872C8642999 @default.
- W3211449872 hasConceptScore W3211449872C93959086 @default.
- W3211449872 hasLocation W32114498721 @default.
- W3211449872 hasOpenAccess W3211449872 @default.
- W3211449872 hasPrimaryLocation W32114498721 @default.
- W3211449872 hasRelatedWork W3092736191 @default.
- W3211449872 hasRelatedWork W3100198908 @default.
- W3211449872 hasRelatedWork W3107474891 @default.
- W3211449872 hasRelatedWork W3164972323 @default.
- W3211449872 hasRelatedWork W3177920269 @default.
- W3211449872 hasRelatedWork W3199116325 @default.
- W3211449872 hasRelatedWork W3211449872 @default.
- W3211449872 hasRelatedWork W3212870931 @default.
- W3211449872 hasRelatedWork W4225144544 @default.
- W3211449872 hasRelatedWork W4304128395 @default.