Mathematical Reasoning on MATH500 (Pass@1, Pass@8)
89.93Pass@1GRPO w/ MAPR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPO w/ MAPRBase Model=Qwen3-14B2025.09 | 89.93 | 96.39 | |
| GRPOBase Model=Qwen3-14B2025.09 | 88.73 | 96.02 | |
| GRPO w/ MAPRBase Model=Qwen3-8B2025.09 | 88.05 | 95.35 | |
| GRPOBase Model=Qwen3-8B2025.09 | 85.75 | 94.31 | |
| GRPO w/ MAPRBase Model=Qwen3-4B2025.09 | 84.52 | 93.74 | |
| GRPOBase Model=Qwen3-4B2025.09 | 79.61 | 90.12 |