AI Glossary Term
What does “Human Preference Model” mean?
Definition
A model trained on human rankings to predict which responses users are likely to prefer.
How Human Preference Model works
Example
From thousands of comparisons in which people chose clearer, more polite answers, a model learns to score new answers. That score then guides further tuning of a chatbot.
Why it matters
It captures qualities that are hard to define in a formula. It reflects the people who gave the ratings and can be exploited.
Common questions
Same as a reward model?
Often the same thing in practice. Human preference model emphasizes what it was trained on.
Who provides the rankings?
Paid raters, volunteers, or users, with guidelines to improve consistency.
What is the weakness?
Raters disagree and may favor longer or more confident-sounding answers.