Mathematical Reasoning on AIME25 (ACC/maj@8)
33.33ACC/maj@8GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GRPOBackbone=Qwen3-8B-Base2026.04 | 33.33 | |
| DAPOBackbone=Qwen3-8B-Base2026.04 | 33.33 | |
| DAPO w/ ELBackbone=Qwen3-8B-Base2026.04 | 33.33 | |
| CDEBackbone=Qwen3-8B-Base2026.04 | 33.33 | |
| DiPOBackbone=Qwen3-8B-Base2026.04 | 33.33 | |
| DAPOBackbone=Qwen3-4B-Base2026.04 | 30 | |
| DAPO w/ ELBackbone=Qwen3-4B-Base2026.04 | 30 | |
| CDEBackbone=Qwen3-4B-Base2026.04 | 30 | |
| DiPOBackbone=Qwen3-4B-Base2026.04 | 30 | |
| GRPOBackbone=Qwen3-4B-Base2026.04 | 26.67 | |
| GRPOBackbone=Qwen2.5-7B2026.04 | 26.67 | |
| DAPOBackbone=Qwen2.5-7B2026.04 | 23.33 | |
| DiPOBackbone=Qwen2.5-7B2026.04 | 23.33 | |
| DAPO w/ ELBackbone=Qwen2.5-7B2026.04 | 20 | |
| CDEBackbone=Qwen2.5-7B2026.04 | 20 | |
| Base modelBackbone=Qwen3-8B-Base2026.04 | 13.33 | |
| Base modelBackbone=Qwen3-4B-Base2026.04 | 6.67 | |
| Base modelBackbone=Qwen2.5-7B2026.04 | 3.33 |