Mathematical Reasoning on GSM8K (accuracy (G8))
91.7Accuracy (GSM8K)GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GRPOModel Size=7B2025.12 | 91.7 | |
| iCLPModel Size=7B2025.12 | 90.1 | |
| FT-SModel Size=7B2025.12 | 88.9 | |
| FT-EModel Size=7B2025.12 | 87.3 | |
| GRPOModel Size=3B2025.12 | 86.6 | |
| BaseModel Size=7B2025.12 | 85.4 | |
| iCLPModel Size=3B2025.12 | 85 | |
| FT-SModel Size=3B2025.12 | 83.2 | |
| FT-EModel Size=3B2025.12 | 81.4 | |
| BaseModel Size=3B2025.12 | 79.1 | |
| GRPOModel Size=0.5B2025.12 | 54.5 | |
| iCLPModel Size=0.5B2025.12 | 51.5 | |
| FT-SModel Size=0.5B2025.12 | 48 | |
| FT-EModel Size=0.5B2025.12 | 43.8 | |
| BaseModel Size=0.5B2025.12 | 41.6 | |
| FT-LEModel Size=0.5B2025.12 | 38.3 |