Difficulty-controllable Question Generation on RACE (test)
2.18Estimated DifficultyQG_dpo
Evaluation Results
| Method | Links | |
|---|---|---|
| QG_dpoSpecified Difficulty=2.0, Approach=Proposed Method2025.10 | 2.18 | |
| QG_sftSpecified Difficulty=2.0, Approach=Proposed Method2025.10 | 1.25 | |
| QG_dpoSpecified Difficulty=0.0, Approach=Proposed Method2025.10 | 0.32 | |
| QG_sftSpecified Difficulty=0.0, Approach=Proposed Method2025.10 | 0.03 | |
| GPT-4.1 miniSpecified Difficulty=2.0, Approach=Few-Shot Learning, Number of shots=30-shot2025.10 | -1.21 | |
| GPT-4.1 miniSpecified Difficulty=-2.0, Approach=Few-Shot Learning, Number of shots=30-shot2025.10 | -1.22 | |
| QG_sftSpecified Difficulty=-2.0, Approach=Proposed Method2025.10 | -1.25 | |
| GPT-4.1 miniSpecified Difficulty=2.0, Approach=Few-Shot Learning, Number of shots=3-shot2025.10 | -1.29 | |
| GPT-4.1 miniSpecified Difficulty=2.0, Approach=Few-Shot Learning, Number of shots=10-shot2025.10 | -1.33 | |
| GPT-4.1 miniSpecified Difficulty=0.0, Approach=Few-Shot Learning, Number of shots=10-shot2025.10 | -1.55 | |
| GPT-4.1 miniSpecified Difficulty=0.0, Approach=Few-Shot Learning, Number of shots=30-shot2025.10 | -1.55 | |
| GPT-4.1 miniSpecified Difficulty=-2.0, Approach=Few-Shot Learning, Number of shots=10-shot2025.10 | -1.56 | |
| GPT-4.1 miniSpecified Difficulty=-2.0, Approach=Few-Shot Learning, Number of shots=3-shot2025.10 | -1.6 | |
| GPT-4.1 miniSpecified Difficulty=0.0, Approach=Few-Shot Learning, Number of shots=3-shot2025.10 | -1.71 | |
| QG_dpoSpecified Difficulty=-2.0, Approach=Proposed Method2025.10 | -1.89 |