AI Paper Atlas

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

Boosting Class Representation via Semantically Related Instances for Robust Long-Tailed Learning with Noisy Labels

Yuhang Li, Zhuying Li, Yuheng Jia

ICCV 2025 · Label-efficient and robust learning › Noisy labels · approach: Noise-aware loss and pairwise methods · 3 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 soft labels (ablation)IBC · CIFAR-10 (noise 0.5, IR 100)overall top-1 accuracy: 74.16 -> 78.17Improved
add soft labels (ablation)IBC · CIFAR-100 (noise 0.5, IR 100)overall top-1 accuracy: 38.82 -> 41.96Improved
add instance-prototype similarity selection (ablation)IBC · CIFAR-100 (noise 0.5, IR 100)overall top-1 accuracy: 39.61 -> 41.96Improved
select semantically related classes instead of random classes (ablation)IBC · CIFAR-10 (noise 0.5, IR 100)overall top-1 accuracy: 76.34 -> 78.17Improved
select semantically related classes instead of random classes (ablation)IBC · CIFAR-100 (noise 0.5, IR 100)few-shot top-1 accuracy: 16.78 -> 19.53Improved
add shot-specific multi-expert ensemble (ablation)IBC · CIFAR-10 (noise 0.5, IR 100)overall top-1 accuracy: 76.68 -> 78.17Improved
add shot-specific multi-expert ensemble (ablation)IBC · CIFAR-100 (noise 0.5, IR 100)few-shot top-1 accuracy: 20.02 -> 19.53Hurt
add MixMatch semi-supervised learning on noisy instances (ablation)IBC · CIFAR-100 (noise 0.5, IR 100)overall top-1 accuracy: 39.70 -> 41.96Improved
enhance tail only with tail soft labels (vs full IBC) (ablation)IBC · CIFAR-100 (noise 0.5, IR 100)few-shot / many-shot accuracy: +0.84% few-shot, -7.14% many-shot, -1.20% medium-shotMixed
use IBC (semantically related classes) instead of label smoothing / random classes at k=30 (ablation)IBC · CIFAR-100 (noise 0.5, IR 100)overall / tail accuracy: +1.10% overall, +2.10% tailImproved
IBC vs previous SOTAIBC · simulated CIFAR-10/100top-1 accuracy: +1.89% to +4.99%Improved
IBC vs previous SOTAIBC · Clothing1Mtop-1 accuracy: 74.97 -> 76.15Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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