Mathematical Reasoning on AMC23 (ACC and mean@8)
71.23Accuracy (ACC)DiPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DiPOBackbone=Qwen3-8B-Base2026.04 | 71.23 | 71.23 | |
| DAPO w/ ELBackbone=Qwen3-8B-Base2026.04 | 69.87 | 69.87 | |
| GRPOBackbone=Qwen3-8B-Base2026.04 | 69.28 | 69.28 | |
| DAPOBackbone=Qwen3-8B-Base2026.04 | 69.12 | 69.12 | |
| CDEBackbone=Qwen3-8B-Base2026.04 | 68.07 | 68.07 | |
| DiPOBackbone=Qwen3-4B-Base2026.04 | 63.7 | 63.7 | |
| DAPO w/ ELBackbone=Qwen3-4B-Base2026.04 | 62.95 | 62.95 | |
| CDEBackbone=Qwen3-4B-Base2026.04 | 62.35 | 62.35 | |
| DAPOBackbone=Qwen3-4B-Base2026.04 | 61.9 | 61.9 | |
| GRPOBackbone=Qwen3-4B-Base2026.04 | 60.24 | 60.24 | |
| DiPOBackbone=Qwen2.5-7B2026.04 | 60.09 | 60.09 | |
| DAPOBackbone=Qwen2.5-7B2026.04 | 59.94 | 59.94 | |
| DAPO w/ ELBackbone=Qwen2.5-7B2026.04 | 58.73 | 58.73 | |
| GRPOBackbone=Qwen2.5-7B2026.04 | 58.43 | 58.43 | |
| CDEBackbone=Qwen2.5-7B2026.04 | 55.87 | 55.87 | |
| Base modelBackbone=Qwen3-8B-Base2026.04 | 39.46 | 39.46 | |
| Base modelBackbone=Qwen3-4B-Base2026.04 | 31.48 | 31.48 | |
| Base modelBackbone=Qwen2.5-7B2026.04 | 22.74 | 22.74 |