“Direct Preference Optimization, a simpler alternative to PPO used in Llama models.”

“Direct Preference Optimization, discussed as a simpler alternative to PPO.”

“you can do DPO which is a form of supervised learning.”