Mathematical Reasoning on MATH-500 (held-out)
89.4Pass@1 ScoreHSD
Evaluation Results
| Method | Links | |
|---|---|---|
| HSDBase Model=Qwen3-32B-Base2026.06 | 89.4 | |
| GRPO + OPSDBase Model=Qwen3-32B-Base2026.06 | 87.9 | |
| RLSDBase Model=Qwen3-32B-Base2026.06 | 87.7 | |
| SDPOBase Model=Qwen3-32B-Base2026.06 | 87.5 | |
| OPSDBase Model=Qwen3-32B-Base2026.06 | 87 | |
| DAPOBase Model=Qwen3-32B-Base2026.06 | 86.8 | |
| GSPOBase Model=Qwen3-32B-Base2026.06 | 86.4 | |
| Dr. GRPOBase Model=Qwen3-32B-Base2026.06 | 85.9 | |
| GRPOBase Model=Qwen3-32B-Base2026.06 | 85.2 | |
| HSDBase Model=Qwen3-8B-Base2026.06 | 83.6 | |
| GRPO + OPSDBase Model=Qwen3-8B-Base2026.06 | 81.9 | |
| RLSDBase Model=Qwen3-8B-Base2026.06 | 81.6 | |
| SDPOBase Model=Qwen3-8B-Base2026.06 | 81.4 | |
| OPSDBase Model=Qwen3-8B-Base2026.06 | 80.9 | |
| DAPOBase Model=Qwen3-8B-Base2026.06 | 80.1 | |
| GSPOBase Model=Qwen3-8B-Base2026.06 | 79.7 | |
| Dr. GRPOBase Model=Qwen3-8B-Base2026.06 | 79.2 | |
| GRPOBase Model=Qwen3-8B-Base2026.06 | 78.4 |