Mathematical Reasoning on AIME Think (test)
72.4AccuracyQwen3-8B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-8BTraining=RL, Mode=Mid-Think2026.01 | 72.4 | 15,318 | 44,142 | |
| Qwen3-8BTraining=RL, Mode=No-Think2026.01 | 72 | 18,980 | 34,585 | |
| Qwen3-8BTraining=No, Mode=Untrained2026.01 | 71.6 | 16,847 | 40,342 | |
| Qwen3-8BTraining=RL, Mode=Think2026.01 | 69.8 | 13,330 | 34,701 | |
| Qwen3-4BTraining=RL, Mode=Mid-Think2026.01 | 69.6 | 15,799 | 40,216 | |
| Qwen3-4BTraining=RL, Mode=No-Think2026.01 | 69.3 | 20,551 | 61,341 | |
| Qwen3-4BTraining=No, Mode=Untrained2026.01 | 68.7 | 16,699 | 40,089 | |
| Qwen3-4BTraining=RL, Mode=Think2026.01 | 61.1 | 13,347 | 32,118 |