| natural-language-inference-on-multinli-dev | Megatron-3.9B | #1 | Matched: 91.4 |
| question-answering-on-squad20-dev | Megatron-3.9B | #1 | F1: 91.2 |
| reading-comprehension-on-race | Megatron-BERT (ensemble) | #1 | Accuracy: 90.9Accuracy (Middle): 93.1Accuracy (High): 90.0 |
| reading-comprehension-on-race | Megatron-BERT | #4 | Accuracy: 89.5Accuracy (Middle): 91.8Accuracy (High): 88.6 |
| language-modelling-on-wikitext-103 | Megatron-LM | #3 | Test perplexity: 10.81Number of params: 8300M |
| language-modelling-on-lambada | Megatron-LM GPT-2 8.3B | #32 | Accuracy: 66.51 |
| question-answering-on-piqa | MT-NLG 530B (0-shot) | #39 | Accuracy: 82.0 |
| question-answering-on-squad11-dev | Megatron-3.9B | #52 | F1: 95.5 |