AI Paper Atlas

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

Shortcut-Resistant CAM Distillation for Long-Tailed Recognition

Wenhai Wan, Teng Zhang, Shao-Yuan Li, Xinrui Wang, Qiang-Sheng Hua, Songcan Chen

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Decoupled and calibrated long-tail training

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 SRCD to BCLBCL · CIFAR-100-LT IR=100 (from scratch)top-1 accuracy (%): 51.76 -> 53.64Improved
add SRCD to BCLBCL · CIFAR-10-LT IR=100 (from scratch)top-1 accuracy (%): 83.61 -> 82.97Hurt
add SRCD to GCLGCL · CIFAR-100-LT IR=10 (from scratch)top-1 accuracy (%): 61.79 -> 60.23Hurt
add SRCD to CE+H2TCE+H2T · CIFAR-100-LT IR=50 (from scratch)top-1 accuracy (%): 47.58 -> 46.97Hurt
add SRCD to GLMCGLMC · ImageNet-LT Tail (from scratch)top-1 accuracy (%): 30.41 -> 34.22Improved
add SRCD to LIFT (CLIP fine-tuning)LIFT · ImageNet-LT Alltop-1 accuracy (%): 77.1 -> 78.4Improved
add SRCD to LIFT (CLIP fine-tuning)LIFT · iNaturalist Tailtop-1 accuracy (%): 81.5 -> 80.8Hurt
use both coherence and concentration cues in teacher weighting instead of one (ablation)CE+FCC+SRCD · ImageNet-LT Alltop-1 accuracy (%): 54.02 -> 55.95Improved
use energy-based weighting instead of simple averaging of head CAMs (ablation)SRCD teacher · CAM similarity to balanced referenceCAM cosine similarityImproved
increase number of head proxies K beyond 3 (ablation)SRCD · long-tailed benchmarkaccuracyNo clear effect
increase distillation strength beta above 0.6 (ablation)SRCD · CIFAR-100-LTaccuracyImproved
replace head-proxy teacher with fixed center mask (ablation)SRCD · long-tailed benchmarkaccuracyHurt

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