| question-answering-on-triviaqa | Branch-Train-MiX 4x7B (sampling top-2 experts) | #30 | EM: 57.1 |
| multi-task-language-understanding-on-mmlu | Branch-Train-MiX 4x7B (sampling top-1 experts) | #60 | Average (%): 53.2 |
| common-sense-reasoning-on-winogrande | Branch-Train-MiX 4x7B (sampling top-1 expert) | #76 | Accuracy: 70.6 |
| code-generation-on-mbpp | Branch-Train-Merge 4x7B (top-2) | #130 | Accuracy: 42.6 |
| code-generation-on-mbpp | Branch-Train-MiX 4x7B (sampling top-2 experts) | #137 | Accuracy: 39.4 |
| math-word-problem-solving-on-math | Branch-Train-MiX 4x7B (sampling top-2 experts) | #164 | Accuracy: 17.8 |
| arithmetic-reasoning-on-gsm8k | Branch-Train-MiX 4x7B (sampling top-2 experts) | #250 | Accuracy: 37.1 |