Mathematical Reasoning on GSM8K (Accuracy, Peak Memory)
39.37Accuracy (%)Adam
Evaluation results
| Method | Links | ||
|---|---|---|---|
| AdamModel=Llama-2 7B, Rank=–2026.04 | 39.37 | 95.18 | |
| OASISModel=Llama-2 7B, Rank=322026.04 | 39.14 | 48.21 | |
| OASISModel=Llama-2 7B, Rank=1282026.04 | 39.1 | 49.6 | |
| LDAdamModel=Llama-2 7B, Rank=322026.04 | 38.74 | 71.48 | |
| LDAdamModel=Llama-2 7B, Rank=1282026.04 | 38.17 | 72.21 | |
| GaLoreModel=Llama-2 7B, Rank=1282026.04 | 37.49 | 72.03 | |
| LoRAModel=Llama-2 7B, Rank=1282026.04 | 37.27 | 66 | |
| LoRAModel=Llama-2 7B, Rank=322026.04 | 37.13 | 64.15 | |
| GaLoreModel=Llama-2 7B, Rank=322026.04 | 35.96 | 71.3 | |
| AdamModel=Llama-3 1B, Rank=–2026.04 | 27.09 | 52.54 | |
| OASISModel=Llama-3 1B, Rank=1282026.04 | 26.88 | 41.59 | |
| LoRAModel=Llama-3 1B, Rank=1282026.04 | 26.61 | 48.62 | |
| LDAdamModel=Llama-3 1B, Rank=1282026.04 | 25.72 | 49.23 | |
| GaLoreModel=Llama-3 1B, Rank=1282026.04 | 25.51 | 47.4 | |
| LDAdamModel=Llama-3 1B, Rank=322026.04 | 24.55 | 49.01 | |
| OASISModel=Llama-3 1B, Rank=322026.04 | 23.78 | 41.03 | |
| LoRAModel=Llama-3 1B, Rank=322026.04 | 22.51 | 48.06 | |
| GaLoreModel=Llama-3 1B, Rank=322026.04 | 21.64 | 47.17 |