AI Paper Atlas

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

On the Reliability of Cue Conflict and Beyond

Pum Jun Kim, Seung-Ah Lee, Seongho Park, Dongyoon Han, Jaejun Yoo

ECCV 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 1 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
evaluate with REFINED-BIAS instead of cue-conflict stylized imagesshape-augmented ResNet-50 · ResNet-50 Shape Augt-test p-value on preference: p=0.181 -> p=0.018Improved
evaluate with REFINED-BIAS instead of cue-conflictcontrastive ResNet-50 · ResNet-50 Contrastivet-test p-value on preference: p=0.684 -> p=0.009Improved
evaluate with REFINED-BIAS instead of cue-conflictadversarially trained ResNet-50 · ResNet-50 Adversarialt-test p-value on preference: p=5.19e-05 -> p=0.489Improved
evaluate with REFINED-BIAS instead of cue-conflictmixed-augmentation ResNet-50 · ResNet-50 Mixed Augpreference direction: p=2.72e-04 -> p=0.002Improved
use ambiguous non-distinctive classes in REFINED-BIAS (ablation)Shape Aug ResNet-50 · ResNet-50 Shape Augt-test p-value on preference: p=0.018 -> p=0.919Hurt
use ambiguous non-distinctive classes in REFINED-BIAS (ablation)Adversarial ResNet-50 · ResNet-50 Adversarialt-test p-value on preference: p=0.489 -> p=0.006Hurt
replace stylized cues with pure curated cues (human agreement)texture cues · Texture cuesFleiss kappa: 0.2937 -> 0.7973Improved
replace stylized cues with pure curated cues (shape)shape cues · Shape cuesFleiss kappa: 0.7276 -> 0.9836Improved
use REFINED-BIAS cues for CNN recognitionImageNet-1k CNNs · shape cuesmean top-1 accuracy: 4 -> 46Improved
remove human curation from dataset generation (ablation)REFINED-BIAS · learning strategiesshape preference significanceNo clear effect
use Swin or CMT instead of ViTViT · REFINED-BIASshape sensitivityImproved
increase quantization strength from 32-bit to 2-bitResNets · REFINED-BIASshape and texture sensitivityHurt

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