Mathematical Reasoning on GSM8K (GSM8K Score)
90GSM8K ScoreGRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GRPOBackbone=LLaMA-3.1-8B2025.05 | 90 | |
| TI-DPOBackbone=LLaMA-3.1-8B2025.05 | 89 | |
| TPOBackbone=LLaMA-3.1-8B2025.05 | 88 | |
| CPOBackbone=LLaMA-3.1-8B2025.05 | 86.5 | |
| KTOBackbone=LLaMA-3.1-8B2025.05 | 86 | |
| DPOBackbone=LLaMA-3.1-8B2025.05 | 85 | |
| SFTBackbone=LLaMA-3.1-8B2025.05 | 84 | |
| TDPOBackbone=LLaMA-3.1-8B2025.05 | 83 | |
| IPOBackbone=LLaMA-3.1-8B2025.05 | 80 | |
| SIMPOBackbone=LLaMA-3.1-8B2025.05 | 78 |