| Benchmark | Model | Rank | Results |
|---|---|---|---|
| arithmetic-reasoning-on-gsm8k | U-PaLM | – | Accuracy: 58.5Parameters (Billion): 540 |
| cross-lingual-question-answering-on-tydiqa-goldp | U-PaLM 62B (fine-tuned) | – | EM: 78.4F1: 88.5 |
| cross-lingual-question-answering-on-tydiqa-goldp | U-PaLM-540B (CoT) | – | EM: 54.6 |
| multi-task-language-understanding-on-mgsm | U-PaLM 540B (CoT) | – | Average (%): 49.9 |
| question-answering-on-strategyqa | Minerva 540B | – | Accuracy: 61.9 |
| question-answering-on-strategyqa | PaLM 540B | – | Accuracy: 76.4 |
| question-answering-on-strategyqa | U-PaLM 540B | – | Accuracy: 76.6 |