Mathematical Reasoning on AIME 2024 (Pass@1, Pass@8)
44.27Pass@1GRPO w/ MAPR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPO w/ MAPRBase Model=Qwen3-14B2025.09 | 44.27 | 68.3 | |
| GRPOBase Model=Qwen3-14B2025.09 | 38.54 | 58.55 | |
| GRPO w/ MAPRBase Model=Qwen3-8B2025.09 | 34.17 | 63.8 | |
| GRPOBase Model=Qwen3-8B2025.09 | 28.54 | 53.96 | |
| GRPO w/ MAPRBase Model=Qwen3-4B2025.09 | 26.15 | 48.82 | |
| GRPOBase Model=Qwen3-4B2025.09 | 17.5 | 33.6 |