BERT MEDIUM
otherentity
ID: bertmedium
Related Papers
Are Larger Pretrained Language Models Uniformly Better? Comparing Performance at the Instance Level
Larger language models have higher accu- racy on average, but are they better on ev- ery single instance (datapoint)? Some work suggests larger models have higher out-of- distribution robustness, while …...
Frustratingly Simple Pretraining Alternatives to Masked Language Modeling
AbstractMasked language modeling (MLM), a self-supervised pretraining objective, is widely used in natural language processing for learning text representations. MLM trains a model to predict a random sample of input …...