ModelRefs / Calibration — AI Glossary

Calibration — AI Glossary

The alignment between a model's expressed confidence (probability) and the actual likelihood of correctness.

Overview

A well-calibrated model saying 70% confidence is correct 70% of the time. LLMs are often overconfident (high confidence on wrong answers) or underconfident post-RLHF. Calibration is measured by ECE (Expected Calibration Error) and reliability diagrams. Critical for applications using logprobs as confidence signals.

Reference details

Topic	prompting
Last reviewed	2026-06-24

Example: What well-calibrated actually means

Take every answer where the model reported 70% confidence. If it is well calibrated, roughly 70 of each 100 are correct. If 95 are correct it is underconfident; if 40 are, it is overconfident and any threshold you set on that number is meaningless. Calibration is a property of the distribution, never of a single answer.

Commonly confused with

Confidence is not accuracy, and a model stating “I am 95% sure” in prose is not reporting a probability — it is generating plausible text. Token logprobs are a genuine signal; a sentence about confidence is not, and the two are routinely confused.

When to use it

Reach for it when:

You route on confidence — auto-approve above a threshold, escalate below
Abstention matters more than coverage
You have enough labelled outcomes to measure calibration honestly

Reach for something else when:

Trusting stated confidence in prose as a probability
On a model you have not measured; RLHF commonly degrades calibration
Where the cost of a confident error is high and nothing else checks it

Continue your research

Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Calibration — AI Glossary.

Frequently asked questions

What is Calibration?