General Language Understanding on MMLU, AlpacaEval, Arena-Hard
73.41MMLU AccuracyQwen2.5-7B + DataFlow-Chat-15K
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-7B + DataFlow-Chat-15KBase Model=Qwen2.5-7B, SFT Data Source=DataFlow-Chat-15K2025.12 | 73.41 | 10.11 | 110 | 28.21 | |
| Qwen2.5-7B + ShareGPT-15KBase Model=Qwen2.5-7B, SFT Data Source=ShareGPT-15K2025.12 | 73.09 | 3.7 | 130 | 26.03 | |
| Qwen2.5-7B + UltraChat-15KBase Model=Qwen2.5-7B, SFT Data Source=UltraChat-15K2025.12 | 72.97 | 3.97 | 80 | 25.91 | |
| Qwen2.5-7BBase Model=Qwen2.5-7B2025.12 | 71.45 | 7.05 | 60 | 26.36 |