AI Paper Atlas

Research map / Label-efficient and robust learning / Noisy labels

ProMeCD: Unifying Long-Tailed and Noisy Label Learning via White-Box Control

Yu Zhu

ICML 2026 · Label-efficient and robust learning › Noisy labels · approach: Noisy cross-modal hashing and multi-label

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
use ProMeCD white-box PI control instead of FMW-NetResNet-32 · CIFAR-100-LT IF=100 (100 epochs)top-1 accuracy: 42.32 -> 47.52 (+5.2%)Improved
train ProMeCD for 200 instead of 100 epochsProMeCD · CIFAR-100-LT IF=100top-1 accuracy: 47.52 -> 53.12Improved
use ProMeCD instead of PaCoResNet-32 · CIFAR-100-LT IF=100top-1 accuracy: 52.00 -> 53.12 (+1.12%)Improved
use ProMeCD instead of standard CEResNet-50 · iNaturalist 2018top-1 accuracy: 57.16 -> 72.94Improved
use ProMeCD instead of PaCo on iNaturalistResNet-50 · iNaturalist 2018top-1 accuracy: 73.20 -> 72.94Hurt
use ProMeCD instead of FMW-Net under flip noiseResNet-32 · CIFAR-10-N Flip 40%test accuracy: 68.80 -> 91.12Improved
use ProMeCD instead of FMW-Net (DBG)ResNet-50 · mini WebVisiontop-1 accuracy: 72.44 -> 73.28Improved
use ProMeCD instead of D-SINKResNet-18 · LTNLL CIFAR-100 IF=100 NR=60% (Sym.)test accuracy: 30.42 -> 32.88Improved
use ProMeCD instead of D-SINKResNet-18 · LTNLL CIFAR-10 IF=100 NR=60% (Sym.)test accuracy: 57.80 -> 60.15Improved
use ProMeCD instead of MW-NetResNet-32 · CIFAR-100-N Sym. 40%test accuracy: 45.79 -> 65.41Improved

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