XLM is a Transformer based architecture that is pre-trained using one of three language modelling objectives:
The authors find that both the CLM and MLM approaches provide strong cross-lingual features that can be used for pretraining models.
Source: Cross-lingual Language Model PretrainingPaper | Code | Results | Date | Stars |
---|
Task | Papers | Share |
---|---|---|
Language Modelling | 17 | 8.76% |
Translation | 15 | 7.73% |
Sentence | 11 | 5.67% |
Machine Translation | 10 | 5.15% |
XLM-R | 8 | 4.12% |
Cross-Lingual Transfer | 8 | 4.12% |
Question Answering | 8 | 4.12% |
Retrieval | 6 | 3.09% |
NER | 6 | 3.09% |