Personalized Distractor Generation on Eedi 100 Student 1361
32.1AccuracyMCTS-guided reasoning reconstruction framework
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MCTS-guided reasoning reconstruction frameworkBackbone=Deepseek-V32025.08 | 32.1 | 3.56 | 2.98 | |
| MCTS-guided reasoning reconstruction frameworkBackbone=GPT-4o2025.08 | 28.2 | 3.17 | 2.64 | |
| CoTBackbone=Deepseek-V32025.08 | 27.1 | 3.35 | 2.83 | |
| MCTS-guided reasoning reconstruction frameworkBackbone=Claude-4-Sonnet2025.08 | 27 | 3.72 | 3.31 | |
| IOBackbone=Deepseek-V32025.08 | 26.4 | 3.29 | 2.83 | |
| RandomBackbone=Deepseek-V32025.08 | 25.8 | 3.1 | 2.63 | |
| MasteryBackbone=Deepseek-V32025.08 | 25.5 | 3.19 | 2.7 | |
| LevelBackbone=GPT-4o2025.08 | 25.4 | 2.67 | 2.39 | |
| MasteryBackbone=GPT-4o2025.08 | 25.4 | 3.06 | 2.61 | |
| CoTBackbone=Claude-4-Sonnet2025.08 | 25.3 | 3.52 | 3.12 | |
| LevelBackbone=Deepseek-V32025.08 | 25.1 | 2.8 | 2.5 | |
| Level+RandomBackbone=Deepseek-V32025.08 | 24.8 | 2.75 | 2.44 | |
| Level+SimilarityBackbone=Deepseek-V32025.08 | 24.7 | 2.75 | 2.44 | |
| Level+RandomBackbone=GPT-4o2025.08 | 24.7 | 2.65 | 2.34 | |
| IOBackbone=GPT-4o2025.08 | 24.4 | 3 | 2.53 | |
| SimilarityBackbone=Deepseek-V32025.08 | 24.2 | 3.15 | 2.71 | |
| IOBackbone=Claude-4-Sonnet2025.08 | 24.1 | 3.51 | 3.1 | |
| Level+RandomBackbone=Claude-4-Sonnet2025.08 | 24 | 2.88 | 2.63 | |
| Level+SimilarityBackbone=GPT-4o2025.08 | 23.9 | 2.63 | 2.33 | |
| Level+SimilarityBackbone=Claude-4-Sonnet2025.08 | 23.7 | 2.83 | 2.56 | |
| CoTBackbone=GPT-4o2025.08 | 23.7 | 3 | 2.54 | |
| MasteryBackbone=Claude-4-Sonnet2025.08 | 23.5 | 3.34 | 2.82 | |
| LevelBackbone=Claude-4-Sonnet2025.08 | 23 | 2.92 | 2.66 | |
| SimilarityBackbone=GPT-4o2025.08 | 22.9 | 2.95 | 2.55 | |
| SimilarityBackbone=Claude-4-Sonnet2025.08 | 22.8 | 3.44 | 3.09 | |
| RandomBackbone=Claude-4-Sonnet2025.08 | 22.3 | 3.36 | 2.96 | |
| MCTS-guided reasoning reconstruction frameworkBackbone=GPT-3.5-turbo2025.08 | 22.1 | 3.12 | 2.53 | |
| RandomBackbone=GPT-4o2025.08 | 18.4 | 2.71 | 2.32 | |
| MCTS-guided reasoning reconstruction frameworkBackbone=LLaMA-3-70B2025.08 | 17.7 | 2.93 | 2.45 | |
| IOBackbone=GPT-3.5-turbo2025.08 | 17.4 | 2.73 | 2.25 | |
| LevelBackbone=GPT-3.5-turbo2025.08 | 16.6 | 2.39 | 2.11 | |
| CoTBackbone=LLaMA-3-70B2025.08 | 16.2 | 2.74 | 2.27 | |
| LevelBackbone=LLaMA-3-70B2025.08 | 15.7 | 2.5 | 2.22 | |
| MasteryBackbone=GPT-3.5-turbo2025.08 | 15.5 | 2.74 | 2.28 | |
| CoTBackbone=GPT-3.5-turbo2025.08 | 14.9 | 2.73 | 2.25 | |
| IOBackbone=LLaMA-3-70B2025.08 | 14.2 | 2.74 | 2.3 | |
| MasteryBackbone=LLaMA-3-70B2025.08 | 14 | 2.68 | 2.21 | |
| Level+RandomBackbone=LLaMA-3-70B2025.08 | 13.7 | 2.43 | 2.17 | |
| Level+SimilarityBackbone=LLaMA-3-70B2025.08 | 13.7 | 2.43 | 2.19 | |
| Level+SimilarityBackbone=GPT-3.5-turbo2025.08 | 13.1 | 2.16 | 1.88 | |
| Level+RandomBackbone=GPT-3.5-turbo2025.08 | 12.9 | 2.16 | 1.9 | |
| SimilarityBackbone=LLaMA-3-70B2025.08 | 12.7 | 2.8 | 2.43 | |
| SimilarityBackbone=GPT-3.5-turbo2025.08 | 12.2 | 2.62 | 2.22 | |
| RandomBackbone=GPT-3.5-turbo2025.08 | 11 | 2.56 | 2.13 | |
| RandomBackbone=LLaMA-3-70B2025.08 | 10.2 | 2.66 | 2.31 |