Reinforcement Learning from Human Feedback

Information technology - software - natural language processing - artificial intelligence - chatbots - training.

(RLHF).

Reinforcement Learning from Human Feedback is a training process for machine learning.

It uses human feedback, or human preferences, to rank - or score - instances of the behaviour or output from the system being trained, for example ChatGPT.

The human-supervised RLHF supplements an initial period of unsupervised training known as generative pre-training.

See also

• Artificial intelligence (AI)
• Bot
• Chatbot
• ChatGPT
• Enterprise-wide resource planning system
• Generative pre-trained transformer (GPT)
• Google Gemini
• Information technology
• Machine learning
• Natural language
• Natural language processing
• Robotics
• Software
• Software robot

Other resource

• Improving Language Understanding by Generative Pre-Training, Radford, Narasimhan, Salimans & Sutskever, 2018