AI Paper Atlas

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

Rank-Guided Pseudo-Bias Learning for Robust Black-Box Adaptation

Rajeev Ranjan Dwivedi, Anshuman Dangwal, Vinod K Kurmi

CVPR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Prototype and concept dictionary models

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 supervised contrastive loss with pseudo-bias labels to CE (ablation)adapter on frozen ResNet-18 · Waterbirdsworst-group accuracy: 38.90 -> 66.67Improved
add contrastive loss to CE (ablation)adapter on frozen ResNet-18 · CMNIST-0.9bias-conflicting accuracy: 61.72 -> 84.56Improved
add contrastive loss to CE (ablation)adapter on frozen ResNet-18 · CelebAworst-group accuracy: 27.20 -> 64.44Improved
add contrastive loss to CE (ablation)adapter on frozen ResNet-18 · CMNIST-0.995bias-conflicting accuracy: 49.09 -> 52.73Improved
add contrastive loss on top of cluster margin (CM) (ablation)CM baseline · Waterbirdsworst-group accuracy: 80.29 -> 83.01Improved
add contrastive loss on top of cluster margin (CM) (ablation)CM baseline · CelebAworst-group accuracy: 80.79 -> 83.10Improved
use oracle true bias labels instead of pseudo labels (ablation)PLD-Debias · Waterbirdsworst-group accuracy: 83.02 -> 85.81Improved
use oracle true bias labels instead of pseudo labels (ablation)PLD-Debias · CelebAworst-group accuracy: 83.11 -> 83.33No clear effect
set contrastive weight to 0.01 rather than 0 (ablation)PLD-Debias · Waterbirds (ResNet-18)worst-group accuracy: 80.85 -> 83.02Improved
increase contrastive weight to 1 or more (ablation)PLD-Debias · Waterbirds (ResNet-18)worst-group accuracy: 83.02 -> 66.04 (lambda 0.01 -> 1); 52.59 at lambda 10Hurt
lower spurious correlation strength (bias ratio 0.8 -> 0.5) with pseudo labelsPLD-Debias · CMNIST ResNet-18pseudo-bias label accuracy: 89.64 -> 60.14Hurt
use PLD-Debias at weak bias ratio 0.5 versus CMPLD-Debias · CMNIST-0.5 ResNet-18bias-conflicting accuracy: 84.54 -> 87.16Improved

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