Fetching the paper…

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