AI Glossary

Model calibration

Also called probability calibration

What does “Model calibration” mean?

Model calibration: How closely a model's confidence matches the frequency with which it is correct.

Model calibration visual guide: How closely a model's confidence matches the frequency with which it is correct.
Visual guide Open full size (new tab)

What is an example of “Model calibration”?

Among predictions assigned 80 percent confidence, about 80 percent should be right in a well-calibrated group.

Keep learning

See also

AlignmentTermAI safetyTermAgent guardrailsTermConfidence scoreTermAbstentionTerm

Same topic

Model cardTermMMLUTermModel extractionTermMembership inferenceTermModel monitoringTerm