| linguistic-acceptability-on-cola | StructBERTRoBERTa ensemble | – | Accuracy: 69.2% |
| natural-language-inference-on-multinli | Adv-RoBERTa ensemble | – | Matched: 91.1Mismatched: 90.7 |
| natural-language-inference-on-qnli | StructBERTRoBERTa ensemble | – | Accuracy: 99.2% |
| natural-language-inference-on-rte | Adv-RoBERTa ensemble | – | Accuracy: 88.7% |
| natural-language-inference-on-wnli | StructBERTRoBERTa ensemble | – | Accuracy: 89.7 |
| paraphrase-identification-on-quora-question-pairs | StructBERTRoBERTa ensemble | – | F1: 74.4Accuracy: 90.7 |
| semantic-textual-similarity-on-mrpc | StructBERTRoBERTa ensemble | – | Accuracy: 91.5%F1: 93.6% |
| semantic-textual-similarity-on-sts-benchmark | StructBERTRoBERTa ensemble | – | Pearson Correlation: 0.928Spearman Correlation: 0.924 |
| sentiment-analysis-on-sst-2-binary-classification | StructBERTRoBERTa ensemble | – | Accuracy: 97.1 |