Mathematical Reasoning on MATH (test) (EM)
48.67Exact Match (EM)MetaTuner-J
Evaluation Results
| Method | Links | |
|---|---|---|
| MetaTuner-JBackbone=Qwen2.5-7B, Method Category=Hybrid2025.09 | 48.67 | |
| MetaTuner-IBackbone=Qwen2.5-7B, Method Category=Hybrid2025.09 | 48.22 | |
| DPOBackbone=Qwen2.5-7B, Method Category=Fine-Tuning2025.09 | 43.78 | |
| PPOBackbone=Qwen2.5-7B, Method Category=Fine-Tuning2025.09 | 41.78 | |
| KTOBackbone=Qwen2.5-7B, Method Category=Fine-Tuning2025.09 | 41.78 | |
| BetterTogetherBackbone=Qwen2.5-7B, Method Category=Hybrid2025.09 | 41.56 | |
| SFTBackbone=Qwen2.5-7B, Method Category=Fine-Tuning2025.09 | 41.33 | |
| MetaTuner-JBackbone=Qwen2.5-3B, Method Category=Hybrid2025.09 | 41.33 | |
| MetaTuner-IBackbone=Qwen2.5-3B, Method Category=Hybrid2025.09 | 40.89 | |
| DPOBackbone=Qwen2.5-3B, Method Category=Fine-Tuning2025.09 | 36.59 | |
| BetterTogetherBackbone=Qwen2.5-3B, Method Category=Hybrid2025.09 | 35.78 | |
| PPOBackbone=Qwen2.5-3B, Method Category=Fine-Tuning2025.09 | 34.22 | |
| KTOBackbone=Qwen2.5-3B, Method Category=Fine-Tuning2025.09 | 33.78 | |
| BPOBackbone=Qwen2.5-7B, Method Category=Prompt Optimization2025.09 | 32.67 | |
| RLPromptBackbone=Qwen2.5-7B, Method Category=Prompt Optimization2025.09 | 31.33 | |
| SFTBackbone=Qwen2.5-3B, Method Category=Fine-Tuning2025.09 | 29.33 | |
| CFPOBackbone=Qwen2.5-7B, Method Category=Prompt Optimization2025.09 | 27.11 | |
| BPOBackbone=Qwen2.5-3B, Method Category=Prompt Optimization2025.09 | 26.44 | |
| RLPromptBackbone=Qwen2.5-3B, Method Category=Prompt Optimization2025.09 | 22.67 | |
| OPROBackbone=Qwen2.5-7B, Method Category=Prompt Optimization2025.09 | 22 | |
| OPROBackbone=Qwen2.5-3B, Method Category=Prompt Optimization2025.09 | 21 | |
| Qwen2.5Backbone=Qwen2.5-7B, Method Category=Vanilla2025.09 | 18.44 | |
| CFPOBackbone=Qwen2.5-3B, Method Category=Prompt Optimization2025.09 | 17.78 | |
| Qwen2.5Backbone=Qwen2.5-3B, Method Category=Vanilla2025.09 | 17.11 |