AI Glossary

Preference dataset

What does “Preference dataset” mean?

Preference dataset: A collection of comparisons or ratings indicating which outputs people or models prefer.

Preference dataset visual guide: A collection of comparisons or ratings indicating which outputs people or models prefer.
Visual guide Open full size (new tab)

What is an example of “Preference dataset”?

Reviewers choose the clearer of two responses to the same question.

Keep learning

See also

Fine-tuningTermInstruction tuningTermKnowledge distillationTermNext-token predictionTermSynthetic supervisionTerm

Same topic

Proximal policy optimizationTermPre-trainingTermQ-learningTermPooling layerTermQLoRATerm