Multiple-choice science question answering on MMLU Physics Larger Dist. Shift 37, 36
32.64AccuracyBLoB
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BLoBBackbone=Llama2-7B, PEFT Method=LoRA, Number of samples (N)=52024.06 | 32.64 | 18.15 | 1.48 | |
| BLoBBackbone=Llama2-7B, PEFT Method=LoRA, Number of samples (N)=02024.06 | 31.94 | 22.9 | 1.53 | |
| LAPBackbone=Llama2-7B, PEFT Method=LoRA2024.06 | 31.6 | 13.91 | 1.36 | |
| BLoBBackbone=Llama2-7B, PEFT Method=LoRA, Number of samples (N)=102024.06 | 31.6 | 18.29 | 1.46 | |
| ENSBackbone=Llama2-7B, PEFT Method=LoRA2024.06 | 30.56 | 17.52 | 1.39 | |
| BBBBackbone=Llama2-7B, PEFT Method=LoRA2024.06 | 30.21 | 26.1 | 1.62 | |
| MCDBackbone=Llama2-7B, PEFT Method=LoRA2024.06 | 29.17 | 29.53 | 1.67 | |
| MLEBackbone=Llama2-7B, PEFT Method=LoRA2024.06 | 28.82 | 29.6 | 1.66 | |
| MAPBackbone=Llama2-7B, PEFT Method=LoRA2024.06 | 28.47 | 30.25 | 1.65 |