Preference Fine-Tuning of LLMs Should Leverage Suboptimal, On-Policy Data 논문리뷰

https://arxiv.org/abs/2404.14367