Standard Question Answering on SQuAD v2
19.53EMPrompt-R1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Prompt-R12025.11 | 19.53 | 29.28 | |
| CoT ReasoningBackbone=GPT-4o-mini2025.11 | 14.06 | 25.73 | |
| BaselineBackbone=GPT-4o-mini2025.11 | 13.28 | 25.61 | |
| GEPACategory=APO (GPT-4o-mini)2025.11 | 13.28 | 25.52 | |
| OPROCategory=APO (GPT-4o-mini)2025.11 | 10.94 | 26.67 | |
| GRPOBackbone=Qwen3-4B2025.11 | 10.16 | 23.1 | |
| BaselineBackbone=Qwen3-4B2025.11 | 6.25 | 16.09 | |
| CoT ReasoningBackbone=Qwen3-4B2025.11 | 6.25 | 16.25 | |
| TextGradCategory=APO (GPT-4o-mini)2025.11 | 6.25 | 22.04 | |
| SFTBackbone=Qwen3-4B2025.11 | 5.47 | 16.18 |