TAAFT
Free mode
100% free
Freemium
Free Trial
Deals
Create tool
Glossary
Machine Learning
Preference Tuning

Preference Tuning

[ˈprɛfərəns ˈtuːnɪŋ]
Machine Learning
Last updated: April 4, 2025

Definition

Fine-tuning methods (like DPO, RLHF) aligning models with human preferences using preferred vs. non-preferred output data.

Detailed Explanation

Fine-tuning methods, like DPO or RLHF, that align model behavior with human preferences by training on data indicating preferred versus less preferred outputs.

Use Cases

Aligning LLMs to be more helpful, honest, and harmless; improving chatbot quality based on user feedback; making AI outputs more desirable or useful to humans.

Related Terms

Deep Q-Networks

A combination of Q-learning with deep neural networks to handle high-dimensional state spaces.

Machine Learning

Synthetic Data Generation

The creation of artificial data that mimics the statistical properties and patterns of real data.

Machine Learning

Mean Absolute Error

The average of the absolute differences between predicted and actual values. Treats all errors equally regardless of direction.

Machine Learning