Mathematical Reasoning on GSM8K (HS & FA)
78.4HSNon-Aligned
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Non-AlignedBackbone=Llama2-7B2024.02 | 78.4 | 27.8 | |
| SFTBackbone=Llama2-7B2024.02 | 68.4 | 23.4 | |
| VaccineBackbone=Llama2-7B2024.02 | 65 | 22.4 | |
| EWCBackbone=Llama2-7B2024.02 | 51.4 | 5.8 |