Fine-tuning on MMLU (val)
69.12MMLU STEM AccuracyFOAM
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| FOAMBackbone Model=Qwen2.5-7B2025.12 | 69.12 | 83.49 | 66.48 | 76.28 | 73.04 | |
| APOLLOBackbone Model=Qwen2.5-7B2025.12 | 68.49 | 82.97 | 65.8 | 76.65 | 72.65 | |
| GaLoreBackbone Model=Qwen2.5-7B2025.12 | 68.32 | 82.87 | 66.08 | 76.1 | 72.55 | |
| LoRABackbone Model=Qwen2.5-7B2025.12 | 67.69 | 82.91 | 66.29 | 75.72 | 72.41 | |
| LoRABackbone Model=LLaMA3.2-3B2025.12 | 47.22 | 63.41 | 49.56 | 61.57 | 54.86 | |
| FOAMBackbone Model=LLaMA3.2-3B2025.12 | 46.72 | 65.19 | 49.29 | 62.77 | 55.33 | |
| GaLoreBackbone Model=LLaMA3.2-3B2025.12 | 46.62 | 65.58 | 48.37 | 63.33 | 55.22 | |
| FullBackbone Model=LLaMA3.2-3B2025.12 | 46.55 | 65.81 | 49.16 | 63.17 | 55.48 | |
| APOLLOBackbone Model=LLaMA3.2-3B2025.12 | 46.16 | 65.91 | 49.03 | 62.8 | 55.29 | |
| FullBackbone Model=Gemma3-1B2025.12 | 27.63 | 26.84 | 25.12 | 25.14 | 26.04 | |
| APOLLOBackbone Model=Gemma3-1B2025.12 | 27.47 | 27.14 | 25.61 | 25.76 | 26.38 | |
| GaLoreBackbone Model=Gemma3-1B2025.12 | 27.4 | 26.71 | 24.8 | 25.69 | 25.99 | |
| LoRABackbone Model=Gemma3-1B2025.12 | 26.61 | 25.93 | 24.78 | 25.02 | 25.48 | |
| FOAMBackbone Model=Gemma3-1B2025.12 | 26.01 | 26.39 | 25.44 | 26.4 | 25.99 |