Mathematical Reasoning on AIME 2024 (Accuracy, CLC)
67.52AccuracySOLAR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SOLARBackbone=Qwen3-8B, Fine-tuning strategy=SOLAR2026.06 | 67.52 | 60.61 | |
| SFTBackbone=Qwen3-8B, Fine-tuning strategy=SFT2026.06 | 65.67 | 58.67 | |
| Soft ThinkingBackbone=Qwen3-8B, Fine-tuning strategy=Soft Thinking2026.06 | 62.52 | 56.91 | |
| Qwen3-8BBackbone=Qwen3-8B, Fine-tuning strategy=Base2026.06 | 62.1 | 56.06 | |
| SOLARBackbone=Qwen3-4B, Fine-tuning strategy=SOLAR2026.06 | 38.52 | 29.67 | |
| MidAlignBackbone=Qwen3-4B, Fine-tuning strategy=MidAlign2026.06 | 36.67 | 27.11 | |
| AlignXBackbone=Qwen3-4B, Fine-tuning strategy=AlignX2026.06 | 36 | 26.88 | |
| SFTBackbone=Qwen3-4B, Fine-tuning strategy=SFT2026.06 | 34.76 | 25.17 | |
| MAPOBackbone=Qwen3-4B, Fine-tuning strategy=MAPO2026.06 | 34.76 | 25.89 | |
| Soft ThinkingBackbone=Qwen3-4B, Fine-tuning strategy=Soft Thinking2026.06 | 31.86 | 24.44 | |
| Qwen3-4BBackbone=Qwen3-4B, Fine-tuning strategy=Base2026.06 | 31.1 | 23.89 |