Mathematical Reasoning on AIME 2024 (pass@1, pass@256)
10Pass@1Qwen2.5-7B-Instruct
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-7B-InstructTraining Mode=Vanilla RLVR2026.06 | 10 | 50 | |
| Qwen2.5-7B-InstructTraining Mode=Curriculum RL2026.06 | 10 | 60 | |
| Qwen2.5-7BTraining Mode=Curriculum RL2026.06 | 10 | 52.2 | |
| LLaMA-3.1-8B-InstructTraining Mode=Curriculum RL2026.06 | 6.7 | 60 | |
| Qwen2.5-7B-InstructTraining Mode=Base2026.06 | 3.3 | 46.7 | |
| Qwen3-8BTraining Mode=Vanilla RLVR2026.06 | 3.3 | 36.7 | |
| Qwen3-8BTraining Mode=Curriculum RL2026.06 | 3.3 | 50 | |
| LLaMA-3.1-8B-InstructTraining Mode=Vanilla RLVR2026.06 | 3.3 | 36.7 | |
| Qwen2.5-7BTraining Mode=Base2026.06 | 1.1 | 40 | |
| Qwen2.5-7BTraining Mode=Vanilla RLVR2026.06 | 0 | 26.7 | |
| Qwen3-8BTraining Mode=Base2026.06 | 0 | 23.3 | |
| LLaMA-3.1-8B-InstructTraining Mode=Base2026.06 | 0 | 53.3 | |
| DeepSeek-R1-0528-Qwen3-8BTraining Mode=Base2026.06 | 0 | 16.7 | |
| DeepSeek-R1-0528-Qwen3-8BTraining Mode=Vanilla RLVR2026.06 | 0 | 13.3 | |
| DeepSeek-R1-0528-Qwen3-8BTraining Mode=Curriculum RL2026.06 | 0 | 20 |