Multilingual Mathematical Reasoning on MGSM (Accuracy, CLC)
88.27AccuracySOLAR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SOLARBackbone=Qwen3-8B, Fine-tuning strategy=SOLAR2026.06 | 88.27 | 87.28 | |
| SFTBackbone=Qwen3-8B, Fine-tuning strategy=SFT2026.06 | 86.9 | 84.71 | |
| Qwen3-8BBackbone=Qwen3-8B, Fine-tuning strategy=Base2026.06 | 83.42 | 80.51 | |
| Soft ThinkingBackbone=Qwen3-8B, Fine-tuning strategy=Soft Thinking2026.06 | 83.04 | 80.01 | |
| SOLARBackbone=Qwen3-4B, Fine-tuning strategy=SOLAR2026.06 | 78.27 | 75.61 | |
| AlignXBackbone=Qwen3-4B, Fine-tuning strategy=AlignX2026.06 | 77.69 | 73 | |
| MAPOBackbone=Qwen3-4B, Fine-tuning strategy=MAPO2026.06 | 77.52 | 73.9 | |
| MidAlignBackbone=Qwen3-4B, Fine-tuning strategy=MidAlign2026.06 | 76.45 | 73.11 | |
| SFTBackbone=Qwen3-4B, Fine-tuning strategy=SFT2026.06 | 76.25 | 72.29 | |
| Soft ThinkingBackbone=Qwen3-4B, Fine-tuning strategy=Soft Thinking2026.06 | 74.23 | 70.37 | |
| Qwen3-4BBackbone=Qwen3-4B, Fine-tuning strategy=Base2026.06 | 74.01 | 69.31 |