| text-classification-on-arxiv-10 | RoBERTa | #2 | Accuracy: 0.779 |
| natural-language-inference-on-anli-test | RoBERTa (Large) | #3 | A1: 72.4A2: 49.8A3: 44.4 |
| semantic-textual-similarity-on-mrpc | RoBERTa (ensemble) | #3 | Accuracy: 92.3% |
| semantic-textual-similarity-on-sts-b-glue-dev | RoBERTa | #3 | Pearson/Spearman Correlation: 92.4 |
| natural-language-inference-on-qnli | RoBERTa (ensemble) | #4 | Accuracy: 98.9% |
| paraphrase-identification-on-mrpc-glue-dev | RoBERTa | #4 | F1: 90.9 |
| natural-language-understanding-on-glue-test | RoBERTa | #5 | GLUE (test): 88.5 |
| question-answering-on-quora-question-pairs | RoBERTa (ensemble) | #5 | Accuracy: 90.2% |
| semantic-textual-similarity-on-sts-benchmark | RoBERTa | #5 | Pearson Correlation: 0.922 |
| type-prediction-on-manytypes4typescript | RoBERTa | #5 | Average Accuracy: 59.84Average Precision: 57.45Average Recall: 57.62… |
| natural-language-inference-on-multinli-dev | RoBERTa | #6 | Matched: 90.2 |
| natural-language-inference-on-rte-glue-dev | RoBERTa | #6 | Accuracy: 86.6 |
| question-answering-on-squad20-dev | RoBERTa (no data aug) | #6 | F1: 89.4EM: 86.5 |
| linguistic-acceptability-on-cola-glue-dev | RoBERTa | #7 | Matthews Correlation: 68.0 |
| natural-language-inference-on-multinli | RoBERTa | #7 | Matched: 90.8 |
| natural-language-inference-on-multinli | RoBERTa (ensemble) | #50 | Mismatched: 90.2 |
| natural-language-inference-on-qnli-glue-dev | RoBERTa | #7 | Accuracy: 94.7 |
| natural-language-inference-on-wnli | RoBERTa (ensemble) | #8 | Accuracy: 89 |
| reading-comprehension-on-race | RoBERTa | #9 | Accuracy: 83.2Accuracy (Middle): 86.5Accuracy (High): 81.3 |
| sentiment-analysis-on-sst-2-binary-classification | RoBERTa (ensemble) | #11 | Accuracy: 96.7 |
| stock-market-prediction-on-astock | RoBERTa WWM Ext (News+Factors) | #11 | Accuray: 62.49F1-score: 62.54Recall: 62.51Precision: 62.59 |
| stock-market-prediction-on-astock | RoBERTa WWM Ext (News) | #12 | Accuray: 61.34F1-score: 61.48Recall: 61.32Precision: 61.97 |
| question-answering-on-social-iqa | RoBERTa-Large 355M (fine-tuned) | #12 | Accuracy: 76.7 |
| question-answering-on-squad20 | RoBERTa (single model) | #12 | EM: 86.820F1: 89.795 |
| linguistic-acceptability-on-cola | RoBERTa (ensemble) | #16 | Accuracy: 67.8% |
| common-sense-reasoning-on-commonsenseqa | RoBERTa-Large 355M | #17 | Accuracy: 72.1 |
| natural-language-inference-on-rte | RoBERTa | #17 | Accuracy: 88.2% |
| natural-language-inference-on-rte | RoBERTa (ensemble) | #18 | Accuracy: 88.2% |
| task-1-grouping-on-ocw | RoBERTa (LARGE) | #22 | # Correct Groups: 29 ± 3Fowlkes Mallows Score (FMS): 26.7 ± .2… |
| document-image-classification-on-rvl-cdip | Roberta base | #27 | Accuracy: 90.06Parameters: 125M |
| paraphrase-identification-on-quora-question-pairs | RoBERTa | #30 | Accuracy: 90.2 |
| sentence-completion-on-hellaswag | RoBERTa-Large Ensemble | #35 | Accuracy: 85.5 |
| sentence-completion-on-hellaswag | RoBERTa-Large 355M | #62 | Accuracy: 81.7 |
| question-answering-on-squad11-dev | RoBERTa | #47 | F1: 94.6 |
| question-answering-on-piqa | RoBERTa-Large 355M | #67 | Accuracy: 79.4 |