Mathematical Reasoning on MATH (ACC/maj@8)
93.2ACC (maj@8)CDE
Evaluation Results
| Method | Links | |
|---|---|---|
| CDEBackbone=Qwen3-8B-Base2026.04 | 93.2 | |
| DiPOBackbone=Qwen3-8B-Base2026.04 | 92.6 | |
| GRPOBackbone=Qwen3-8B-Base2026.04 | 92.4 | |
| DAPOBackbone=Qwen3-8B-Base2026.04 | 92.4 | |
| DAPO w/ ELBackbone=Qwen3-8B-Base2026.04 | 92 | |
| DAPO w/ ELBackbone=Qwen3-4B-Base2026.04 | 90.4 | |
| DiPOBackbone=Qwen3-4B-Base2026.04 | 90.2 | |
| CDEBackbone=Qwen3-4B-Base2026.04 | 89.8 | |
| GRPOBackbone=Qwen3-4B-Base2026.04 | 89.6 | |
| DAPOBackbone=Qwen3-4B-Base2026.04 | 89.6 | |
| CDEBackbone=Qwen2.5-7B2026.04 | 85 | |
| GRPOBackbone=Qwen2.5-7B2026.04 | 84.4 | |
| DiPOBackbone=Qwen2.5-7B2026.04 | 84 | |
| DAPOBackbone=Qwen2.5-7B2026.04 | 83.4 | |
| DAPO w/ ELBackbone=Qwen2.5-7B2026.04 | 83.4 | |
| Base modelBackbone=Qwen3-8B-Base2026.04 | 80.2 | |
| Base modelBackbone=Qwen3-4B-Base2026.04 | 64 | |
| Base modelBackbone=Qwen2.5-7B2026.04 | 52.2 |