AI Dictionary › AI Fundamentals

RLHF

RLHF (Reinforcement Learning from Human Feedback)

RLHF is the training technique that aligns language models with human preferences. After pre-training on text, the model is refined like this: several answers are generated for the same question, human raters indicate which they prefer, and these preferences become the signal the model uses to learn to answer "the way people like".

Definition

It is the step that turns a raw text-completion model into a useful, polite, instruction-following assistant. ChatGPT owes much of its success precisely to RLHF, which made it conversational and safe.

RLHF also explains some typical model behaviors: the tendency to be verbose (raters rewarded thorough answers), excess caution on certain topics, and willingness to admit mistakes. Knowing this helps you understand "who" you're talking to when you write a prompt.

Related terms

• Reinforcement Learning
• Fine-tuning
• LLM (Large Language Model)
• Training Data
• System Prompt

More in AI Fundamentals

• LLM (Large Language Model)
• Token
• Context Window
• Temperature
• AI Hallucination
• Fine-tuning
• RAG (Retrieval Augmented Generation)
• Embedding

Put it into practice

• Real scenario: apply this concept with an instant AI verdict.
• Daily gym: short drills to train dictionary concepts.
• Playoff game: challenge other players, elimination-style.

📱 Download the Android app (beta) iOS coming soon

Say what you mean. Get what you need.

Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.