Mathematical Reasoning on AIME 25 (Accuracy, Performance Gain)
73.33AccuracyQwen3-32B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-32BSetting=Recur2026.06 | 73.33 | 6 | |
| Qwen3-14BSetting=Recur2026.06 | 71.66 | 5.91 | |
| Qwen3-32BSetting=Base2026.06 | 69.17 | — | |
| Qwen3-14BSetting=Base2026.06 | 67.66 | — | |
| Qwen3-8BSetting=Recur2026.06 | 61.66 | 2.77 | |
| Qwen3-4BSetting=Recur2026.06 | 60 | 5.97 | |
| Qwen3-8BSetting=Base2026.06 | 60 | — | |
| LLaMA-70BSetting=Recur2026.06 | 58.33 | 4.44 | |
| Qwen3-4BSetting=Base2026.06 | 56.66 | — | |
| LLaMA-70BSetting=Base2026.06 | 55.83 | — | |
| Qwen3-1.7BSetting=Recur2026.06 | 40 | 29.76 | |
| LLaMA-8BSetting=Recur2026.06 | 36.67 | 10.04 | |
| LLaMA-8BSetting=Base2026.06 | 33.33 | — | |
| Qwen3-1.7BSetting=Base2026.06 | 30.83 | — |