Mathematical Computation on MathQA
52.34Exact Match (EM)Prompt-R1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Prompt-R12025.11 | 52.34 | 61.59 | |
| CoT ReasoningBackbone=GPT-4o-mini2025.11 | 49.22 | 57.03 | |
| GRPOBackbone=Qwen3-4B2025.11 | 46.88 | 54.43 | |
| BaselineBackbone=GPT-4o-mini2025.11 | 46.09 | 54.04 | |
| TextGradCategory=APO (GPT-4o-mini)2025.11 | 44.53 | 61.46 | |
| OPROCategory=APO (GPT-4o-mini)2025.11 | 43.75 | 60.08 | |
| GEPACategory=APO (GPT-4o-mini)2025.11 | 40.63 | 61.59 | |
| BaselineBackbone=Qwen3-4B2025.11 | 28.91 | 32.29 | |
| CoT ReasoningBackbone=Qwen3-4B2025.11 | 27.34 | 30.6 | |
| SFTBackbone=Qwen3-4B2025.11 | 17.97 | 22.66 |