Mathematical Reasoning on AIME '25 (Pass@1, Pass@8)
31.25Pass@1GRPO w/ MAPR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPO w/ MAPRBase Model=Qwen3-14B2025.09 | 31.25 | 53.57 | |
| GRPO w/ MAPRBase Model=Qwen3-8B2025.09 | 28.44 | 45.96 | |
| GRPOBase Model=Qwen3-14B2025.09 | 27.92 | 45.56 | |
| GRPOBase Model=Qwen3-8B2025.09 | 22.19 | 38.74 | |
| GRPO w/ MAPRBase Model=Qwen3-4B2025.09 | 21.56 | 37.17 | |
| GRPOBase Model=Qwen3-4B2025.09 | 11.77 | 25.56 |