Language Understanding on MMLU (Overall Accuracy)
79.9MMLU AccuracySFT
Evaluation Results
| Method | Links | |
|---|---|---|
| SFTBackbone=Qwen2.5-32B2025.05 | 79.9 | |
| RawBackbone=Qwen2.5-32B2025.05 | 79.6 | |
| RawBackbone=Qwen2.5-7B2025.05 | 71.3 | |
| SFTBackbone=Qwen2.5-7B2025.05 | 71.3 | |
| GRPOBackbone=Qwen2.5-7B2025.05 | 71.3 | |
| SFTBackbone=Llama3.1-8B2025.05 | 61.4 | |
| GRPOBackbone=Llama3.1-8B2025.05 | 61.1 | |
| RawBackbone=Llama3.1-8B2025.05 | 60.4 | |
| RawBackbone=Llama3.2-3B2025.05 | 52.7 | |
| SFTBackbone=Llama3.2-3B2025.05 | 52.1 | |
| GRPOBackbone=Llama3.2-3B2025.05 | 52 |