AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Sample-Dependent Adaptive Temperature Scaling for Improved Calibration

Tom Joy, Francesco Pinto, Ser-Nam Lim, Philip H.S. Torr, Puneet K. Dokania

AAAI 2023 · Architectures and efficiency › RNNs and normalization · approach: Dataset distillation and data-efficient training · 28 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
Replace a single global temperature with a sample-dependent temperature predicted from VAE feature-space likelihoodsResNet50 trained with cross-entropy · CIFAR10ECE: 1.41 ± 0.43 -> 1.13 ± 0.60Improved
Replace a single global temperature with a sample-dependent temperatureWideResNet28-10 trained with cross-entropy · CIFAR100ECE: 3.76 ± 0.29 -> 2.95 ± 0.41Improved
Replace a single global temperature with a sample-dependent temperatureResNet50 trained with cross-entropy · Tiny-ImageNetECE: 7.64 ± 1.53 -> 4.51 ± 1.76Improved
Replace a single global temperature with a sample-dependent temperatureResNet50 trained with cross-entropy · CIFAR10-CECE: 12.72 ± 0.64 -> 10.86 ± 1.88Improved
Replace a single global temperature with a sample-dependent temperatureWideResNet28-10 trained with cross-entropy · CIFAR100-CECE: 12.28 ± 1.14 -> 12.17 ± 0.10Improved
Replace a single global temperature with a sample-dependent temperatureResNet50 · CIFAR-10.1ECE: 4.57 ± 0.32 -> 3.67 ± 1.41Improved
Apply adaptive temperature scaling instead of vanilla temperature scalingResNet50 · CIFAR10accuracy: 93.13 ± 1.97 -> 93.13 ± 1.97No clear effect
Apply adaptive temperature scaling instead of vanilla temperature scalingResNet50 · CIFAR100AURRA-C: 93.62 ± 0.06 -> 94.03 ± 0.18Improved
Apply adaptive temperature scaling instead of vanilla temperature scalingResNet50 · Tiny-ImageNetAURRA-C: 79.58 ± 2.06 -> 81.26 ± 0.49Improved
Use per-sample temperatures rather than one global temperatureCIFAR100-C models · CIFAR100-CECE: 12.28 ± 1.14 -> 12.17 ± 0.10; 11.78 ± 0.91 -> 11.69 ± 0.74Mixed

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