| question-answering-on-naturalqa | SpanBERT | #3 | F1: 82.5 |
| question-answering-on-squad11 | SpanBERT (single model) | #4 | EM: 88.8F1: 94.6Hardware Burden: 586G |
| relation-extraction-on-re-tacred | SpanBERT | #4 | F1: 85.3 |
| relation-classification-on-tacred | SpanBERT | #7 | F1: 70.8 |
| semantic-textual-similarity-on-mrpc | SpanBERT | #7 | Accuracy: 90.9% |
| coreference-resolution-on-ontonotes | SpanBERT | #10 | F1: 79.6 |
| open-domain-question-answering-on-searchqa | SpanBERT | #10 | F1: 84.8 |
| question-answering-on-squad20-dev | SpanBERT | #11 | F1: 86.8 |
| question-answering-on-newsqa | SpanBERT | #12 | F1: 73.6 |
| natural-language-inference-on-multinli | SpanBERT | #13 | Matched: 88.1 |
| paraphrase-identification-on-quora-question-pairs | SpanBERT | #13 | F1: 71.9Accuracy: 89.5 |
| semantic-textual-similarity-on-sts-benchmark | SpanBERT | #14 | Pearson Correlation: 0.899 |
| question-answering-on-squad20 | SpanBERT | #16 | EM: 85.7F1: 88.7 |
| relation-extraction-on-tacred | SpanBERT-large | #17 | F1: 70.8 |
| natural-language-inference-on-qnli | SpanBERT | #18 | Accuracy: 94.3% |
| linguistic-acceptability-on-cola | SpanBERT | #20 | Accuracy: 64.3% |
| sentiment-analysis-on-sst-2-binary-classification | SpanBERT | #30 | Accuracy: 94.8 |
| natural-language-inference-on-rte | SpanBERT | #37 | Accuracy: 79.0% |
| question-answering-on-triviaqa | SpanBERT | #37 | F1: 83.6 |