Currently, mainstream AI alignment methods such as Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO) rely on high-quality human preference feedback data.

We are in the middle of a huge boom in artificial intelligence (AI), with unprecedented investment in research, a supercharged pace of innovation and sky-high expectations. But what is driving this ...

UC Riverside researchers have created a certified unlearning method that removes sensitive or copyrighted data from AI models ...