Mathematical Reasoning on DM (test)
70.1AccuracyDeepSeekMath-CRPS-60K
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepSeekMath-CRPS-60KBackbone=DeepSeekMath-7B, #Samples=60K2026.04 | 70.1 | |
| DeepSeekMath-CRPS-30KBackbone=DeepSeekMath-7B, #Samples=30K2026.04 | 68.5 | |
| SIGMA-60KBackbone=DeepSeekMath-7B, #Samples=60K2026.04 | 68.4 | |
| SIGMA-30KBackbone=DeepSeekMath-7B, #Samples=30K2026.04 | 67.2 | |
| DeepSeekMath-CRPS-15KBackbone=DeepSeekMath-7B, #Samples=15K2026.04 | 65.8 | |
| MathFusionBackbone=DeepSeekMath-7B, #Samples=60K2026.04 | 65.8 | |
| MathFusion (Sequential)Backbone=DeepSeekMath-7B, #Samples=30K2026.04 | 64.6 | |
| DeepSeekMath-DARTBackbone=DeepSeekMath-7B, #Samples=60K2026.04 | 62.8 | |
| DeepSeekMath-DARTBackbone=DeepSeekMath-7B, #Samples=590K2026.04 | 61.6 | |
| DeepSeekMath-RFTBackbone=DeepSeekMath-7B, #Samples=590K2026.04 | 60.2 | |
| DeepSeekMath-MMIQCBackbone=DeepSeekMath-7B, #Samples=2.3M2026.04 | 52.9 | |
| DeepSeekMath-InstructBackbone=DeepSeekMath-7B, #Samples=780K2026.04 | 52.2 | |
| DeepSeekMath-MetaMathBackbone=DeepSeekMath-7B, #Samples=60K2026.04 | 45.9 |