| linguistic-acceptability-on-cola-glue-dev | ELECTRA-1.75M | #2 | Matthews Correlation: 69.1 |
| natural-language-inference-on-rte-glue-dev | ELECTRA-1.75M | #2 | Accuracy: 88.0 |
| semantic-textual-similarity-on-sts-b-glue-dev | ELECTRA-1.75M | #2 | Pearson/Spearman Correlation: 92.6 |
| natural-language-inference-on-multinli-dev | ELECTRA-1.75M | #3 | Matched: 90.9 |
| natural-language-inference-on-qnli-glue-dev | ELECTRA-1.75M | #3 | Accuracy: 95.0 |
| natural-language-understanding-on-glue-dev | ELECTRA-1.75M | #4 | GLUE (dev): 89.5 |
| natural-language-understanding-on-glue-test | ELECTRA-1.75M | #4 | GLUE (test): 89.4 |
| natural-language-inference-on-wnli | ELECTRA-1.75M | #5 | Accuracy: 92.5 |
| paraphrase-identification-on-mrpc-glue-dev | ELECTRA-1.75M | #5 | F1: 90.8 |
| question-answering-on-quora-question-pairs | ELECTRA | #7 | Accuracy: 90.1% |
| linguistic-acceptability-on-cola | ELECTRA-1.75M | #8 | Accuracy: 71.7% |
| natural-language-inference-on-qnli | ELECTRA-1.75M | #8 | Accuracy: 95.8% |
| sentiment-analysis-on-sst-2-binary-classification | ELECTRA | #8 | Accuracy: 96.9 |
| natural-language-inference-on-rte | ELECTRA-1.75M | #15 | Accuracy: 89.8% |
| question-answering-on-squad20 | ELECTRA-1.75M | #28 | F1: 91.4 |
| paraphrase-identification-on-quora-question-pairs | ELECTRA-1.75M | #32 | Accuracy: 90.8 |
| question-answering-on-squad11-dev | ELECTRA-1.75M | #49 | F1: 94.9 |