| question-answering-on-piqa | UL2 20B | #1 | Accuracy: 90.7 |
| coreference-resolution-on-winograd-schema-challenge | UL2 20B (fine-tuned) | #2 | Accuracy: 98.1 |
| coreference-resolution-on-winograd-schema-challenge | UL2 20B (0-shot) | #20 | Accuracy: 79.9 |
| reading-comprehension-on-race | UL2 20B | #2 | Accuracy: 90.9 |
| question-answering-on-copa | UL2 20B (fine-tuned) | #3 | Accuracy: 99 |
| question-answering-on-copa | UL2 20B (0-shot) | #26 | Accuracy: 85 |
| common-sense-reasoning-on-winogrande | UL2 20B | #4 | Accuracy: 90.1 |
| word-sense-disambiguation-on-words-in-context | UL2 20B (fine-tuned) | #5 | Accuracy: 77.3 |
| word-sense-disambiguation-on-words-in-context | UL2 20B (0-shot) | #23 | Accuracy: 49.8 |
| long-range-modeling-on-scrolls | UL2 | #6 | Avg.: 37.87GovRep: 53.6 / 26.1 / 28.8SumScr: 32.9 / 7.8 / 19.4… |
| long-range-modeling-on-scrolls | UL2 20B | #12 | CNLI: 88.7 |
| question-answering-on-boolq | UL2 20B (fine-tuned) | #6 | Accuracy: 90.8 |
| question-answering-on-boolq | UL2 20B (0-shot) | #83 | Accuracy: 63.1 |
| semantic-parsing-on-wikisql | UL2 20B | #6 | Denotation accuracy (test): 87.3 |
| sentence-completion-on-hellaswag | UL2 20B | #8 | Accuracy: 94.1 |
| natural-language-inference-on-rte | UL2 20B (fine-tuned) | #9 | Accuracy: 92.1% |
| natural-language-inference-on-rte | UL2 20B (0-shot) | #66 | Accuracy: 60.7% |
| common-sense-reasoning-on-arc-challenge | UL2 20B (chain-of-thought + self-consistency) | #26 | Accuracy: 49.5 |
| common-sense-reasoning-on-arc-challenge | UL2 20B (chain-of-thought) | #33 | Accuracy: 42.9 |
| common-sense-reasoning-on-arc-challenge | UL2 20B (zero-shot) | #41 | Accuracy: 29.8 |
| common-sense-reasoning-on-commonsenseqa | UL2 20B (chain-of-thought + self-consistency) | #29 | Accuracy: 55.7 |
| common-sense-reasoning-on-commonsenseqa | UL2 20B (chain-of-thought) | #31 | Accuracy: 51.4 |
| common-sense-reasoning-on-commonsenseqa | UL2 20B (zero-shot) | #33 | Accuracy: 34.2 |
| common-sense-reasoning-on-arc-easy | UL2 20B (chain-of-thought + self-consistency) | #59 | Accuracy: 69.8 |
| common-sense-reasoning-on-arc-easy | UL2 20B (chain-of-thought) | #94 | Accuracy: 38.4 |
| common-sense-reasoning-on-arc-easy | UL2 20B (0-shot) | #96 | Accuracy: 32.2 |
| multi-task-language-understanding-on-mmlu | UL2 20B (5-shot) | #73 | Average (%): 39.2 |
| arithmetic-reasoning-on-gsm8k | UL2 20B (chain-of-thought) | #275 | Accuracy: 4.4Parameters (Billion): 20 |
| arithmetic-reasoning-on-gsm8k | UL2 20B (0-shot) | #277 | Accuracy: 4.1Parameters (Billion): 20 |