| natural-language-inference-on-qnli | ALBERT | #1 | Accuracy: 99.2% |
| semantic-textual-similarity-on-mrpc | ALBERT | #2 | Accuracy: 93.4% |
| natural-language-understanding-on-glue-test | ALBERT | #3 | GLUE (test): 89.4 |
| question-answering-on-quora-question-pairs | ALBERT | #3 | Accuracy: 90.5% |
| question-answering-on-squad20 | ALBERT (ensemble model) | #3 | EM: 89.731F1: 92.215 |
| question-answering-on-squad20 | ALBERT (single model) | #7 | EM: 88.107F1: 90.902 |
| semantic-textual-similarity-on-sts-benchmark | ALBERT | #3 | Pearson Correlation: 0.925 |
| natural-language-inference-on-multinli | ALBERT | #4 | Matched: 91.3 |
| reading-comprehension-on-race | ALBERT ensemble | #5 | Accuracy: 89.4 |
| multimodal-intent-recognition-on-photochat | ALBERT-base | #6 | F1: 52.2Precision: 44.8Recall: 62.7 |
| natural-language-inference-on-wnli | ALBERT | #6 | Accuracy: 91.8 |
| sentiment-analysis-on-sst-2-binary-classification | ALBERT | #6 | Accuracy: 97.1 |
| question-answering-on-squad20-dev | ALBERT xxlarge | #8 | F1: 88.1EM: 85.1 |
| question-answering-on-squad20-dev | ALBERT xlarge | #12 | F1: 85.9EM: 83.1 |
| question-answering-on-squad20-dev | ALBERT large | #14 | F1: 82.1EM: 79.0 |
| question-answering-on-squad20-dev | ALBERT base | #16 | F1: 79.1EM: 76.1 |
| question-answering-on-multitq | ALBERT | #10 | Hits@1: 10.8Hits@10: 45.9 |
| common-sense-reasoning-on-commonsenseqa | Albert Lan et al. (2020) (ensemble) | #11 | Accuracy: 76.5 |
| linguistic-acceptability-on-cola | ALBERT | #11 | Accuracy: 69.1% |
| natural-language-inference-on-rte | ALBERT | #16 | Accuracy: 89.2% |
| paraphrase-identification-on-quora-question-pairs | ALBERT | #31 | Accuracy: 90.5 |
| question-answering-on-squad11-dev | ALBERT ensemble | #46 | F1: 95.5 |