Mathematical Reasoning on AIME24 (ACC, mean@8)
35AccuracyDiPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DiPOBackbone=Qwen3-8B-Base2026.04 | 35 | 35 | |
| DAPO w/ ELBackbone=Qwen3-8B-Base2026.04 | 33.75 | 33.75 | |
| GRPOBackbone=Qwen3-8B-Base2026.04 | 31.67 | 31.67 | |
| CDEBackbone=Qwen3-8B-Base2026.04 | 31.67 | 31.67 | |
| DAPOBackbone=Qwen3-8B-Base2026.04 | 30.08 | 30.08 | |
| DiPOBackbone=Qwen3-4B-Base2026.04 | 29.17 | 29.17 | |
| GRPOBackbone=Qwen3-4B-Base2026.04 | 26.67 | 26.67 | |
| DAPO w/ ELBackbone=Qwen3-4B-Base2026.04 | 26.67 | 26.67 | |
| CDEBackbone=Qwen3-4B-Base2026.04 | 26.67 | 26.67 | |
| DAPOBackbone=Qwen3-4B-Base2026.04 | 26.25 | 26.25 | |
| DiPOBackbone=Qwen2.5-7B2026.04 | 22.92 | 22.92 | |
| GRPOBackbone=Qwen2.5-7B2026.04 | 20.42 | 20.42 | |
| DAPOBackbone=Qwen2.5-7B2026.04 | 20.42 | 20.42 | |
| DAPO w/ ELBackbone=Qwen2.5-7B2026.04 | 20 | 20 | |
| CDEBackbone=Qwen2.5-7B2026.04 | 20 | 20 | |
| Base modelBackbone=Qwen3-4B-Base2026.04 | 9.58 | 9.58 | |
| Base modelBackbone=Qwen3-8B-Base2026.04 | 8.33 | 8.33 | |
| Base modelBackbone=Qwen2.5-7B2026.04 | 7.08 | 7.08 |