| Benchmark | Model | Rank | Results |
|---|---|---|---|
| question-answering-on-piqa | LoRA-Sparse (LLaMA-7B) (extra data) | – | Accuracy: 79.4 |
| reading-comprehension-on-race | LoRA-Sparse (LLaMA-7B) (extra data) | – | Accuracy: 42.5 |
| sentence-completion-on-hellaswag | LoRA-Sparse (LLaMA-7B) (extra data) | – | Accuracy: 60.1 |
| visual-question-answering-vqa-on-seed-seedbench | LoRA-Sparse (Vicuna-7B) (extra data) | – | Accuracy: 58.8 |