[QA] Preference Fine-Tuning of LLMs Should Leverage Suboptimal, On-Policy Data

Arxiv Papers - [QA] Preference Fine-Tuning of LLMs Should Leverage Suboptimal, On-Policy Data