Preference Tuning LLMs with Direct Preference Optimization Methods