AI Paper Atlas

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

Conformal Prediction Meets Long-tail Classification

Shuqi Liu, Jianguo Huang, Luke Ong

AAAI 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Confidence calibration objectives · 5 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
add selective rank regularization on head classes (TACP) to standard CPAPS score · ImageNet-LT rho=0.6head-tail coverage gap CovGap-HT (%): 2.18 -> 1.11Improved
apply TACP to TOPK scoreTOPK · ImageNet-LT rho=0.6head-tail coverage gap CovGap-HT (%): 8.83 -> 1.02Improved
apply TACP to TOPK scoreTOPK · ImageNet-LT rho=0.6average set size: 23.15 -> 25.63Hurt
apply TACP instead of STANDARD (APS)APS · CIFAR100-LT mu=100head-tail coverage gap CovGap-HT (%): 4.81 -> 0.78Improved
TACP vs Partition-Wise calibrationLAC · ImageNet-LT rho=0.6head-tail coverage gap CovGap-HT (%): 1.96 -> 1.18Improved
replace hard head indicator with soft class-prior weighting (sTACP)APS · ImageNet-LT rho=0.6 (alpha=10%)class-conditional coverage gap CovGap (%): 19.00 -> 15.86Improved
use sTACP instead of ClassWiseAPS · ImageNet-LT rho=0.6 (alpha=5%)set size: 983.53 -> 76.83Improved
use sTACP at alpha=5% vs STANDARDAPS · ImageNet-LT rho=0.6 (alpha=5%)class-conditional coverage gap CovGap (%): 9.62 -> 8.08Improved
use sTACP instead of STANDARD at alpha=5% on set sizeAPS · ImageNet-LT rho=0.6 (alpha=5%)average set size: 67.49 -> 76.83Hurt

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