Mathematical Computation on GSM8K (EM, F1)
97.66EMPrompt-R1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Prompt-R1Backbone=GPT-4o-mini2025.11 | 97.66 | 97.66 | |
| GRPOBackbone=Qwen3-4B2025.11 | 92.97 | 92.97 | |
| GEPAOptimization Framework=APO, Backbone=GPT-4o-mini2025.11 | 87.5 | 90.1 | |
| BaselineBackbone=Qwen3-4B2025.11 | 84.38 | 84.38 | |
| CoT ReasoningBackbone=GPT-4o-mini2025.11 | 84.38 | 88.02 | |
| BaselineBackbone=GPT-4o-mini2025.11 | 83.59 | 86.72 | |
| CoT ReasoningBackbone=Qwen3-4B2025.11 | 82.81 | 82.81 | |
| TextGradOptimization Framework=APO, Backbone=GPT-4o-mini2025.11 | 70.31 | 85.99 | |
| OPROOptimization Framework=APO, Backbone=GPT-4o-mini2025.11 | 63.28 | 83.65 | |
| SFTBackbone=Qwen3-4B2025.11 | 32.03 | 32.03 |