Preference Pair
زوج التفضيل
مثال تدريبي يتألف من موجِّه وإجابتين — إحداهما مُفضَّلة والأخرى مرفوضة — يُستخدم لتعليم النموذج تفضيلات البشر.
A training example consisting of a prompt and two responses — one preferred and one rejected — used to teach the model human preferences.
Also translated asزوج المقارنة البشرية، ثنائية التفضيل
First appears in this corpus in: Deep Reinforcement Learning from Human Preferences (2017)
Appears in these papers
- Deep Reinforcement Learning from Human Preferences2017in the sky ✦
- Direct Preference Optimization: Your Language Model Is Secretly a Reward Model2023in the sky ✦
- Direct Preference Optimization: Your Language Model Is Secretly a Reward Model2023in the sky ✦
- Learning to Summarize from Human Feedback2020in the sky ✦
- Learning to Summarize from Human Feedback2020in the sky ✦
- LIMA: Less Is More for Alignment2023in the sky ✦