AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

H-Calibration: Rethinking Classifier Recalibration With Probabilistic Error-Bounded Objective

Wenjian Huang, Guiping Cao, Jiahao Xia, Jingkun Chen, Hao Wang, Jianguo Zhang

TPAMI 2025 · Label-efficient and robust learning › Recognition and long-tail · approach: Confidence calibration objectives · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
recalibrate with h-calibration instead of no calibrationResNet110 · CIFAR100ECE (top-label): 0.1848 -> 0.0118Improved
use h-calibration instead of temperature scalingResNet110 · CIFAR100ECE (top-label): 0.0187 -> 0.0118Improved
use h-calibration instead of Spline calibrationResNet101pre on CARS · CARSECE (top-label): 0.0109 -> 0.0052Improved
use h-calibration instead of Spline calibrationResNet50(NTS) on BIRDS · BIRDSECE (top-label): 0.0103 -> 0.0096Improved
use h-calibration instead of SoftBinResNet110 · CIFAR10ECE (top-label): 0.0030 -> 0.0043Hurt
use h-calibration instead of ScaBinDenseNet40 · CIFAR10ECE (top-label): 0.0020 -> 0.0073Hurt
use h-calibration instead of TS+LECEWideResNet32 · CIFAR100ECE (top-label): 0.0088 -> 0.0149Hurt
use h-calibration instead of GPWideResNet32 · CIFAR10ECE (top-label): 0.0013 -> 0.0015Mixed
use h-calibration instead of Spline calibrationResNet101 · CARSECE (top-label): 0.0050 -> 0.0082Hurt
use h-calibration instead of temperature scalingResNet101pre · CARSECE (top-label): 0.0297 -> 0.0052Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy