Direct Preference Optimization. An alternative to RLHF that skips the separate reward model step. Instead of training a reward model and then doing reinforcement learning, DPO directly optimizes the language model on human preference data. Simpler, cheaper, and increasingly popular for alignment.
Reinforcement Learning from Human Feedback.
The process of taking a pre-trained model and continuing to train it on a smaller, specific dataset to adapt it for a particular task or domain.
A mathematical function applied to a neuron's output that introduces non-linearity into the network.
An optimization algorithm that combines the best parts of two other methods — AdaGrad and RMSProp.
Artificial General Intelligence.
An autonomous AI system that can perceive its environment, make decisions, and take actions to achieve goals.
Browse our complete glossary or subscribe to our newsletter for the latest AI news and insights.