Multiple-choice Reasoning on Arc-bo
48.39AccuracyOurs-SFT
Evaluation Results
| Method | Links | |
|---|---|---|
| Ours-SFTalignment=SFT2025.07 | 48.39 | |
| Ours-Base-32kcontext-length=32k2025.07 | 44.62 | |
| Ours-Base2025.07 | 44.32 | |
| Tibetan-Alpaca-7B2025.07 | 27.94 | |
| Tibetan-Llama2-7B2025.07 | 27.54 | |
| Yak-Llama2-7B2025.07 | 25.64 |