---
title: "On Calibration of Modern Neural Networks"
description: "A classifier can become more accurate while its probabilities become less honest. Follow reliability diagrams, expected calibration error, and the one learned temperature that softens confidence without changing a single predicted class."
canonical_url: "https://fanout.sh/daily/2026-11-06-calibration-temperature-scaling"
md_url: "https://fanout.sh/daily/2026-11-06-calibration-temperature-scaling.md"
last_updated: "2026-11-06"
access: "public"
---

# On Calibration of Modern Neural Networks

A classifier can become more accurate while its probabilities become less honest. Follow reliability diagrams, expected calibration error, and the one learned temperature that softens confidence without changing a single predicted class.

## Paper details

- Authors: Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q. Weinberger

- Venue: ICML 2017

- Track: AI Research

- Difficulty: Easy

- Reading time: 13 min

- Original paper: https://proceedings.mlr.press/v70/guo17a.html

## What you will learn

- Why accuracy and calibrated confidence answer different questions, and how a reliability diagram exposes the gap.

- How one validation-fitted temperature rescales every logit while preserving the model's class ranking and accuracy.

- What the paper's cross-dataset results establish, and why binning, distribution shift, and newer architectures limit the recipe.

## Continue exploring

[Compare Batch Normalization](https://fanout.sh/daily/2026-07-10-batch-normalization): Review the normalization method the calibration paper associates with lower error but worse confidence estimates in its controlled ConvNet comparison.

---
This representation contains public Fanout content only. Protected Pro lessons, account data, billing, checkout, and pricing are not included.

Browse the public content map: https://fanout.sh/sitemap.md