Review Writing on LaMP Review Writing
35.3ROUGE-1C-BPO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| C-BPOTraining Paradigm=Preference Opt., Backbone=Mistral-7B-Instruct-v0.32026.05 | 35.3 | 15.4 | |
| CoPETraining Paradigm=Preference Opt., Backbone=Mistral-7B-Instruct-v0.32026.05 | 33.5 | 14.6 | |
| OPPUTraining Paradigm=SFT-based, Backbone=Mistral-7B-Instruct-v0.32026.05 | 31.9 | 13.4 | |
| BCOTraining Paradigm=Preference Opt., Backbone=Mistral-7B-Instruct-v0.32026.05 | 31.5 | 13.2 | |
| KTOTraining Paradigm=Preference Opt., Backbone=Mistral-7B-Instruct-v0.32026.05 | 29.8 | 12.6 | |
| TAMTraining Paradigm=SFT-based, Backbone=Mistral-7B-Instruct-v0.32026.05 | 28.9 | 12.2 | |
| Base ModelTraining Paradigm=Non-Tuned, Backbone=Mistral-7B-Instruct-v0.32026.05 | 28.7 | 12.6 | |
| RAGTraining Paradigm=Non-Tuned, Backbone=Mistral-7B-Instruct-v0.32026.05 | 27.2 | 12.8 | |
| PAGTraining Paradigm=Non-Tuned, Backbone=Mistral-7B-Instruct-v0.32026.05 | 25.6 | 12.5 |