AI Paper Atlas

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

Just Y-Prediction: Enabling Historical Cumulative Inconsistency in Label Diffusion for Learning with Noisy Label

Senyu Hou, Gaoxia Jiang, Xinyi Zheng, Yaqing Guo, Shuna Liang, Wenjian Wang

ICML 2026 · Label-efficient and robust learning › Noisy labels · approach: Noise-aware loss and pairwise methods

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 pre-correction (ablation)JYP · CIFAR-10 Sym. 50%accuracy: 97.01 -> 97.35Improved
add dual-view sample selection (ablation)JYP · CIFAR-10 Sym. 50%accuracy: 96.58 -> 97.35Improved
add HCI and dual-view on top of y-prediction baseline (ablation)JYP · CIFAR-10 Sym. 50%accuracy: 96.10 -> 97.35Improved
add pre-correction (ablation)JYP · CIFAR-10 IDN 40%accuracy: 97.26 -> 97.45Improved
add dual-view sample selection (ablation)JYP · CIFAR-10 IDN 40%accuracy: 96.66 -> 97.45Improved
add HCI and dual-view on top of y-prediction baseline (ablation)JYP · CIFAR-10 IDN 40%accuracy: 96.03 -> 97.45Improved
add all modules (pre-correction, dual-view, HCI) (ablation)JYP · CIFAR-10 Sym. 50%accuracy: 95.39 -> 97.35Improved
use y-prediction instead of epsilon-prediction (ablation)label diffusion model · CIFAR-10 (50% Sym., 40% IDN, Human)accuracyImproved
use y-prediction for one-step inference instead of epsilon-prediction (ablation)label diffusion model · CIFAR-10 noisyone-step test accuracyImproved
one-step inference (JYP) instead of iterative inference (LRAD)JYP · WebVisioninference time: 21s to 4sImproved
one-step inference (JYP) instead of iterative inference (LRAD)JYP · ILSVRC12inference time: 27s to 5sImproved
use dual-view strict partition in low-noise settingsJYP · CIFAR low-noiseaccuracyHurt

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