AI Paper Atlas

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

I Am Big, You Are Little; I Am Right, You Are Wrong

David A. Kelly, Akchunya Chanchal, Nathan Blake

ICCV 2025 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 4 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
classification is incorrect rather than correctMPS size across models · ImageNet-1k validationMPS size increase (%): +2.6Hurt
classification is incorrect rather than correctMPS size across models · Caltech-256MPS size increase (%): +3.3Hurt
use ConvNext-V2 Huge v1 instead of Inception V3model choice · ImageNet-1kaverage MPS area: 0.239 -> 0.05Improved
use EVA Giant instead of Inception V4model choice · ImageNet-1kaverage MPS area: 0.23 -> 0.054Improved
use ResNet152-D instead of ViT Huge V1model choice · ImageNet-1kaverage MPS area: 0.156 -> 0.134Improved
compare architectures with Kruskal-Wallis testMPS size · ImageNet-1kH statistic: H(4) = 1176.134, p < 0.001Improved
compare models within the Inception family (Friedman test)MPS size · ImageNet-1kp-value: p=0.36No clear effect
compare models within ConvNext, EVA, ResNet, ViT (Friedman test)MPS size · ImageNet-1kp-value: p < 0.01Improved
use larger pretrained models (ViT Large vs ViT Huge V1)MPS area · ImageNet-1kaverage MPS area: 0.099 vs 0.156Mixed

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