Preference Labeling on Summarization
89Preference Labeling AccuracyCurriculum-RLAIF
Evaluation Results
| Method | Links | |
|---|---|---|
| Curriculum-RLAIFBase Model=LLaMA-3-8B2025.05 | 89 | |
| Conventional RLAIFBase Model=LLaMA-3-8B2025.05 | 82 | |
| Curriculum-RLAIFBase Model=Gemma-1-2B2025.05 | 79 | |
| RLCDBase Model=LLaMA-3-8B2025.05 | 78 | |
| RLCDBase Model=Gemma-1-2B2025.05 | 72 | |
| Conventional RLAIFBase Model=Gemma-1-2B2025.05 | 71 | |
| CAIBase Model=LLaMA-3-8B2025.05 | 70 | |
| CAIBase Model=Gemma-1-2B2025.05 | 67 |